Salvo User Manual

User Manual:

Open the PDF directly: View PDF .
Page Count: 529

Download
Open PDF In Browser	View PDF

(inside front cover)

User Manual
version 4.2.2

for all distributions

Quick Start Guide
Thanks for purchasing Salvo, The RTOS that runs in tiny places.™
Pumpkin is dedicated to providing powerful, efficient and low-cost
embedded programming solutions. We hope you'll like what we've
made for you.
If this is the first time you've encountered Salvo, please review
Chapter 1 • Introduction to get a flavor for what Salvo is, what it
can do, and what other tools you'll need to use it successfully. See
Chapter 2 • RTOS Fundamentals if you haven't used an RTOS
before. Then try the steps below in the order listed.

Note You don't need to purchase Salvo to run the demo programs, try the tutorial or use the freeware libraries to build your
own multitasking Salvo application – they're all part of Salvo Lite,
the freeware version of Salvo.

Running on Your Hardware
If you have a compatible target environment, you can run one of
the standalone Salvo example applications contained in Pumpkin\Salvo\Example on your own hardware. Open the demo's project, build it, download or program it into your hardware, and let it
run. Most demo programs provide real-time feedback. If it's a
Salvo Lite demo and uses commonly available hardware, you can
even build your own application by modifying the source and rebuilding it.
See Appendix C • File and Program Descriptions for more information on the demo programs.

Trying the Tutorial
Chapter 4 • Tutorial builds a multitasking, event-driven Salvo ap-

plication in six easy steps. The tutorial will familiarize you with
Salvo's terminology, user services, and the process of building a
working application. A set of tutorial projects is included with
every Salvo distribution for embedded targets, enabling you to
build each tutorial application by simply loading and building the
project in the appropriate development environment.

Salvo Lite
A compiler that's certified for use with Salvo is all you need to use
Salvo Lite, the freeware version of Salvo. You can write your own,
small multitasking application with calls to Salvo services and link
it to the freeware libraries. See Chapter 4 • Tutorial and the Salvo
Application Note for your compiler and/or target for more information.

Salvo LE
Salvo LE adds the standard Salvo libraries to Salvo Lite. This
means that the numbers of tasks, events, etc. in your application
are limited only by the available RAM.

Salvo Pro
With Salvo Pro, you'll have full access to all its source code, standard libraries, test programs and priority support. If you haven't
done so already, try the tutorial in Chapter 4 • Tutorial as a first
step towards creating your own application. Then use the configuration options in Chapter 5 • Configuration and the services outlined in Chapter 7 • Reference, along with their examples, to finetune Salvo to your application's requirements. If you run into problems or have questions, you'll find lots of useful information in
Chapter 6 • Frequently Asked Questions (FAQ) and Chapter 11 •
Tips, Tricks and Troubleshooting.

Getting Help
Some of the best resources for new and experienced Salvo users
are the Salvo User Forums, hosted on Pumpkin's web site,
http://www.pumpkininc.com/. Check there for up-to-date information on the latest Salvo releases.

Contact Information & Technical
Support

Contacting Pumpkin
Pumpkin's mailing address and phone and fax numbers are:
Pumpkin, Inc.
750 Naples Street
San Francisco, CA 94112 USA
tel: 415-584-6360
fax: 415-585-7948
info@pumpkininc.com
sales@pumpkininc.com
support@pumpkininc.com
Time Zone: GMT–0800 (Pacific Standard Time)

Connecting to Pumpkin's Web Site
Use your web browser to access the Pumpkin web site at
•

http://www.pumpkininc.com/

Information available on the web site includes
• Latest News
• Software Downloads & Upgrades
• User Manuals
• Compiler Reference Manuals
• Application Notes
• Assembly Guides
• Release Notes
• User Forums

Salvo User Forums
Pumpkin maintains User Forums for Salvo at Pumpkin's web site.
The forums contain a wealth of practical information on using
Salvo, and is visited by Salvo users as well as Pumpkin technical
support.

How to Contact Pumpkin for Support
Pumpkin provides online Salvo support via the Salvo Users Forums on the Pumpkin World Wide Web (WWW) site. Files and
information are available to all Salvo users via the web site. To
access the site, you'll need web access and a browser (e.g. Netscape, Opera, Internet Explorer).

Internet (WWW)
The Salvo User Forums are located at:
•

http://www.pumpkininc.com

and are the preferred method for you to post your pre-sales, general or technical support questions.

Email
Normally, we ask that you post your technical support questions to
the Salvo User Forums on our website. We monitor the forums and
answer technical support questions on-line.
In an emergency, you can reach technical support via email:
•

support@pumpkininc.com

We will make every effort to respond to your email requests for
technical support within 1 working day. Please be sure to provide
as much information about your problem as possible.

Mail, Phone & Fax
If you were unable to find an answer to your question in this manual, check the Pumpkin website and the Salvo user Forums (see
below) for additional information that may have been recently

posted. If you are still unable to resolve your questions, please contact us directly at the numbers above.

What To Provide when Requesting Support
Registered users requesting Salvo technical support should supply:
• The Salvo version number
• The compiler name and version number
• The user's source code snippet(s) in question
• The user's salvocfg.h file
• All other relevant files, details, etc.
Small code sections can be posted directly to the Salvo User Forums – see the on-line posting FAQ on how to use the UBB code
tags ([code] and [/code]) to preserve the code's formatting and
make it more legible.
If the need arises to send larger code sections, or even a complete,
buildable project, please compress the files and email them directly
to Salvo Technical support (see below). Please be sure to provide
all necessary files to enable Technical Support to build your Salvo
application locally in an attempt to solve your problem. Keep in
mind that without the appropriate target system hardware, support
in these cases is generally limited to non-runtime problem solving.
Technical Support will keep all user code in strictest confidence.

Salvo User Manual
Copyright © 1995-2010 by Pumpkin, Inc.
All rights reserved worldwide. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording or otherwise, without prior
permission of Pumpkin, Inc.
Pumpkin, Inc.
750 Naples Street
San Francisco, CA 94112 USA
tel: 415-584-6360
fax: 415-585-7948
web: www.pumpkininc.com
email: sales@pumpkininc.com

Disclaimer
Pumpkin, Incorporated ("Pumpkin") has taken every precaution to provide complete and accurate information in this
document. However, due to continuous efforts being made to improve and update the product(s), Pumpkin and its
Licensor(s) shall not be liable for any technical or editorial errors or omissions contained in this document, or for
any damage, direct or indirect, from discrepancies between the document and the product(s) it describes.
The information is provided on an as-is basis, is subject to change without notice and does not represent a commitment on the part of Pumpkin, Incorporated or its Licensor(s).

Trademarks
The Pumpkin name and logo, the Salvo name and logo, the CubeSat Kit name and logo, "The RTOS that runs in tiny
places." and "Don’t leave Earth without It." are trademarks of Pumpkin, Incorporated.
The absence of a product or service name or logo from this list does not constitute a waiver of Pumpkin's trademark
or other intellectual property rights concerning that name or logo.
All other products and company names mentioned may be trademarks of their respective owners. All words and
terms mentioned that are known to be trademarks or service marks have been appropriately capitalized. Pumpkin,
Incorporated cannot attest to the accuracy of this information. Use of a term should not be regarded as affecting the
validity of any trademark or service mark.
This list may be partial.

Patent Information
The software described in this document is manufactured under one or more of the following U.S. patents:
Patents Pending

Life Support Policy
Pumpkin, Incorporated's products are not authorized for use as critical components in life support devices or systems
without the express written approval of the president of Pumpkin, Incorporated. As used herein:
•

1) Life support devices or systems are devices or systems which, (a) are intended for surgical
implant into the body, or (b) support or sustain life, and whose failure to perform, when

properly used in accordance with instructions for use provided in the labeling, can be
reasonably expected to result in significant injury to the user.
•
•

2) A critical component is any component of a life support device or system whose failure to
perform can be reasonably expected to cause the failure of the life support device or system,
or to affect its safety or effectiveness.

Refund Policy and Limited Warranty on Media
Pumpkin wants you to be happy with your Salvo purchase. That's why Pumpkin invites you to test drive Salvo before you buy. You can download and evaluate the fully functional Salvo freeware version Salvo Lite from the Salvo
web site. If you have questions while you are using Salvo Lite, please don't hesitate to consult the Salvo User Forums, contact our support staff at support@pumpkininc.com, or contact Pumpkin directly.
Because of this free evaluation practice, and because the purchased version contains the complete source code for
Salvo, Pumpkin does not offer refunds on software purchases.
Pumpkin will replace defective distribution media or manuals at no charge, provided you return the item to be replaced with proof of purchase to Pumpkin during the 90-day period after purchase. More details can be found in
Section 11 Limited Warranty on Media of the Pumpkin Salvo License.

Documentation Creation Notes
This documentation was produced using Microsoft Word, Creative Softworx Capture Professional, CorelDRAW!,
Adobe Photoshop, Adobe Illustrator and Adobe Acrobat.
Document name:
Template used:
Last saved on:
Total pages:
Total words:

SalvoUserManual.doc (a Master document)
User's Manual - Template (TT).dot
16:06, Thursday, June 3, 2010
528
97163

Credits
Author:
Artwork:
C-language Advice:
Compiler Advice:

Andrew E. Kalman
Laura Macey, Elizabeth Peartree, Andrew E. Kalman
Russell K. Kadota, Clyde Smith-Stubbs, Dan Henry
Matthew Luckman, Jeffrey O'Keefe, Paul Curtis, Richard Man

Pumpkin Salvo Software License Agreement v1.2
Please Read this Carefully and Completely Before Using this Software.
(Note: The Terms used herein are defined below in Section 1 Definitions)

Grant of License
This License Agreement is a legal agreement between You and Pumpkin, which owns the Software accompanied by
this License or identified above or on the Product Identification Card accompanying this License or on the Product
Identification Label attached to the product package. By clicking the Yes (i.e. Accept) button or by installing, copying, or otherwise using the Software or any Software Updates You agree to be bound by the terms of this License. If
You do not agree to the terms of this License, Pumpkin is unwilling to license the Software to You, and You must
not install, copy, or use the Software, including all Updates that You received as part of the Software. In such event,
You should click the No (i.e. Decline) button and promptly contact Pumpkin for instructions on returning the entire
unused Software and any accompanying product(s) for a refund. By installing, copying, or otherwise using an Update, You agree to be bound by the additional License terms that accompany such Update. If You do not agree to
the terms of the additional License terms that accompany the Update, disregard the Update and the additional License terms that accompany the Update. In this event, Customer's rights to use the Software shall continue to be
governed by the then-existing License.

1 Definitions
"License" means this document, a license agreement.
"You" means an individual or a legal entity exercising rights under, and complying with all of the terms of, this License or a future version of this License. For legal entities, "You" includes any entity that controls, is controlled by,
or is under common control with You. For purposes of this definition, "control" means (i) the power, direct or indirect, to cause the direction or management of such entity, whether by contract or otherwise, or (ii) ownership of fifty
percent (50%) or more of the outstanding shares or beneficial ownership of such entity.
"Pumpkin" means Pumpkin, Incorporated and its Supplier(s).
"Original Code" means Source Code of computer software that is described in the Source Code Notice (below) as
Original Code, and which, at the time of its release under this License is not already Covered Code governed by this
License.
"Source Code" means the preferred form of the Covered Code for making modifications to it, including all modules
it contains, plus any associated interface definition files, scripts used to control compilation and installation of an
Executable, or a list of source code differential comparisons against either the Original Code or another well known,
available Covered Code of Your choice.
"Covered Code" means the Original Code or Modifications or the combination of the Original Code and Modifications, in each case including portions thereof.
"Executable" means Covered Code in any form other than Source Code.
"Application" means computer software or firmware that is created in combination with Covered Code.
"Software" means the proprietary computer software system owned by Pumpkin that includes but is not limited to
software components (including, but not limited to Covered Code), product documentation and associated media,
sample files, extension files, tools, utilities and miscellaneous technical information, in whole or in part.
"Update" means any Software Update.

"Larger Work" means a work that combines Covered Code or portions thereof with code not governed by the terms
of this License.
"Modifications" means any addition to or deletion from the substance or structure of either the Original Code or any
previous Modifications. When Covered Code is released as a series of files, a Modification is (i) any addition to or
deletion from the contents of a file containing Original Code or previous Modifications, or (ii) any new file that contains any part of the Original Code or Previous Modifications.
"Support" means customer support.
"Prerelease Code" means portions of the Software identified as prerelease code or "beta" versions.

2 Copyright
The Software, including all applicable rights to patents, copyrights, trademarks and trade secrets, is the sole and exclusive property of Pumpkin, Incorporated and its Licensor(s) and is provided for Your exclusive use for the purposes of this License. The Software is protected by United States copyright laws and international treaty provisions.
Therefore, You must treat the Software like any other copyrighted material, except that You may either (i) make one
copy of the Software in machine readable form solely for backup or archival purposes, or (ii) transfer the Software
to a hard disk, provided You keep the original solely for backup and archival purposes. Additionally, only so long
as the Software is installed only on the permanent memory of a single computer and that single computer is used by
one user for at least 80% of the time the computer is in use, that same user may also make a copy of the Software to
use on a portable or home computer which is primarily used by such user. As an express condition of this License,
You must reproduce and include on each copy any copyright notice or other proprietary notice that is on the original
copy of the Software supplied by Pumpkin. You may not copy the printed materials accompanying the Software.

3 Source Code License
3.1 The Software is licensed, not sold, to You by Pumpkin for use only under the terms of this License, and Pumpkin reserves any rights not expressly granted to You. Except where explicitly identified as such, the Software is
neither "shareware" nor "freeware" nor "communityware." The Software contains intellectual property in the form of
Source Code, algorithms and other manifestations. You own the media on which the Software is recorded or fixed,
but Pumpkin, Incorporated and its Licensor(s) retains ownership of the Software, related documentation and fonts.
3.2 Pumpkin grants You the use of the Software only if You have registered the Software with Pumpkin by returning the registration card or by other means specified by Pumpkin.
3.3 Pumpkin grants You a non-exclusive, worldwide License, subject to third-party intellectual property claims, (i)
to use and modify ("Utilize") the Software (or portions thereof) with or without Modifications, or as part of a Larger
Work, on a single computer for the purpose of creating, modifying, running, debugging and testing Your own Application and any of its updates, enhancements and successors, and (ii) under patents now or hereafter owned or controlled by Pumpkin, to Utilize the Software (or portions thereof), but solely to the extent that any such patent is
reasonably necessary to enable You to Utilize the Software (or portions thereof) and not to any greater extent that
may be necessary to Utilize further Modifications or combinations. To use ("Use") the Software means that the
Software is either loaded in the temporary memory (i.e. RAM) of a computer or installed on the permanent memory
of a computer (i.e. hard disk, etc.). You may Use the Software on a network, provided that a licensed copy of the
software has been acquired for each person permitted to access the Software through the network. You may also Use
the Software in object form only (i.e. as an Executable) on a single, different computer or computing device (e.g.
target microcontroller or microprocessor, demonstration or evaluation board, in-circuit emulator, test system, prototype, etc.).
3.4 Any supplemental software code or other materials provided to You as part of Pumpkin's Support shall be considered part of the Software and subject to the terms and conditions of this License. With respect to technical information You provide to Pumpkin as part of the Support, Pumpkin may use such information for its business
purposes, including product support and development. Pumpkin will not utilize such technical information in a form
that personally identifies You without Your permission.

3.5 The Software shall be deemed accepted by You upon payment of the Software by You and shall not be granted a
refund of any license fees for the Software, except for Your rights defined in this License.

4 Software Distribution Obligations
4.1 You may not under any circumstances release or distribute the Source Code, with or without Modifications, or as
part of a Larger Work, without Pumpkin's express written permission.
4.2 You may distribute the Software in Executable form only and as part of a Larger Work only (i.e. in conjunction
with and as part of Your Application. Additionally, You must (i) not permit the further redistribution of the Software
in any form by Your customers, (ii) include a valid copyright notice in Your application (where possible - if it is not
possible to put such a notice in Your Application due to its structure, then You must include such a notice in a location (such as a relevant directory file) where a user would be likely to look for such a notice), (iii) include the existing copyright notice(s) in all Pumpkin Software used in Your Application, (iv) agree to indemnify, hold harmless
and defend Pumpkin from and against any and all claims and lawsuits, including attorney's fees, that arise or result
from the use or distribution of Your Application, (v) otherwise comply with the terms of this License, and (vi) agree
that Pumpkin reserves all rights not expressly granted.
4.3 You may freely distribute the demonstration programs (identified as "Demo") that are part of the Software as
long as they are accompanied by this License.
4.4 The freeware version (consisting of pre-compiled libraries, a limited number of source code files, and various
other files and documentation) and identified as "Freeware" is governed by this license, with the following exceptions: The sole exception shall be for a Larger Work created exclusively with the freeware libraries that are part of
the Software; in this case Pumpkin automatically grants You the right to distribute Your Application freely.
4.5 You may not under any circumstances, other than those explicitly mentioned in Sections 4.2, 4.3 and 4.4 above,
release or distribute the Covered Code, with or without Modifications, or as part of a Larger Work, without Pumpkin's express written permission.

5 Other Restrictions
5.1 You may not permit other individuals to use the Software except under the terms of this License.
5.2 You may not rent, lease, grant a security interest in, loan or sublicense the Software; nor may You create derivative works based upon the Software in whole or in part.
5.3 You may not translate, decompile, reverse engineer, disassemble (except and solely to the extent an applicable
statute expressly and specifically prohibits such restrictions), or otherwise attempt to create a human-readable version of any parts of the Software supplied exclusively in binary form.
5.4 If the Software was licensed to You for academic use, You may not use the software for commercial product
development.
5.5 You may not remove any designation mark from any supplied material that identifies such material as belonging
to or developed by Pumpkin.
5.6 You may permanently transfer all of Your rights under this License, provided You retain no copies, You transfer
all of the Software (including all component parts, the media and printed materials, any upgrades, and this License),
You provide Pumpkin notice of Your name, company, and address and the name, company, and address of the person to whom You are transferring the rights granted herein, and the recipient agrees to the terms of this License and
pays to Pumpkin a transfer fee in an amount to be determined by Pumpkin and in effect at the time in question. If
the Software is an upgrade, any transfer must include all prior versions of the Software. If the Software is received
as part of a subscription, any transfer must include all prior deliverables of Software and all other subscription deliverables. Upon such transfer, Your License under this Agreement is automatically terminated.

xv

5.7 You may use or transfer the Updates to the Software only in conjunction with Your then-existing Software. The
Software and all Updates are licensed as a single product and the Updates may not be separated from the Software
for use at any time.

6 Termination
This License is effective until terminated. This License will terminate immediately without notice from Pumpkin or
judicial resolution if You fail to comply with any provision of this License, and You may terminate this License at
any time. Upon such termination You must destroy the Software, all accompanying written materials and all copies
thereof. Provisions which, by their nature, must remain in effect beyond the termination of this License shall survive.

7 Multiple Media
Even if this Pumpkin product includes the Software on more than one medium (e.g., on both a CD-ROM and on
magnetic disk(s); or on both 3.5 inch disk(s) and 5.25 inch disk(s)), You are only licensed to use one copy of the
Software as described in Section 2.3. The restrictions contained herein apply equally to hybrid media that may contain multiple versions of the Software for use on different operating systems. Regardless of the type of media You
receive, You may only use the portion appropriate for Your single user computer / workstation. You may not use
the Software stored on the other medium on another computer or common storage device, nor may You rent, lease,
loan or transfer it to another user except as part of a transfer pursuant to Section 5.7.

8 Prerelease Code
Prerelease Code may not be at the level of performance and compatibility of the final, generally available product
offering, and may not operate correctly and may be substantially modified prior to first commercial shipment.
Pumpkin is not obligated to make this or any later version of the Prerelease Code commercially available. The grant
of license to use Prerelease Code expires upon availability of a commercial release of the Prerelease Code from
Pumpkin.

9 Export Law Assurances
You may not use or otherwise export or re-export the Software except as authorized by United States law and the
laws of the jurisdiction in which the Software was obtained. In particular, but without limitation, the Software may
not be exported or re-exported to (i) into (or to a national or resident of) any U.S. embargoed country or (ii) to anyone on the U.S. Treasury Department's list of Specially Designated Nations or the U.S. Department of Commerce's
Table of Denial Orders. By using the Software You represent and warrant that You are not located in, under control
of, or a national or resident of any such country or on any such list.

10 U.S. Government End Users
If You are acquiring the Software and fonts on behalf of any unit or agency of the United States Government, the
following provisions apply. The Government agrees that the Software and fonts shall be classified as "commercial
computer software" and "commercial computer software documentation" as such terms are defined in the applicable
provisions of the Federal Acquisition Regulation ("FAR") and supplements thereto, including the Department of
Defense ("DoD") FAR Supplement ("DFARS"). If the Software and fonts are supplied for use by DoD, it is delivered subject to the terms of this Agreement and either (i) in accordance with DFARS 227.7202-1(a) and 227.72023(a), or (ii) with restricted rights in accordance with DFARS 252.227-7013(c)(1)(ii) (OCT 1988), as applicable. If
the Software and fonts are supplied for use by any other Federal agency, it is restricted computer software delivered
subject to the terms of this Agreement and (i) FAR 12.212(a); (ii) FAR 52.227-19; or (iii) FAR 52.227-14(ALT III),
as applicable.

11 Limited Warranty on Media
Pumpkin warrants for a period of ninety (90) days from Your date of purchase (as evidenced by a copy of Your receipt) that the media provided by Pumpkin, if any, on which the Software is recorded will be free from defects in
materials and workmanship under normal use. Pumpkin will have no responsibility to replace media damaged by
accident, abuse or misapplication. PUMPKIN'S ENTIRE LIABILITY AND YOUR SOLE AND EXCLUSIVE
REMEDY WILL BE, AT PUMPKIN'S OPTION, REPLACEMENT OF THE MEDIA, REFUND OF THE
PURCHASE PRICE OR REPAIR OR REPLACEMENT OF THE SOFTWARE. ANY IMPLIED WARRANTIES
ON THE MEDIA, INCLUDING THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
A PARTICULAR PURPOSE, ARE LIMITED IN DURATION TO NINETY (90) DAYS FROM THE DATE OF
DELIVERY. THIS WARRANTY GIVES YOU SPECIFIC LEGAL RIGHTS, AND YOU MAY ALSO HAVE
OTHER RIGHTS THAT VARY BY JURISDICTION.

12 Disclaimer of Warranty
THIS LIMITED WARRANTY IS THE ONLY WARRANTY PROVIDED BY PUMPKIN. PUMPKIN
EXPRESSLY DISCLAIMS ALL OTHER WARRANTIES AND/OR CONDITIONS, ORAL OR WRITTEN,
EITHER EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO, IMPLIED WARRANTIES OR
CONDITIONS OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE WITH REGARD TO
THE SOFTWARE AND ACCOMPANYING WRITTEN MATERIALS, AND NONINFRINGEMENT.
PUMPKIN DOES NOT WARRANT THAT THE FUNCTIONS CONTAINED IN THE SOFTWARE WILL
MEET YOUR REQUIREMENTS, OR THAT THE OPERATION OF THE SOFTWARE WILL BE
UNINTERRUPTED OR ERROR-FREE, OR THAT DEFECTS IN THE SOFTWARE WILL BE CORRECTED.
FURTHERMORE, PUMPKIN DOES NOT WARRANT OR MAKE ANY REPRESENTATIONS REGARDING
THE USE OR THE RESULTS OF THE USE OF THE SOFTWARE OR RELATED DOCUMENTATION IN
TERMS OF THEIR CORRECTNESS, ACCURACY, RELIABILITY, OR OTHERWISE. AS A RESULT, THE
SOFTWARE IS LICENSED "AS-IS", AND YOU THE LICENSEE EXPRESSLY ASSUME ALL LIABILITIES
AND RISKS, FOR USE OR OPERATION OF ANY APPLICATION PROGRAMS YOU MAY CREATE WITH
THE SOFTWARE, INCLUDING WITHOUT LIMITATION, APPLICATIONS DESIGNED OR INTENDED FOR
MISSION CRITICAL APPLICATIONS AND HIGH-RISK ACTIVITIES, SUCH AS THE OPERATION OF
NUCLEAR FACILITIES, PACEMAKERS, DIRECT LIFE SUPPORT MACHINES, WEAPONRY, AIR
TRAFFIC CONTROL, AIRCRAFT NAVIGATION OR COMMUNICATIONS SYSTEMS, FACTORY
CONTROL SYSTEMS, ETC., IN WHICH THE FAILURE OF THE SOFTWARE COULD LEAD DIRECTLY TO
DEATH, PERSONAL INJURY, OR SEVERE PHYSICAL OR ENVIRONMENTAL DAMAGE. NO PUMPKIN
DEALER, DIRECTOR, OFFICER, EMPLOYEE OR AGENT IS AUTHORIZED TO MAKE ANY
MODIFICATION, EXTENSION, OR ADDITION TO THIS WARRANTY. BECAUSE SOME JURISDICTIONS
DO NOT ALLOW THE EXCLUSION OR LIMITATION OF IMPLIED WARRANTIES, THE ABOVE
LIMITATION MAY NOT APPLY TO YOU. THIS WARRANTY GIVES YOU SPECIFIC LEGAL RIGHTS,
AND YOU MAY ALSO HAVE OTHER RIGHTS THAT VARY BY JURISDICTION.

13 Limitation of Liabilities, Remedies and Damages
TO THE MAXIMUM EXTENT PERMITTED BY APPLICABLE LAW, IN NO EVENT WILL PUMPKIN,
INCORPORATED, OR ANY OF ITS LICENSORS, SUPPLIERS, DIRECTORS, OFFICERS, EMPLOYEES OR
AGENTS (COLLECTIVELY "PUMPKIN AND ITS SUPPLIER(S)") BE LIABLE TO YOU FOR ANY
CONSEQUENTIAL, INCIDENTAL, INDIRECT OR SPECIAL DAMAGES WHATSOEVER (INCLUDING,
WITHOUT LIMITATION, DAMAGES FOR LOSS OF BUSINESS PROFITS, BUSINESS INTERRUPTION,
LOSS OF BUSINESS INFORMATION AND THE LIKE, OR ANY OTHER PECUNIARY LOSS), WHETHER
FORESEEABLE OR UNFORESEEABLE, ARISING OUT OF THE USE OF OR INABILITY TO USE THE
SOFTWARE OR ACCOMPANYING WRITTEN MATERIALS, REGARDLESS OF THE BASIS OF THE
CLAIM AND EVEN IF PUMPKIN AND ITS SUPPLIER(S) HAS BEEN ADVISED OF THE POSSIBILITY OF
SUCH DAMAGES. THIS LIMITATION WILL NOT APPLY IN CASE OF PERSONAL INJURY ONLY
WHERE AND TO THE EXTENT THAT APPLICABLE LAW REQUIRES SUCH LIABILITY. BECAUSE
SOME JURISDICTIONS DO NOT ALLOW THE EXCLUSION OF LIMITATION OF LIABILITY FOR
CONSEQUENTIAL OR INCIDENTAL DAMAGES, THE ABOVE LIMITATIONS MAY NOT APPLY TO
YOU. IN NO EVENT SHALL PUMPKIN AND ITS SUPPLIER(S)' TOTAL LIABILITY TO YOU FOR ALL

xvii

DAMAGES, LOSSES AND CAUSES OF ACTION (WHETHER IN CONTRACT, TORT (INCLUDING
NEGLIGENCE), PRODUCT LIABILITY OR OTHERWISE) EXCEED $50.00.
PUMPKIN SHALL BE RELIEVED OF ANY AND ALL OBLIGATIONS WITH RESPECT TO THIS SECTION
FOR ANY PORTIONS OF THE SOFTWARE THAT ARE REVISED, CHANGED, MODIFIED, OR
MAINTAINED BY ANYONE OTHER THAN PUMPKIN.

14 Complete Agreement, Controlling Law and Severability
This License constitutes the entire agreement between You and Pumpkin with respect to the use of the Software, the
related documentation and fonts, and supersedes all prior or contemporaneous understandings or agreements, written
or oral, regarding such subject matter. No amendment to or modification of this License will be binding unless in
writing and signed by a duly authorized representative of Pumpkin. The acceptance of any purchase order placed by
You is expressly made conditional on Your assent to the terms set forth herein, and not those in Your purchase order. This License will be construed under the laws of the State of California, except for that body of law dealing
with conflicts of law. If any provision of this License shall be held by a court of competent jurisdiction to be contrary to law, that provision will be enforced to the maximum extent permissible, and the remaining provisions of this
License will remain in full force and effect. The application of the United Nations Convention on Contracts for the
International Sale of Goods is expressly excluded. Any law or regulation that provides that the language of a contract shall be construed against the drafter shall not apply to this License. In the event of any action to enforce this
Agreement, the prevailing party shall be entitled to recover from the other its court costs and reasonable attorneys'
fees, including costs and fees on appeal.

15 Additional Terms
Nothing in this License shall be interpreted to prohibit Pumpkin from licensing under terms different from this License any code which Pumpkin otherwise would have a right to License.
This License does not grant You any rights to use the trademarks or logos that are the property of Pumpkin, Inc.,
even if such marks are included in the Software. You may contact Pumpkin for permission to display the abovementioned marks.
Pumpkin may publish revised and/or new versions of this License from time to time. Each version will be given a
distinguishing version number.
Should You have any questions or comments concerning this License, please do not hesitate to write to Pumpkin,
Inc., 750 Naples Street, San Francisco, CA 94112 USA, Attn: Warranty Information. You may also send email to
support@pumpkininc.com.

Source Code Notice
The contents of this file are subject to the Pumpkin Salvo License (the "License"). You may not use this file except
in compliance with the License. You may obtain a copy of the License at http://www.pumpkininc.com, or from warranty@pumpkininc.com.
Software distributed under the License is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY
KIND, either express or implied. See the License for specific language governing the warranty and the rights and
limitations under the License.
The Original Code is Salvo - The RTOS that runs in tiny places(tm). Copyright (C) 1995-2002 Pumpkin, Inc. and its
Licensor(s). All Rights Reserved.

Contents
Contents ............................................................................................................... i
Figures .............................................................................................................. xv
Listings............................................................................................................ xvii
Tables ............................................................................................................... xix
Release Notes .................................................................................................. xxi
Introduction ............................................................................................................................xxi
What's New ............................................................................................................................xxi
Release Notes .........................................................................................................................xxi
Third-Party Tool Versions......................................................................................................xxi

Supported Targets and Compilers............................................................... xxiii
Preface ............................................................................................................ xxv
Historical Information ........................................................................................................... xxv
Typographic Conventions ..................................................................................................... xxv
Standardized Numbering Scheme ........................................................................................xxvi
The Salvo Coding Mindset..................................................................................................xxvii
Configurability Is King.................................................................................................xxvii
Conserve Precious Resources ......................................................................................xxviii
Learn to Love the Preprocessor ...................................................................................xxviii
Document, But Don't Duplicate...................................................................................xxviii
We're Not Perfect.........................................................................................................xxviii

Chapter 1 • Introduction..................................................................................... 1
Welcome....................................................................................................................................1
What Is Salvo?...........................................................................................................................2
Why Should I Use Salvo? .........................................................................................................2
What Kind of RTOS Is Salvo? ..................................................................................................3
What Does a Salvo Program Look Like? ..................................................................................3
What Resources Does Salvo Require? ......................................................................................5
How Is Salvo Different?............................................................................................................6
What Do I Need to Use Salvo?..................................................................................................7
Which Processors and Compilers does Salvo Support? ............................................................8
Salvo User Manual

i

How Is Salvo Distributed? ........................................................................................................8
What Is in this Manual?.............................................................................................................8

Chapter 2 • RTOS Fundamentals..................................................................... 11
Introduction .............................................................................................................................11
Basic Terms.............................................................................................................................12
Foreground / Background Systems .........................................................................................14
Reentrancy...............................................................................................................................15
Resources ................................................................................................................................16
Multitasking and Context Switching.......................................................................................16
Tasks and Interrupts ................................................................................................................17
Preemptive vs. Cooperative Scheduling..................................................................................18
Preemptive Scheduling .....................................................................................................19
Cooperative Scheduling....................................................................................................20
More on Multitasking..............................................................................................................21
Task Structure ...................................................................................................................21
Simple Multitasking..........................................................................................................22
Priority-based Multitasking ..............................................................................................22
Task States ........................................................................................................................23
Delays and the Timer ........................................................................................................24
Event-driven Multitasking ................................................................................................26
Events and Intertask Communications ....................................................................................29
Semaphores.......................................................................................................................29
Event Flags.................................................................................................................30
Task Synchronization.................................................................................................31
Resources ...................................................................................................................33
Messages...........................................................................................................................35
Message Queues ...............................................................................................................37
Summary of Task and Event Interaction .................................................................................37
Conflicts ..................................................................................................................................38
Deadlock ...........................................................................................................................38
Priority Inversions.............................................................................................................39
RTOS Performance .................................................................................................................39
A Real-World Example ...........................................................................................................39
The Conventional Superloop Approach............................................................................40
The Event-Driven RTOS Approach..................................................................................41
Step By Step......................................................................................................................43
Initializing the Operating System...............................................................................43
Structuring the Tasks..................................................................................................43
Prioritizing the Tasks..................................................................................................44
Interfacing with Events ..............................................................................................45
Adding the System Timer...........................................................................................45
Starting the Tasks .......................................................................................................45
Enabling Multitasking ................................................................................................46
Putting It All Together ...............................................................................................46
The RTOS Difference.......................................................................................................49

Chapter 3 • Installation..................................................................................... 51
Introduction .............................................................................................................................51

ii

Contents

Salvo User Manual

Running the Installer ...............................................................................................................51
Network Installation .........................................................................................................56
Installing Salvo on non-Wintel Platforms.........................................................................57
A Completed Installation.........................................................................................................57
Uninstalling Salvo ...................................................................................................................58
Uninstalling Salvo on non-Wintel Machines....................................................................59
Installations with Multiple Salvo Distributions.......................................................................60
Installer Behavior..............................................................................................................60
Installing Multiple Salvo Distributions.............................................................................60
Uninstalling with Multiple Salvo Distributions................................................................60
Copying Salvo Files ................................................................................................................60
Modifying Salvo Files .............................................................................................................61

Chapter 4 • Tutorial........................................................................................... 63
Introduction .............................................................................................................................63
Part 1: Writing a Salvo Application ........................................................................................63
Tut1: Initializing Salvo and Starting to Multitask ............................................................63
Tut2: Creating, Starting and Switching tasks ...................................................................65
Tut3: Adding Functionality to Tasks ................................................................................68
Tut4: Using Events for Better Performance......................................................................70
Tut5: Delaying a Task.......................................................................................................74
Signaling from Multiple Tasks .........................................................................................78
Wrapping Up.....................................................................................................................81
Food For Thought .............................................................................................................82
Part 2: Building a Salvo Application.......................................................................................82
Working Environment ......................................................................................................82
Creating a Project Directory .............................................................................................83
Including salvo.h...............................................................................................................84
Configuring your Compiler...............................................................................................84
Setting Search Paths ...................................................................................................84
Using Libraries vs. Using Source Files.............................................................................85
Using Libraries .................................................................................................................85
Using Source Files ............................................................................................................86
Setting Configuration Options....................................................................................86
Linking to Salvo Object Files.....................................................................................90

Chapter 5 • Configuration ................................................................................ 93
Introduction .............................................................................................................................93
The Salvo Build Process..........................................................................................................93
Library Builds ...................................................................................................................93
Source-Code Builds ..........................................................................................................96
Benefits of Different Build Types.....................................................................................98
Configuration Option Overview..............................................................................................98
Configuration Options for all Distributions ............................................................................99
OSCOMPILER: Identify Compiler in Use .....................................................................100
OSEVENTS: Set Maximum Number of Events.............................................................101
OSEVENT_FLAGS: Set Maximum Number of Event Flags ........................................102
OSLIBRARY_CONFIG: Specify Precompiled Library Configuration .........................103

Salvo User Manual

Contents

iii

OSLIBRARY_GLOBALS: Specify Memory Type for Global Salvo Objects in
Precompiled Library....................................................................................................104
OSLIBRARY_OPTION: Specify Precompiled Library Option.....................................105
OSLIBRARY_TYPE: Specify Precompiled Library Type ............................................106
OSLIBRARY_VARIANT: Specify Precompiled Library Variant.................................107
OSMESSAGE_QUEUES: Set Maximum Number of Message Queues........................108
OSTARGET: Identify Target Processor.........................................................................109
OSTASKS: Set Maximum Number of Tasks and Cyclic Timers...................................110
OSUSE_LIBRARY: Use Precompiled Library..............................................................111
Configuration Options for Source Code Distributions..........................................................112
OSBIG_SEMAPHORES: Use 16-bit Semaphores.........................................................113
OSBYTES_OF_COUNTS: Set Size of Counters...........................................................114
OSBYTES_OF_DELAYS: Set Length of Delays ..........................................................115
OSBYTES_OF_EVENT_FLAGS: Set Size of Event Flags...........................................116
OSBYTES_OF_TICKS: Set Maximum System Tick Count .........................................117
OSCALL_OSCREATEEVENT: Manage Interrupts when Creating Events..................118
OSCALL_OSGETPRIOTASK: Manage Interrupts when Returning a Task's Priority..121
OSCALL_OSGETSTATETASK: Manage Interrupts when Returning a Task's State ..121
OSCALL_OSMSGQCOUNT: Manage Interrupts when Returning Number of
Messages in Message Queue .......................................................................................121
OSCALL_OSMSGQEMPTY: Manage Interrupts when Checking if Message Queue
is Empty.......................................................................................................................121
OSCALL_OSRETURNEVENT: Manage Interrupts when Reading and/or Trying
Events ..........................................................................................................................122
OSCALL_OSSIGNALEVENT: Manage Interrupts when Signaling Events and
Manipulating Event Flags............................................................................................122
OSCALL_OSSTARTTASK: Manage Interrupts when Starting Tasks..........................122
OSCLEAR_GLOBALS: Explicitly Clear all Global Parameters...................................123
OSCLEAR_UNUSED_POINTERS: Reset Unused Tcb and Ecb Pointers....................124
OSCOLLECT_LOST_TICKS: Configure Timer System For Maximum Versatility ....125
OSCOMBINE_EVENT_SERVICES: Combine Common Event Service Code............126
OSCTXSW_METHOD: Identify Context-Switching Methodology in Use...................127
OSCUSTOM_LIBRARY_CONFIG: Select Custom Library Configuration File..........128
OSDISABLE_ERROR_CHECKING: Disable Runtime Error Checking......................129
OSDISABLE_FAST_SCHEDULING: Configure Round-Robin Scheduling ...............130
OSDISABLE_TASK_PRIORITIES: Force All Tasks to Same Priority........................131
OSENABLE_BINARY_SEMAPHORES: Enable Support for Binary Semaphores .....132
OSENABLE_BOUNDS_CHECKING: Enable Runtime Pointer Bounds Checking.....133
OSENABLE_CYCLIC_TIMERS: Enable Cyclic Timers .............................................134
OSENABLE_EVENT_FLAGS: Enable Support for Event Flags..................................135
OSENABLE_EVENT_READING: Enable Support for Event Reading........................136
OSENABLE_EVENT_TRYING: Enable Support for Event Trying.............................137
OSENABLE_FAST_SIGNALING: Enable Fast Event Signaling.................................138
OSENABLE_IDLE_COUNTER: Track Scheduler Idling.............................................139
OSENABLE_IDLING_HOOK: Call a User Function when Idling ...............................140
OSENABLE_MESSAGES: Enable Support for Messages............................................141
OSENABLE_MESSAGE_QUEUES: Enable Support for Message Queues.................142
OSENABLE_OSSCHED_DISPATCH_HOOK: Call User Function Inside Scheduler 143
OSENABLE_OSSCHED_ENTRY_HOOK: Call User Function Inside Scheduler.......144
OSENABLE_OSSCHED_RETURN_HOOK: Call User Function Inside Scheduler....145
OSENABLE_SEMAPHORES: Enable Support for Semaphores ..................................146

iv

Contents

Salvo User Manual

OSENABLE_STACK_CHECKING: Monitor Call ... Return Stack Depth...................147
OSENABLE_TCBEXT0|1|2|3|4|5: Enable Tcb Extensions ...........................................148
OSENABLE_TIMEOUTS: Enable Support for Timeouts.............................................151
OSGATHER_STATISTICS: Collect Run-time Statistics..............................................152
OSINTERRUPT_LEVEL: Specify Interrupt Level for Interrupt-callable Services.......153
OSLOC_ALL: Storage Type for All Salvo Objects .......................................................154
OSLOC_COUNT: Storage Type for Counters ...............................................................156
OSLOC_CTCB: Storage Type for Current Task Control Block Pointer........................157
OSLOC_DEPTH: Storage Type for Stack Depth Counters ...........................................157
OSLOC_ECB: Storage Type for Event Control Blocks and Queue Pointers.................157
OSLOC_EFCB: Storage Type for Event Flag Control Blocks.......................................157
OSLOC_ERR: Storage Type for Error Counters............................................................158
OSLOC_GLSTAT: Storage Type for Global Status Bits...............................................158
OSLOC_LOGMSG: Storage Type for Log Message String ..........................................158
OSLOC_LOST_TICK: Storage Type for Lost Ticks .....................................................158
OSLOC_MQCB: Storage Type for Message Queue Control Blocks.............................159
OSLOC_MSGQ: Storage Type for Message Queues.....................................................159
OSLOC_PS: Storage Type for Timer Prescalar .............................................................159
OSLOC_TCB: Storage Type for Task Control Blocks ..................................................160
OSLOC_SIGQ: Storage Type for Signaled Events Queue Pointers...............................160
OSLOC_TICK: Storage Type for System Tick Counter ................................................160
OSLOGGING: Log Runtime Errors and Warnings........................................................161
OSLOG_MESSAGES: Configure Runtime Logging Messages ....................................162
OS_MESSAGE_TYPE: Configure Message Pointers ...................................................164
OSMPLAB_C18_LOC_ALL_NEAR: Locate all Salvo Objects in Access Bank
(MPLAB-C18 Only)....................................................................................................165
OSOPTIMIZE_FOR_SPEED: Optimize for Code Size or Speed..................................166
OSPIC18_INTERRUPT_MASK: Configure PIC18 Interrupt Mode.............................167
OSRPT_HIDE_INVALID_POINTERS: OSRpt() Won't Display Invalid Pointers.......169
OSRPT_SHOW_ONLY_ACTIVE: OSRpt() Displays Only Active Task and Event
Data .............................................................................................................................170
OSRPT_SHOW_TOTAL_DELAY: OSRpt() Shows the Total Delay in the Delay
Queue...........................................................................................................................171
OSRTNADDR_OFFSET: Offset (in bytes) for Context-Switching Saved Return
Address........................................................................................................................172
OSSCHED_RETURN_LABEL(): Define Label within OSSched() ..............................173
OSSET_LIMITS: Limit Number of Runtime Salvo Objects..........................................174
OSSPEEDUP_QUEUEING: Speed Up Queue Operations............................................175
OSTIMER_PRESCALAR: Configure Prescalar for OSTimer()....................................176
OSTYPE_TCBEXT0|1|2|3|4|5: Set Tcb Extension Type ...............................................177
OSUSE_CHAR_SIZED_BITFIELDS: Pack Bitfields into Chars .................................178
OSUSE_EVENT_TYPES: Check for Event Types at Runtime.....................................179
OSUSE_INLINE_OSSCHED: Reduce Task Call…Return Stack Depth ......................180
OSUSE_INLINE_OSTIMER: Eliminate OSTimer() Call…Return Stack Usage..........182
OSUSE_INSELIG_MACRO: Reduce Salvo's Call Depth.............................................183
OSUSE_MEMSET: Use memset() (if available) ...........................................................184
Organization ..........................................................................................................................185
Choosing the Right Options for your Application ................................................................186
Predefined Configuration Constants......................................................................................188
Obsolete Configuration Parameters.......................................................................................189

Salvo User Manual

Contents

v

Chapter 6 • Frequently Asked Questions (FAQ) .......................................... 191
General ..................................................................................................................................191
What is Salvo? ................................................................................................................191
Is there a shareware / freeware / open source version of Salvo? ....................................191
Just how small is Salvo? .................................................................................................192
Why should I use Salvo? ................................................................................................192
What should I consider Salvo Pro over Salvo LE?.........................................................193
What can I do with Salvo?..............................................................................................193
What kind of RTOS is Salvo?.........................................................................................194
What are Salvo's minimum requirements? .....................................................................194
What kind of processors can Salvo applications run on?................................................194
My compiler doesn't implement a stack. It allocates variables using a static overlay
model. Can it be used with Salvo? ..............................................................................195
How many tasks and events does Salvo support?...........................................................195
How many priority levels does Salvo support? ..............................................................195
What kind of events does Salvo support? .......................................................................195
Is Salvo Y2K compliant?................................................................................................195
Where did Salvo come from? .........................................................................................196
Getting Started.......................................................................................................................196
Where can I find examples of projects that use Salvo? ..................................................196
Which compiler(s) do you recommend for use with Salvo?...........................................196
Is there a tutorial? ...........................................................................................................196
Apart from the Salvo User Manual, what other sources of documentation are
available?.....................................................................................................................197
I'm on a tight budget. Can I use Salvo? ..........................................................................197
I only have an assembler. Can I use Salvo?....................................................................197
Performance...........................................................................................................................197
How can using Salvo improve the performance of my application? ..............................197
How do delays work under Salvo? .................................................................................198
What's so great about having task priorities?..................................................................198
When does the Salvo code in my application actually run? ...........................................199
How can I perform fast, timing-critical operations under Salvo?...................................199
Memory .................................................................................................................................199
How much will Salvo add to my application's ROM and RAM usage?.........................199
How much RAM will an application built with the libraries use?..................................200
Do I need to worry about running out of memory? ........................................................200
If I define a task or event but never use it, is it costing me RAM?.................................201
How much call ... return stack depth does Salvo use? ....................................................201
Why must I use pointers when working with tasks? Why can't I use explicit task IDs? 202
How can I avoid re-initializing Salvo's variables when I wake up from sleep on a
PIC12C509 PICmicro MCU?......................................................................................203
Libraries ................................................................................................................................203
What kinds of libraries does Salvo include?...................................................................203
What's in each Salvo library?..........................................................................................204
Why are there so many libraries?....................................................................................204
Should I use the libraries or the source code when building my application?................204
What's the difference between the freeware and standard Salvo libraries? ....................204
My library-based application is using more RAM than I can account for. Why? ..........204
I'm using a library. Why does my application use more RAM than one compiled
directly from source files? ...........................................................................................205

vi

Contents

Salvo User Manual

I'm using a freeware library and I get the message "#error: OSXYZ exceeds library
limit – aborting." Why? ...............................................................................................205
Why can't I alter the functionality of a library by adding configuration options to my
salvocfg.h?...................................................................................................................205
The libraries are very large – much larger than the ROM size of my target processor.
Won't that affect my application?................................................................................206
I'm using a library. Can I change the bank where Salvo variables are located? .............206
Configuration.........................................................................................................................206
I'm overwhelmed by all the configuration options. Where should I start? .....................206
Do I have to use all of Salvo's functionality? .................................................................207
What file(s) do I include in my main.c? .........................................................................207
What is the purpose of OSENABLE_SEMAPHORES and similar configuration
options? .......................................................................................................................207
Can I collect run-time statistics with Salvo?...................................................................207
How can I clear my processor's watchdog timer with Salvo?.........................................207
I enabled timeouts and my RAM and ROM grew substantially– why? .........................208
Timer and Timing..................................................................................................................208
Do I have to install the timer?.........................................................................................208
How do I install the timer?..............................................................................................208
I added the timer to my ISR and now my ISR is huge and slow. What should I do? .....209
How do I pick a tick rate for Salvo? ...............................................................................209
How do I use the timer prescalar?...................................................................................209
I enabled the prescalar and set it to 1 but it didn't make any difference. Why?..............209
What is the accuracy of the system timer?......................................................................210
What is Salvo's interrupt latency?...................................................................................210
What if I need to specify delays larger than 8 bits of ticks? ...........................................210
How can I achieve very long delays via Salvo? Can I do that and still keep task
memory to a minimum?...............................................................................................210
Can I specify a timeout when waiting for an event?.......................................................211
Does Salvo provide functions to obtain elapsed time? ...................................................211
How do I choose the right value for OSBYTES_OF_TICKS?.......................................212
My processor has no interrupts. Can I still use Salvo's timer services?..........................213
Context Switching .................................................................................................................213
How do I know when I'm context switching in Salvo?...................................................213
Why can't I context switch from something other than the task level?...........................213
Why does Salvo use macros to do context switching? ...................................................213
Can I context switch in more than one place per task?...................................................214
When must I use context-switching labels?....................................................................214
Tasks & Events......................................................................................................................214
What are taskIDs?...........................................................................................................214
Does it matter which taskID I assign to a particular task?..............................................215
Is there an idle task in Salvo? .........................................................................................215
How can I monitor the tasks in my application?.............................................................215
What exactly happens in the scheduler? .........................................................................215
What about reentrant code and Salvo?............................................................................216
What are "implicit" and "explicit" OS task functions? ...................................................216
How do I setup an infinite loop in a task? ......................................................................216
Why must tasks use static local variables? .....................................................................217
Doesn't using static local variables take more memory than with other RTOSes?.........217
Can tasks share the same priority?..................................................................................217
Can I have multiple instances of the same task?.............................................................218

Salvo User Manual

Contents

vii

Does the order in which I start tasks matter?..................................................................218
How can I reduce code size when starting tasks? ...........................................................219
What is the difference between a delayed task and a waiting task?................................219
Can I create a task to immediately wait an event?..........................................................220
I started a task but it never ran. Why? ............................................................................220
What happens if I forget to loop in my task?..................................................................220
Why did my low-priority run-time tasks start running before my high-priority startup
task completed? ...........................................................................................................221
When I signaled a waiting task, it took much longer than the context switching time
to run. Why? ................................................................................................................221
Can I destroy a task and (re-) create a new one in its place? ..........................................221
Can more than one task wait on an event?......................................................................222
Does Salvo preserve the order in which events occur?...................................................222
Can a task wait on more than one event at a time? .........................................................222
How can I implement event flags?..................................................................................223
What happens when a task times out waiting for an event? ...........................................224
Why is my high-priority task stuck waiting, while other low-priority tasks are
running?....................................................................................................................... 224
When an event occurs and there are tasks waiting for it, which task(s) become
eligible? .......................................................................................................................224
How can I tell if a task timed out waiting for an event? .................................................225
Can I create an event from inside a task?........................................................................225
What kind of information can I pass to a task via a message?........................................226
My application uses messages and binary semaphores. Is there any way to make the
Salvo code smaller?.....................................................................................................226
Why did RAM requirements increase substantially when I enabled message queues?..227
Can I signal an event from outside a task? .....................................................................227
When I signal a message that has more than one task waiting for it, why does only
one task become eligible?............................................................................................227
I'm using a message event to pass a character variable to a waiting task, but I don't
get the right data when I dereference the pointer. What's going on?...........................227
What happens when there are no tasks in the eligible queue? ........................................228
In what order do messages leave a message queue?.......................................................229
What happens if an event is signaled before any task starts to wait it? Will the event
get lost or it will be processed after task starts to wait it? ...........................................229
What happens if an event is signaled several times before waiting task gets a chance
to run and process that event? Will the last one signal be processed and previous
lost? Or the first will be processed and the following signals lost?.............................229
What is more important to create first, an event or the task that waits it? Does the
order of creation matter? .............................................................................................229
What if I don't need one event anymore and want to use its slot for another event?
Can I destroy event? ....................................................................................................229
Can I use messages or message queues to pass raw data between tasks?.......................230
How can I test if there's room for additional messages in a message queue without
signaling the message queue?......................................................................................230
Interrupts ...............................................................................................................................230
Why does Salvo disable all interrupts during a critical section of code?........................230
I'm concerned about interrupt latency. Can I modify Salvo to disable only certain
interrupts during critical sections of code?..................................................................231
How big are the Salvo functions I might call from within an interrupt? ........................231

viii

Contents

Salvo User Manual

Why did my interrupt service routine grow and become slower when I added a call to
OSTimer()?..................................................................................................................232
My application can't afford the overhead of signaling from an ISR. How can I get
around this problem? ...................................................................................................232
Building Projects ...................................................................................................................233
What warning level should I use when building Salvo projects? ...................................233
What optimization level should I use when building Salvo projects? ............................233
Miscellaneous........................................................................................................................233
Can Salvo run on a 12-bit PICmicro with only a 2-level call…return stack?.................233
Will Salvo change my approach to embedded programming? .......................................233

Chapter 7 • Reference .................................................................................... 235
Run-Time Architecture..........................................................................................................235
Rule #1: Every Task Needs a Context Switch ................................................................235
Rule #2: Context Switches May Only Occur in Tasks ...................................................236
Rule #3: Persistent Local Variables Must be Declared as Static ....................................237
User Services.........................................................................................................................240
OS_Delay(): Delay the Current Task and Context-switch.............................................243
OS_DelayTS(): Delay the Current Task Relative to its Timestamp and Contextswitch...........................................................................................................................245
OS_Destroy(): Destroy the Current Task and Context-switch .......................................247
OS_Replace(): Replace the Current Task and Context-switch.......................................249
OS_SetPrio(): Change the Current Task's Priority and Context-switch .........................251
OS_Stop(): Stop the Current Task and Context-switch ..................................................253
OS_WaitBinSem(): Context-switch and Wait the Current Task on a Binary
Semaphore ...................................................................................................................255
OS_WaitEFlag(): Context-switch and Wait the Current Task on an Event Flag............257
OS_WaitMsg(): Context-switch and Wait the Current Task on a Message ...................261
OS_WaitMsgQ(): Context-switch and Wait the Current Task on a Message Queue .....263
OS_WaitSem(): Context-switch and Wait the Current Task on a Semaphore ...............265
OS_Yield(): Context-switch ...........................................................................................267
OSClrEFlag(): Clear Event Flag Bit(s)...........................................................................269
OSCreateBinSem(): Create a Binary Semaphore ...........................................................271
OSCreateCycTmr(): Create a Cyclic Timer ...................................................................273
OSCreateEFlag(): Create an Event Flag .........................................................................275
OSCreateMsg(): Create a Message.................................................................................277
OSCreateMsgQ(): Create a Message Queue...................................................................279
OSCreateSem(): Create a Semaphore.............................................................................281
OSCreateTask(): Create and Start a Task .......................................................................283
OSDestroyCycTmr(): Destroy a Cyclic Timer ...............................................................285
OSDestroyTask(): Destroy a Task ..................................................................................287
OSGetPrio(): Return the Current Task's Priority ............................................................289
OSGetPrioTask(): Return the Specified Task's Priority .................................................291
OSGetState(): Return the Current Task's State...............................................................293
OSGetStateTask(): Return the Specified Task's State ....................................................295
OSGetTicks(): Return the System Timer........................................................................297
OSGetTS(): Return the Current Task's Timestamp ........................................................299
OSInit(): Prepare for Multitasking..................................................................................301
OSMsgQCount(): Return Number of Messages in Message Queue...............................303
OSMsgQEmpty(): Check for Available Space in Message Queue.................................305

Salvo User Manual

Contents

ix

OSReadBinSem(): Obtain a Binary Semaphore Unconditionally ..................................307
OSReadEFlag(): Obtain an Event Flag Unconditionally................................................309
OSReadMsg():Obtain a Message's Message Pointer Unconditionally ...........................311
OSReadMsgQ(): Obtain a Message Queue's Message Pointer Unconditionally............313
OSReadSem(): Obtain a Semaphore Unconditionally....................................................315
OSResetCycTmr(): Reset a Cyclic Timer.......................................................................317
OSRpt(): Display the Status of all Tasks, Events, Queues and Counters .......................319
OSSched(): Run the Highest-Priority Eligible Task.......................................................321
OSSetCycTmrPeriod(): Set a Cyclic Timer's Period ......................................................323
OSSetEFlag(): Set Event Flag Bit(s) ..............................................................................325
OSSetPrio(): Change the Current Task's Priority ...........................................................327
OSSetPrioTask(): Change a Task's Priority....................................................................329
OSSetTicks(): Initialize the System Timer .....................................................................331
OSSetTS(): Initialize the Current Task's Timestamp......................................................333
OSSignalBinSem(): Signal a Binary Semaphore............................................................335
OSSignalMsg(): Send a Message....................................................................................337
OSSignalMsgQ(): Send a Message via a Message Queue..............................................339
OSSignalSem(): Signal a Semaphore .............................................................................341
OSStartCycTmr(): Start a Cyclic Timer .........................................................................343
OSStartTask(): Make a Task Eligible To Run................................................................345
OSStopCycTmr(): Stop a Cyclic Timer..........................................................................347
OSStopTask(): Stop a Task.............................................................................................349
OSSyncTS(): Synchronize the Current Task's Timestamp .............................................351
OSTimer(): Run the Timer..............................................................................................353
OSTryBinSem(): Obtain a Binary Semaphore if Available............................................355
OSTryMsg(): Obtain a Message if Available .................................................................357
OSTryMsgQ(): Obtain a Message from a Message Queue if Available ........................359
OSTrySem(): Obtain a Semaphore if Available .............................................................361
Additional User Services.......................................................................................................363
OSAnyEligibleTasks (): Check for Eligible Tasks .........................................................363
OScTcbExt0|1|2|3|4|5, OStcbExt0|1|2|3|4|5(): Return a Tcb Extension..........................365
OSCycTmrRunning(): Check Cyclic Timer for Running...............................................367
OSProtect(), OSUnprotect(): Protect Services Against Corruption by ISR....................369
OSTaskStopped(): Check whether Task has Stopped.....................................................371
OSTimedOut(): Check for Timeout................................................................................372
OSVersion(), OSVERSION: Return Version as Integer ................................................374
User Macros ..........................................................................................................................376
OSECBP(), OSEFCBP(),OSMQCBP(), OSTCBP(): Return a Control Block Pointer ..376
User-Defined Services...........................................................................................................378
OSDisableIntsHook(), OSEnableIntsHook(): Interrupt-control Hooks..........................378
OSIdlingHook(): Idle Function Hook.............................................................................380
OSSchedDispatchHook(), OSSchedEntryHook(), OSSchedReturnHook(): Scheduler
Hooks...........................................................................................................................382
Return Codes .........................................................................................................................384
Salvo Defined Types .............................................................................................................384
Salvo Variables......................................................................................................................388
Salvo Source Code ................................................................................................................389
Locations of Salvo Functions ................................................................................................391
Abbreviations Used by Salvo ................................................................................................393

x

Contents

Salvo User Manual

Chapter 8 • Libraries....................................................................................... 395
Library Types ........................................................................................................................395
Libraries for Different Environments....................................................................................395
Native Compilers ............................................................................................................395
Non-native Compilers.....................................................................................................396
Using the Libraries ................................................................................................................396
Overriding Default RAM Settings ..................................................................................397
Library Functionality.............................................................................................................398
Types...............................................................................................................................399
Memory Models..............................................................................................................399
Options............................................................................................................................399
Global Variables .............................................................................................................399
Configurations ................................................................................................................400
Variants...........................................................................................................................401
Library Reference..................................................................................................................403
Rebuilding the Libraries........................................................................................................403
GNU Make and the bash Shell........................................................................................404
Rebuilding Salvo Libraries .............................................................................................404
Linux/Unix Environment .........................................................................................404
Multiple Compiler Versions ...........................................................................................405
Win32 Environment........................................................................................................405
Customizing the Libraries...............................................................................................406
Creating a Custom Library Configuration File ........................................................406
Building the Custom Library....................................................................................407
Using the Custom Library in a Library Build ..........................................................407
Example – Custom Library with 16-bit Delays and Non-Zero Prescalar.................407
Preserving a User's salvoclcN.h Files.......................................................................409
Restoring the Standard Libraries..............................................................................409
Custom Libraries for non-Salvo Pro Users ..............................................................409
Makefile Descriptions.....................................................................................................409
Pumpkin\Salvo\Src\Makefile ...................................................................................409
Pumpkin\Salvo\Src\Makefile2 .................................................................................410
Pumpkin\Salvo\Src\CODE\Makefile .......................................................................410
Pumpkin\Salvo\Src\CODE\targets.mk .....................................................................410

Chapter 9 • Performance................................................................................ 411
Introduction ...........................................................................................................................411
Interrupts ...............................................................................................................................411
Context Switcher.............................................................................................................411
Summary ..................................................................................................................412
Critical Sections..............................................................................................................412
Effect on Runtime Performance ...............................................................................413
Controlling Interrupts Globally ................................................................................414
Controlling Interrupts Individually ..........................................................................415
Avoiding Interrupt Control Altogether.....................................................................417
Side Effects of Interrupt Hooks................................................................................420
The Fallacy of Avoiding Critical Sections at the Interrupt Level ............................421
User Hooks ............................................................................................................................422
OSDisableHook(), OSEnableHook()..............................................................................422

Salvo User Manual

Contents

xi

OSClrWDTHook()..........................................................................................................422

Chapter 10 • Porting ....................................................................................... 425
Chapter 11 • Tips, Tricks and Troubleshooting ........................................... 427
Introduction ...........................................................................................................................427
Compile-Time Troubleshooting ............................................................................................428
I'm just starting, and I'm getting lots of errors. ...............................................................428
My compiler can't find salvo.h. ......................................................................................428
My compiler can't find salvocfg.h. .................................................................................428
My compiler can't find certain target-specific header files.............................................428
My compiler can't locate a particular Salvo service. ......................................................428
My compiler has issued an "undefined symbol" error for a context-switching label
that I've defined properly.............................................................................................429
My compiler is saying something about OSIdlingHook.................................................429
My compiler has no command-line tools. Can I still build a library?.............................429
Run-Time Troubleshooting ...................................................................................................430
Nothing's happening. ......................................................................................................430
It only works if I single-step through my program. ........................................................431
It still doesn't work. How should I begin debugging?.....................................................431
My program's behavior still doesn't make any sense. .....................................................432
Compiler Issues .....................................................................................................................432
Where can I get a free C compiler? ................................................................................432
Where can I get a free make utility? ...............................................................................433
Where can I get a Linux/Unix-like shell for my Windows PC? .....................................433
My compiler behaves strangely when I'm compiling from the DOS command line,
e.g. "This program has performed an illegal operation and will be terminated."........433
My compiler is issuing redeclaration errors when I compile my program with Salvo's
source files...................................................................................................................434
HI-TECH PICC Compiler ..............................................................................................434
Running HPDPIC under Windows 2000 Pro ...........................................................434
Setting PICC Error/Warning Format under Windows 2000 Pro..............................435
Linker reports fixup errors .......................................................................................435
Placing variables in RAM ........................................................................................436
Link errors when working with libraries..................................................................436
Avoiding absolute file pathnames ............................................................................436
Compiled code doesn't work ....................................................................................437
PIC17CXXX pointer passing bugs...........................................................................437
While() statements and context switches .................................................................437
Library generation in HPDPIC.................................................................................437
Problems banking Salvo variables on 12-bit devices ...............................................438
Working with Salvo messages .................................................................................438
Adding OSTimer() to an Interrupt Service Routine .................................................438
Using the interrupt_level pragma .............................................................................440
HI-TECH V8C Compiler................................................................................................440
Simulators.................................................................................................................440
HI-TECH 8051C Compiler.............................................................................................441
Problems with static initialization and small and medium memory models. ...........441
IAR PICC Compiler........................................................................................................441

xii

Contents

Salvo User Manual

Target-specific header files ......................................................................................441
Interrupts ..................................................................................................................441
Mix Power C Compiler...................................................................................................442
Required compile options.........................................................................................442
Application crashes after adding long C source lines to a Salvo task ......................442
Application crashes after adding complex expressions to a Salvo task ...................443
Application crashes when compiling with /t option .................................................444
Compiler crashes when using a make system ..........................................................444
Metrowerks CodeWarrior Compiler ...............................................................................444
Compiler has a fatal internal error when compiling your source code.....................444
Microchip MPLAB .........................................................................................................445
The Stack window shows nested interrupts..............................................................445
Controlling the Size of your Application ..............................................................................445
Working with Message Pointers............................................................................................446

Appendix A • Recommended Reading.......................................................... 449
Salvo Publications .................................................................................................................449
Learning C.............................................................................................................................449
K&R................................................................................................................................449
C, A Reference Manual ..................................................................................................449
Power C...........................................................................................................................449
Real-time Kernels..................................................................................................................450
µC/OS & MicroC/OS-II..................................................................................................450
CTask..............................................................................................................................450
Embedded Programming.......................................................................................................450
RTOS Issues ..........................................................................................................................451
Priority Inversions...........................................................................................................451
Microcontrollers ....................................................................................................................451
PIC16 ..............................................................................................................................451

Appendix B • Other Resources...................................................................... 453
Web Links to Other Resources..............................................................................................453

Appendix C • File and Program Descriptions .............................................. 457
Overview ...............................................................................................................................457
Online File Locations ............................................................................................................457
Salvo Distributions .........................................................................................................457
Local/User File Locations .....................................................................................................458
Salvo Uninstaller(s) ........................................................................................................458
Salvo Documentation......................................................................................................458
Salvo Header Files ..........................................................................................................458
Salvo Source Files ..........................................................................................................458
Salvo Libraries................................................................................................................459
Salvo Applications..........................................................................................................459
Salvo Graphics Files .......................................................................................................459
Other Pumpkin Products.................................................................................................459
Target and Compiler Abbreviations ......................................................................................459
Projects ..................................................................................................................................460

Salvo User Manual

Contents

xiii

Nomenclature..................................................................................................................460
Project Files ....................................................................................................................461

Index ................................................................................................................ 463

xiv

Contents

Salvo User Manual

Figures
Figure 1: Foreground / Background Processing ............................................................................ 14
Figure 2: Interrupts Can Occur While Tasks Are Running............................................................ 18
Figure 3: Preemptive Scheduling................................................................................................... 19
Figure 4: Cooperative Scheduling ................................................................................................. 20
Figure 5: Task States...................................................................................................................... 23
Figure 6: Binary and Counting Semaphores .................................................................................. 29
Figure 7: Signaling a Binary Semaphore ....................................................................................... 30
Figure 8: Waiting a Binary Semaphore When the Event Has Already Occurred .......................... 30
Figure 9: Signaling a Binary Semaphore When a Task is Waiting for the Corresponding
Event........................................................................................................................................... 31
Figure 10: Synchronizing Two Tasks with Event Flags ................................................................ 32
Figure 11: Using a Counting Semaphore to Implement a Ring Buffer.......................................... 34
Figure 12: Signaling a Message with a Pointer to the Message's Contents ................................... 36
Figure 13: Welcome Screen........................................................................................................... 51
Figure 14: Salvo License Agreement Screen................................................................................. 52
Figure 15: Choose Components Screen......................................................................................... 53
Figure 16: Choose Destination Location Screen............................................................................ 54
Figure 17: Choose Start Menu Folder Screen................................................................................ 55
Figure 18: Installation Complete Screen........................................................................................ 55
Figure 19: Finish Screen ................................................................................................................ 56
Figure 20: Typical Salvo Install Directory Contents
(Lib Subdirectory View)...................... 57
Figure 21: Location of the Uninstaller(s)....................................................................................... 58
Figure 22: Confirming the Uninstall Operation............................................................................. 58
Figure 23: Uninstallation Complete Screen................................................................................... 59
Figure 24: Uninstall Complete Screen........................................................................................... 59
Figure 25: Salvo Library Build Overview ..................................................................................... 95
Figure 26: Salvo Source-Code Build Overview ............................................................................ 97
Figure 27: How to call OSCreateBinSem() when OSCALL_OSCREATEEVENT is set to
OSFROM_BACKGROUND ................................................................................................... 119
Figure 28: How to call OSCreateBinSem() when OSCALL_OSCREATEBINSEM is set to
OSFROM_FOREGROUND .................................................................................................... 119
Figure 29: How to call OSCreateBinSem() when OSCALL_CREATEBINSEM is set to
OSFROM_ANYWHERE......................................................................................................... 120
Figure 30: Tcb Extension Example Program Output................................................................... 150
Figure 31: OSRpt() Output to Terminal Screen........................................................................... 320

Salvo User Manual

xv

xvi

Figures

Salvo User Manual

Listings
Listing 1: A Simple Salvo Program ................................................................................................. 4
Listing 2: C Compiler Feature Requirements .................................................................................. 7
Listing 3: Reentrancy Errors with printf() ..................................................................................... 15
Listing 4: Task Structure for Preemptive Multitasking.................................................................. 21
Listing 5: Task Structure for Cooperative Multitasking ................................................................ 22
Listing 6: Delay Loop .................................................................................................................... 24
Listing 7: Delaying via the RTOS ................................................................................................. 26
Listing 8: Examples of Events ....................................................................................................... 27
Listing 9: Task Synchronization with Binary Semaphores............................................................ 32
Listing 10: Using a Binary Semaphore to Control Access to a Resource...................................... 33
Listing 11: Using a Counting Semaphore to Control Access to a Resource.................................. 35
Listing 12: Signaling a Message with a Pointer............................................................................. 36
Listing 13: Receiving a Message and Operating on its Contents................................................... 37
Listing 14: Vending Machine Superloop....................................................................................... 40
Listing 15: Task Version of ReleaseItem() .................................................................................... 44
Listing 16: Task Version of CallPolice() ....................................................................................... 44
Listing 17: Prioritizing a Task ....................................................................................................... 45
Listing 18: Creating a Message Event ........................................................................................... 45
Listing 19: Calling the System Timer ............................................................................................ 45
Listing 20: Starting all Tasks ......................................................................................................... 46
Listing 21: Multitasking Begins..................................................................................................... 46
Listing 22: RTOS-based Vending Machine................................................................................... 49
Listing 23: A Minimal Salvo Application ..................................................................................... 64
Listing 24: A Multitasking Salvo Application with two Tasks...................................................... 65
Listing 25: Multitasking with two Non-trivial Tasks..................................................................... 68
Listing 26: Multitasking with an Event ......................................................................................... 71
Listing 27: Multitasking with a Delay ........................................................................................... 76
Listing 28: Calling OSTimer() at the System Tick Rate................................................................ 76
Listing 29: Signaling from Multiple Tasks .................................................................................... 79
Listing 30: salvocfg.h for Tutorial Program .................................................................................. 90
Listing 31: Tcb Extension Example............................................................................................. 149
Listing 32: Task with a Proper Context Switch ........................................................................... 235
Listing 33: Tasks that Fail to Context Switch.............................................................................. 236
Listing 34: Incorrectly Context-Switching Outside of a Task ..................................................... 237
Listing 35: Task Using Persistent Local Variable ....................................................................... 238
Listing 36: Task Using Auto Local Variables ............................................................................. 239
Listing 37: Source Code Files...................................................................................................... 391
Listing 38: Location of Functions in Source Code ...................................................................... 393
Listing 39: List of Abbreviations................................................................................................. 394
Listing 40: Example salvocfg.h for Use with Standard Library .................................................. 397
Listing 41: Example salvocfg.h for Use with Standard Library and Reduced Number of Tasks 397
Listing 42: Additional Lines in salvocfg.h for Reducing Memory Usage with Salvo Libraries . 398
Listing 43: Partial Listing of Services than can be called from Interrupts................................... 402
Listing 44: Making a Single Salvo Library.................................................................................. 404
Listing 45: Making all Salvo Libraries for a Particular Compiler ............................................... 404

Salvo User Manual

xvii

Listing 46: Making all Salvo Libraries for a Particular Target.................................................... 405
Listing 47: Obtaining a List of Library Targets in the Makefile.................................................. 405
Listing 48: Making Salvo Libraries for IAR's MSP430 C Compiler v2.x................................... 405
Listing 49: Example Custom Library Configuration File salvoclc4.h......................................... 407
Listing 50: Making a Custom Salvo Library with Custom Library Configuration 4................... 408
Listing 51: Example salvocfg.h for Library Build Using Custom Library Configuration 4 and
Archelon / Quadravox AQ430 Development Tools ................................................................. 408
Listing 52: Making a Custom Salvo Library with Custom Library Configuration 4................... 408
Listing 53: Use of interrupt hooks in Salvo source code. ............................................................ 413
Listing 54: Most general configuration for Salvo's interrupt hooks. ........................................... 414
Listing 55: Application-specific configuration for Salvo's interrupt hooks. Relevant ISRs also
shown. Target is TI's MSP430FG4619..................................................................................... 416
Listing 56: Interrupt hooks for applications that do not call Salvo services from any
interrupts................................................................................................................................... 417
Listing 57: Passing interrupt activity up from an ISR to call a Salvo service without a
corresponding interrupt hook. Target is Microchip PIC18F452............................................... 418
Listing 58: Interrupt hooks to avoid interrupt nesting. ................................................................ 421
Listing 59: Example watchdog hook. Target is TI's MSP430F1612. .......................................... 423

xviii

Listings

Salvo User Manual

Tables
Table 1: Allowable Storage Types / Type Qualifiers for Salvo Objects...................................... 155
Table 2: Configuration Options by Category............................................................................... 186
Table 3: Configuration Options by Desired Feature.................................................................... 188
Table 4: Predefined Symbols ....................................................................................................... 189
Table 5: Return Codes ................................................................................................................. 384
Table 6: Normal Types ................................................................................................................ 386
Table 7: Normal Pointer Types.................................................................................................... 386
Table 8: Qualified Types ............................................................................................................. 387
Table 9: Qualified Pointer Types................................................................................................. 387
Table 10: Salvo Variables............................................................................................................ 389
Table 11: Type Codes for Salvo Libraries................................................................................... 399
Table 12: Configuration Codes for Salvo Libraries..................................................................... 400
Table 13: Features Common to all Salvo Library Configurations............................................... 401
Table 14: Variant Codes for Salvo Libraries ............................................................................... 403

Salvo User Manual

xix

xx

Tables

Salvo User Manual

Release Notes

Introduction
What's New
Please refer to the distribution's salvo-whatsnew.txt file for
more information on what's new in the v4.2.2 release.

Release Notes
Please refer to the general (salvo-release.txt) and distributionspecific (salvo-release-targetname.txt) release notes for
more information on release-related changes and updates in the
v4.2.2 release.

Third-Party Tool Versions
Please

to the distribution-specific (salvo-releasetargetname.txt) release notes for the version numbers of thirdparty tools (compilers, linkers, librarians, etc.) in the v4.2.2 release.

Salvo User Manual

refer

xxi

xxii

Release Notes

Salvo User Manual

Supported Targets and Compilers
As of v4.2.2, Salvo supports a variety of 8-, 16- and 32-bit targets
and compilers:
Please refer to the distribution-specific (salvo-releasetargetname.txt) release notes for the version numbers of thirdparty tools (compilers, linkers, librarians, etc.) in the v4.2.2 release. If you have a named compiler that is older than the ones
listed, you may need to upgrade it to work with Salvo. Contact the
compiler vendor for upgrade information.

Salvo User Manual

xxiii

Preface

Historical Information
Pumpkin, Inc.'s Salvo v1 was an internal release, written in assembly language and targeted specifically for the Microchip
PIC17C756 PICmicro® MCU in a proprietary, in-house data acquisition system. This 1998 version provided much of the basic functionality that would later makes its way into the later Salvo
releases.
After a market analysis Pumpkin, Inc. decided to expand on Salvo
v1's functionality by rewriting it in C. In doing so, opportunities
arose for many configuration options and optimizations, to the
point where not only was the C version more powerful and flexible
than its assembly-language predecessor, but it was completely
portable, too.
In 2000, Salvo v2 became the first commercial release of Pumpkin,
Inc.'s cooperative priority-based multitasking RTOS. It was targeted at the entire range of Microchip PICmicro® MCUs.
In 2002, Salvo v3 was released. This marked the expansion of the
Salvo RTOS into new embedded targets, like the 8-bit 8051 and
the 16-bit MSP430.
Salvo 4 was released in 2005. Not only did this release mark the
first Salvo support for 32-bit embedded targets, but it also included
many of the lessons learned over the previous six years in terms of
usability and maximum configurability for high performance.
Salvo 4 is the first Salvo release to remove all non-instruction-set
hardware dependencies from the core Salvo code. This gives
end-users complete flexibility in configuring Salvo for maximum
real-time performance.

Typographic Conventions
Various text styles are used throughout this manual to improve
legibility. Code examples, code excerpts, path names and file
names are shown in a monospaced font. New and particularly
useful terms, and terms requiring emphasis, are shown italicized.
Salvo User Manual

xxv

User input (e.g. at the DOS command line) is shown in this manner. Certain terms and sequence numbers are shown in bold. Important notes, cautions and warnings have distinct borders around
them:

Note Salvo source code uses tab settings of 4, i.e. tabs are
equivalent to 4 spaces.
The letters xyz are used to denote one of several possible names,
e.g. OSSignalXyz() refers to OSSignalBinSem(), OSSignalMsg(), OSSignalMsgQ(), OSSignalSem(), etc. Xyz is caseinsensitive.
The symbol | is used as a shorthand to denote multiple, similar
names, e.g. sysa|e|f denotes sysa and/or syse and/or sysf.
DOS and Windows pathnames use '\'. Linux and Unix pathnames
use '/'. They are used interchangeably throughout this document.

Standardized Numbering Scheme
Salvo employs a standardized numbering scheme for all software
releases. The version/revision numbering scheme uses multiple
fields1 as shown below:
salvo-distribution-targetMAJOR.MINOR.SUBMINOR[-PATCH]

where
•

distribution

refers to Salvo Lite, tiny, SE, LE

or Pro
•

•
•
•
•

1

xxvi

target refers to the target processor(s)
supported in the distribution
MAJOR changes when major features (e.g. array
mode) are added.
MINOR changes when minor features (e.g. new
user services) are added to or changed.
SUBMINOR changes during alpha and beta testing
and when support files (e.g. new Salvo
Application Notes) are added.
PATCH is present and changes each time a bug
fix is applied and/or new documentation is

The final field is present only on patches.

Preface

Salvo User Manual

added. PATCH may also be used for release
candidates, e.g. rc4.
All MAJOR.MINOR.SUBMINOR versions are released with their own,
complete installer. -PATCH may be used on complete installers or
on minimal installers or archives that add new or modified files to
an existing Salvo code and documentation installation.
Examples include:
•
salvo-litepic-2.2.0

•
3.1.0-rc3

•

•
salvo-le-8051- •

•
salvo-promsp430-4.1.0

•
•

v2.2 Salvo
Lite for PICmicro®
MCUs installer,
released
v3.1.0 Salvo
LE for 8051 family
installer, release
candidate #3
version
4.1.0 Salvo Pro for
TI's MSP430 installer,
released

•
Salvo releases are generically referred to by their MAJOR.MINOR
numbering, i.e. "the 3.0 release."

The Salvo Coding Mindset
Configurability Is King
Salvo is extremely configurable to meet the requirements of the
widest possible target audience of embedded microcontrollers. It
also provides you, the user, with all the necessary header files, user
hooks, predefined constants, data types, useable functions, etc. that
will enable you to create your own Salvo application as quickly
and as error-free as possible.

Salvo User Manual

Preface

xxvii

Conserve Precious Resources
The Salvo source code is written first and foremost to use as few
resources as possible in the target application. Resources include
RAM, ROM, stack call…return levels and instruction cycles. Most
of Salvo's RAM- and ROM-hungry functionality is disabled by default. If you want a particular feature (e.g. event flags), you must
enable it via a configuration option (e.g. OSENABLE_EVENT_FLAGS)
and re-make your application. This allows you to manage the
Salvo code in your application from a single point – the Salvo configuration file salvocfg.h.

Learn to Love the Preprocessor
Salvo makes heavy use of the C preprocessor and symbols predefined by the compiler, Salvo and/or the user in order to configure
the source code for compilation. Though this may seem somewhat
daunting at first, you'll find that it makes managing Salvo projects
much simpler.

Document, But Don't Duplicate
Wherever possible, neither source code nor documentation is repeated in Salvo. This makes it easier for us to maintain and test the
code, and provide accurate and up-to-date information.

We're Not Perfect
While every effort has been made to ensure that Salvo works as
advertised and without error, it's entirely possible that we may
have overlooked a problem or failed to catch a mistake. Should
you find what you think is an error or ambiguity, please contact us
so that we can resolve the issue(s) as quickly as possible and enable you to continue coding your Salvo applications worry-free.2

Note We feel that it should not be necessary for you to modify
the source code to achieve functionality close to what Salvo already provides. We urge you to contact us first with your questions
before modifying the source code, as we cannot support modified
versions of Salvo. In many instances, we can both propose a solution to your problem, and perhaps also incorporate it into the next
Salvo release.
2

xxviii

See Pumpkin Salvo Software License Agreement for more information.

Preface

Salvo User Manual

Chapter 1 • Introduction

Welcome
In the race to innovate, time-to-market is crucial in launching a
successful new product. If you don't take advantage of in-house or
commercially available software foundations and support tools,
your competition will. But cost is also an important issue, and with
silicon (as in real life) prices go up as things get bigger. If your design can afford lots memory and maybe a big microprocessor, too,
go out and get those tools. That's what everybody else is doing …
But what if it can't?
What if you've been asked to do the impossible – fit complex, realtime functionality into a low-cost microcontroller and do it all on a
tight schedule? What if your processor has only a few KB of ROM
and even less RAM? What if the only tools you have are a compiler, some debugging equipment, a couple of books and your
imagination? Are you really going to be stuck again with state machines, jump tables, complex interrupt schemes and code that you
can't explain to anyone else? After a while, that won't be much fun
anymore. Why should you be shut out of using the very same
software frameworks the big guys use?
They say that true multitasking needs plenty of memory, and it's
not an option for your design. But is that really true?
Not any more. Not with Salvo. Salvo is full-blown multitasking in
a surprisingly small memory space – it's about as big as
printf()!3 Multitasking, priorities, events, a system timer – it's all
in there. No interrupts available? That's not a problem, either.
You'll get more functionality out of your processor quicker than
you ever thought possible. And you can put Salvo to work for you
right away.

3

Salvo User Manual

Comparison based on implementations with full printf() functionality.

1

What Is Salvo?
Salvo is a proven, powerful, high-performance and royalty-free
real-time operating system (RTOS) that requires very little program and data memory, and no task stacks. It is an easy-to-use
software tool to help you quickly create powerful, reliable and sophisticated applications (programs) for embedded systems.
Salvo was designed from the ground up for use in microprocessors
and microcontrollers with severely limited resources, and will
typically require from 5 to 100 times less memory than other
RTOSes. In fact, Salvo's memory requirements are so minimal that
it will run where no other RTOS can.
Salvo is ROMable, easily scaleable and extremely portable. It runs
on just about any processor, from a PIC to a Pentium.

Why Should I Use Salvo?
If you're designing the next hot embedded product, you know that
time-to-market is crucial to guarantee success. Salvo provides a
powerful and flexible framework upon which you can quickly
build your application.
If you're faced with a complex design and limited processing resources, Salvo can help you make the most of what's available in
your system.
And if you're trying to cost-reduce or add functionality to an existing design, Salvo may be what you need because it helps you leverage the processing power you already have.
Before Salvo, embedded systems programmers could only dream
of running an RTOS in their low-end processors. They were locked
out of the benefits that an RTOS can bring to a project, including
reducing time-to-market, managing complexity, enhancing robustness and improving code sharing and re-use. They were unable to
take advantage of the many well-established RTOS features designed to solve common and recurring problems in embedded systems programming.
That dream is now a reality. With Salvo, you can stop worrying
about the underlying structure and reliability of your program and
start focusing on the application itself.

2

Chapter 1 • Introduction

Salvo User Manual

What Kind of RTOS Is Salvo?
Salvo is a purely event-driven cooperative multitasking RTOS,
with full support for event and timer services. Multitasking is priority-based, with sixteen separate priority levels supported. Tasks
that share the same priority will execute in a round-robin fashion.
Salvo provides services for employing semaphores (binary and
counting), messages, message queues and event flags for intertask
communications and resource management. A full complement of
RTOS functions (e.g. context-switch, stop a task, wait on a semaphore, etc.) is supported. Timer functions, including delays, timeouts and cyclic timers, are also supported.
Salvo is written in ANSI C, with a very small number of processor-specific extensions, some of which are written in native assembly language. It is highly configurable to support the unique
demands of your particular application.
While Salvo is targeted towards embedded applications, it is universally applicable and can also be used to create applications for
other types of systems (e.g. 16-bit DOS applications).

What Does a Salvo Program Look Like?
A Salvo program looks a lot like any other that runs under a multitasking RTOS. Listing 1 shows (with comments) the source code
for a remote automotive seat warmer with user-settable temperature. The microcontroller is integrated into the seat, and requires
just four wires for communication with the rest of the car's electronics – power, ground, Rx (to receive the desired seat temperature from a control mounted elsewhere) and Tx (to indicate status).
The desired temperature is maintained via TaskControl(). TaskStatus() sends, every second, either a single 50ms pulse to indicate that the seat has not yet warmed up, or two consecutive 50ms
pulses to indicate that the seat is at the desired temperature.
#include 
typedef unsigned char t_boolean;
typedef unsigned char t_temp;
/* Local flag.
t_boolean warm = FALSE;

*/

/* Seat temperature functions.
*/
extern t_temp UserTemp( void );
extern t_temp SeatTemp( void );
extern t_boolean CtrlTemp( t_temp user, seat );

Salvo User Manual

Chapter 1 • Introduction

3

/* Moderate-priority (i.e. 8) task (i.e. #1)
/* to maintain seat temperature. CtrlTemp()
/* returns TRUE only if the seat is at the
/* the desired (user) temperature.
void TaskControl( void )
{
while (1) {
warm = CtrlTemp(UserTemp(), SeatTemp());
OS_Yield();
}
}

*/
*/
*/
*/

/* High-priority (i.e. 3) task (i.e. #2) to
/* generate pulses. System ticks are 10ms.
void TaskStatus( void )
{
/* initialize pulse output (low).
TX_PORT &= ~0x01;

*/
*/

*/

while (1) {
OS_Delay(100);
TX_PORT |= 0x01;
OS_Delay(5);
TX_PORT &= ~0x01;
if (warm) {
OS_Delay(5);
TX_PORT |= 0x01;
OS_Delay(5);
TX_PORT &= ~0x01;
}
}
}
/* Initialize Salvo, create and assign
/* priorities to the tasks, and begin
/* multitasking.
int main( void )
{
OSInit();

*/
*/
*/

OSCreateTask(TaskControl, OSTCBP(1), 8);
OSCreateTask(TaskStatus, OSTCBP(2), 3);
while (1) {
OSSched();
}
}
Listing 1: A Simple Salvo Program

It's important to note that when this program runs, temperature
control continues while TaskStatus() is delayed. The calls to
OS_Delay() do not cause the program to loop for some amount of
time and then continue. After all, that would be a waste of processor resources (i.e. instruction cycles). Instead, those calls simply

4

Chapter 1 • Introduction

Salvo User Manual

instruct Salvo to suspend the pulse generator and ensure that it resumes running after the specified time period. TaskControl()
runs whenever TaskStatus() is suspended.
Apart from creating a simple Salvo configuration file and tying
Salvo's timer to a 10ms periodic interrupt in your system, the C
code above is all that is needed to run these two tasks concurrently.
Imagine how easy it is to add more tasks to this application to enhance its functionality.
See Chapter 4 • Tutorial for more information on programming
with Salvo.

What Resources Does Salvo Require?
The amount of ROM Salvo requires will depend on how much of
Salvo you are using. A minimal multitasking application on an 8bit RISC processor might use a few hundred instructions. A fullblown Salvo application on the same processor will use around 1K
instructions.
In conventional RTOSes, a large amount of RAM is dedicated to
the individual task stacks. Since Salvo does not need or maintain
task tasks, its RAM requirements are commensurately (much)
smaller, since C compilers may use the stack for local/auto variables, function parameters, etc. Also because of this fundamental
design aspect of Salvo, Salvo can run on targets that have limited
hardware call/return stacks4 instead of a more common generalpurpose stack.
The amount of RAM Salvo requires is also dependent on your particular configuration. In an 8-bit RISC application,5 each task will
require 4-12 (typically 7) bytes, each event 3-4 bytes,6 and 4-6
more bytes are required to manage all the tasks, events and delays.
That's it!
In all cases, the amount of RAM required is primarily dependent
on the size of pointers (i.e. 8, 16 or 32 bits) to ROM and RAM in
4

5
6

Salvo User Manual

A hardware call/return stack is used only to store the caller function’s return
address, and is limited to some depth (e.g. 16 levels on PIC17 processors). A
hardware call/retun stack cannot be used for local (auto variables), for
example. Additionally, processors with hardware call/return stacks do not
implement PUSH and POP instructions.
PIC16 series (e.g. PIC16C64). Pointers to ROM take two bytes, and pointers
to RAM take one byte.
Message queues require additional RAM.

Chapter 1 • Introduction

5

your application, i.e. it's application-dependent. In some applications (e.g. CISC processors) additional RAM may be required for
general-purpose register storage.
If you plan to use the delay and timeout services, Salvo requires
that you call its timer service at a regular rate. While there are noninterrupt-driven ways of achieving this, this requirement is often
satisfied by calling the timer service via a single interrupt. However, this interrupt need not be dedicated to Salvo – it can be used
for your own purposes, too.
The number of tasks and events is limited only by the amount of
available memory.
See Chapter 6 • Frequently Asked Questions (FAQ) for more information.

How Is Salvo Different?
Salvo is a cooperative RTOS that doesn't need a stack.7 Virtually
all other RTOSes use a stack, and many are preemptive as well as
cooperative. This means that compared to other RTOSes, Salvo
differs primarily in these ways:
•
•

•

7

8

6

• Salvo is a cooperative RTOS, so you must
explicitly manage task switching8.
• Task switching can only occur at the task
level, i.e. directly inside your tasks, and not
from within a function called by your task, or
elsewhere. This is due to the absence of a
general-purpose stack and the concomitant
ability of the RTOS to save task & state
information on the stack. This may have a small
impact on the structure of your program.
• Compared to other cooperative or preemptive
RTOSes, which need lots of RAM memory
(usually in the form of a general-purpose stack),
Salvo needs very little. For processors without
much RAM, Salvo may be your only RTOS
choice.

By "doesn’t need a stack" we mean that Salvo does not need RAM to store the
data that a conventional RTOS usually stores on the (general-purpose) stack,
including return addresses, local/auto variables, saved registers, and other
(usually compiler-dependent) task-specific data.
We'll explain this term later, but for now it means being in one task and
relinquishing control of the processor so that another task may run.

Chapter 1 • Introduction

Salvo User Manual

Salvo is able to provide most of the performance and features of a
full-blown RTOS while using only a fraction as much memory.
With Salvo you can quickly create powerful, fast, sophisticated
and robust multitasking applications.

What Do I Need to Use Salvo?
A working knowledge of C is recommended. But even if you're a
C beginner, you shouldn't have much difficulty learning to use
Salvo.
Some knowledge of RTOS fundamentals is useful, but not required. If working with an RTOS is new to you, be sure to review
Chapter 2 • RTOS Fundamentals.
You will need a good ANSI-C-compliant compiler for the processor(s) you're using. It must be capable of compiling the Salvo
source code, which makes use of many C features, including (but
not limited to):
•
•
•
•
•
•
•
•
•
•
•
•
•

• arrays,
• unions,
• bit fields,
• structures,
• static variables,
• multiple source files,
• indirect function calls,
• multiple levels of indirection,
• passing of all types of parameters,
• multiple bytes of parameter passing,
• extensive use of the C preprocessor,
• pointers to functions, arrays, structures,
unions, etc., and
• support for variable arguments lists9 (via
va_arg(), etc.)
Listing 2: C Compiler Feature Requirements

A compiler with the ability to perform in-line assembly is a plus.
The more fully-featured the in-line assembler, the better.

9

Salvo User Manual

This is not absolutely necessary, but is desireable. va_arg() is part of the
ANSI C standard.

Chapter 1 • Introduction

7

Lastly, your compiler should be capable of compiling to object
(*.o) modules and libraries (*.lib), and linking object modules
and libraries together to form a final executable (usually *.hex).
We recommend that you use a compiler that is already certified for
use with Salvo. If your favorite compiler and/or processor are not
yet supported and it meets Salvo’s requirements, you can probably
do a port to them in a few hours. Chapter 10 • Porting will guide
you through the process. Always check with the factory for the latest news concerning supported compilers and processors.

Which Processors and Compilers does Salvo
Support?
Please visit Pumpkin’s website for up-to-date information.

How Is Salvo Distributed?
Salvo is supplied on downloadable over the Internet as a Windows
95 / 98 / ME / NT / 2000 / XP installer program. After you install
Salvo onto your computer you will have a group of subdirectories
that contain the Salvo source code, Salvo libraries, Salvo examples, and various other support files.

What Is in this Manual?
Chapter 1 • Introduction is this chapter.
Chapter 2 • RTOS Fundamentals is an introduction to RTOS programming. If you're only familiar with traditional "superloop" or
"foreground / background" programming architectures, you should
definitely review this chapter.
Chapter 3 • Installation covers how to install Salvo onto your com-

puter.
Chapter 4 • Tutorial is a guide to using Salvo. It contains examples

to introduce you to all of Salvo's functionality and how to use it in
your application. Even programmers familiar with other RTOSes
should still review this chapter.
Chapter 5 • Configuration explains all of Salvo's configuration pa-

rameters. Beginners and experienced users need this information to

8

Chapter 1 • Introduction

Salvo User Manual

optimize Salvo's size and performance to their particular application.
Chapter 6 • Frequently Asked Questions (FAQ) contains answers

to many frequently asked questions.
Chapter 7 • Reference is a guide to all of Salvo's user services

(callable functions).
Chapter 8 • Libraries lists the available freeware and standard li-

braries and explains how to use them.
Chapter 9 • Performance has actual data on the size and speed of

Salvo in various configurations. It also has tips on how to characterize Salvo's performance in your particular system.
Chapter 10 • Porting covers the issues you'll face if you're porting

Salvo to a compiler and/or processor that is not yet formally certified or supported by Salvo.
Chapter 11 • Tips, Tricks and Troubleshooting has information on

a variety of problems you may encounter, and how to solve them.
Appendix A • Recommended Reading contains references to mul-

titasking and related documents.
Appendix B • Other Resources has information on other resources

that may be useful to you in conjunction with Salvo.
Appendix C • File and Program Descriptions contains descriptions

of all of the files and file types that are part of a Salvo installation.

Salvo User Manual

Chapter 1 • Introduction

9

10

Chapter 1 • Introduction

Salvo User Manual

Chapter 2 • RTOS Fundamentals
Note If you're already familiar with RTOS fundamentals you
may want to skip directly to Chapter 3 • Installation.

Introduction
•

"I've built polled systems. Yech. Worse are
applications that must deal with several
different things more or less concurrently,
without using multitasking. The software in both
situations is invariably a convoluted mess.
Twenty years ago, I naïvely built a steel
thickness gauge without an RTOS, only to later
have to shoehorn one in. Too many
asynchronous things were happening; the inline code grew to outlandish complexity." Jack
G. Ganssle10

Most programmers are familiar with traditional systems that employ a looping construct for the main part of the application and
use interrupts to handle time-critical events. These are so-called
foreground / background (or superloop) systems, where the interrupts run in the foreground (because they take priority over everything else) and the main loop runs in the background when no
interrupts are active. As applications grow in size and complexity
this approach loses its appeal because it becomes increasingly difficult to characterize the interaction between the foreground and
background.
An alternative method for structuring applications is to use a software framework that manages overall program execution according to a set of clearly defined rules. With these rules in place, the
application's performance can be characterized in a relatively
straightforward manner, regardless of its size and complexity.
Many embedded systems can benefit from using an approach involving the use of multiple, concurrent tasks communicating
amongst themselves, all managed by a kernel, and with clearly10

Salvo User Manual

"Interrupt Latency", Embedded Systems Programming, Vol. 14 No. 11,
October 2001, p. 73.

11

defined run-time behavior. This is the RTOS approach to programming. These and other terms are defined below.

Note This chapter is only a quick introduction to the operation
and use of an RTOS. Appendix A • Recommended Reading contains references for further, in-depth reading.

Basic Terms
A task is a sequence of instructions, sometimes done repetitively,
to perform an action (e.g. read a keypad, display a message on an
LCD, flash an LED or generate a waveform). In other words, it's
usually a small program inside a bigger one. When running on a
relatively simple processor (e.g. Z80, 68HC11, PIC), a task may
have all of the system's resources to itself regardless of how many
tasks are used in the application.
An interrupt is an internal or external hardware event that causes
program execution to be suspended. Interrupts must be enabled for
an interrupt to occur. When this occurs, the processor vectors to a
user-defined interrupt service routine (ISR), which runs to completion. Then program execution picks up where it left off. Because of
their ability to suspend program execution, interrupts are said to
run in the foreground, and the rest of the program runs in the background.
A task's priority suggests the task's importance relative to other
tasks. It may be fixed or variable, unique or shared with other
tasks.
A task switch occurs when one task suspends running and another
starts or resumes running. It may also be called a context switch,
because a task's context (generally the complete contents of the
stack and the values of the registers) is usually saved for re-use
when the task resumes.
Preemption occurs when a task is interrupted and another task is
made ready to run. An alternative to a preemptive system is a cooperative system, in which a task must voluntarily relinquish control of the processor before another task may run. It is up to the
programmer to structure the task so that this occurs. If a running
task fails to cooperate, then no other tasks will execute, and the
application will fail to work properly.

12

Chapter 2 • RTOS Fundamentals

Salvo User Manual

Preemptive and cooperative context switching are handled by a
kernel. Kernel software manages the switching of tasks (also called
scheduling) and intertask communication. A kernel generally ensures that the highest-priority eligible task is the task that's running
(preemptive scheduling) or will run next (cooperative scheduling).
Kernels are written to be as small and as fast as possible to guarantee high performance in the overlying application program.11
A delay is an amount of time (often specified in milliseconds) during which a task's execution can be suspended. While suspended, a
task should use as few of the processor's resources as possible to
maximize the performance of the overall application, which is
likely to include other tasks that are not concurrently suspended.
Once the delay has elapsed (or expired), the task resumes executing. The programmer specifies how long the delay is, and how often it occurs.
An event is an occurrence of something (e.g. a key was pressed, an
error occurred or an expected response failed to occur) that a task
can wait for. Also, just about any part of a program can signal the
occurrence of an event, thus letting others know that the event happened.
Intertask communication is an orderly means of passing information from one task to another following some well-established programming concepts. Semaphores, messages, message queues and
event flags can be used to pass information in one form or another
between tasks and, in some cases, ISRs.
A timeout is an amount of time (often specified in milliseconds)
that a task can wait for an event. Timeouts are optional – a task can
also wait for an event indefinitely. If a task specifies a timeout
when waiting for an event and the event doesn't occur, we say that
a timeout has occurred, and special handling is invoked.
A task's state describes what the task is currently doing. Tasks
change from one state to another via clearly defined rules. Common task states might be ready / eligible, running, delayed, waiting, stopped and destroyed / uninitialized.
The timer is another piece of software that keeps track of elapsed
time and/or real time for delays, timeouts and other time-related
services. The timer is only as accurate as the timer clock provided
by your system.
11

Salvo User Manual

Some kernels also provide I/O functions and other services such as memory
management. Those are not discussed here.

Chapter 2 • RTOS Fundamentals

13

A system is idling when there are no tasks to run.
The operating system (OS) contains the kernel, the timer and the
remaining software (called services) to handle tasks and events
(e.g. task creation, signaling of an event). One chooses a real-time
operating system (RTOS) when certain operations are critical and
must be completed correctly and within a certain amount of time.
An RTOS-enabled application or program is the end product of
combining your tasks, ISRs, data structures, etc, with an RTOS to
form single program.
Now let's examine all these terms, and some others, in more detail.

Foreground / Background Systems
The simplest program structure is one of a main loop (sometimes
called a superloop) calling functions in an ordered sequence. Because program execution can switch from the main loop to an ISR
and back, the main loop is said to run in the background, whereas
the ISRs run in the foreground. This is the sort of programming
that many beginners encounter when learning to program simple
systems.

3

9

ISR2
7 8 10

11

2 4

ISR1
superloop
functions

1

5

6

12

13
10

time
Figure 1: Foreground / Background Processing

In Figure 1 we see a group of functions repeated over and over [1,
5, 13] in a main loop. Interrupts may occur at any time, and even at
multiple levels. When an interrupt occurs (high-priority interrupt at
[2] and [8], low-priority interrupt at [6]), processing in the function

14

Chapter 2 • RTOS Fundamentals

Salvo User Manual

is suspended until the interrupt is finished, whereupon the program
returns to the main loop or to a previous interrupted ISR. The main
loop functions are executed in strictly serial manner, all at the same
priority, without any means of changing when or even if the function should execute. ISRs must be used in order to respond quickly
to external events, and can be prioritized if multiple interrupt levels
are supported.
Foreground / background systems are relatively simple from a programming standpoint as long as there is little interaction amongst
the functions in the main loop and between them and the ISRs. But
they have several drawbacks: Loop timing is affected by any
changes in the loop and/or ISR code. Also, the response of the system to inputs is poor because information made available by an
ISR to a function in the loop cannot be processed by the function
until its turn to execute. This rigidly sequential nature of program
execution in the super loop affords very little flexibility to the programmer, and complicates time-critical operations. State machines
may be used to partially solve this problem. As the application
grows, the loop timing becomes unpredictable, and a variety of
other complicating factors arise.

Reentrancy
One such factor is reentrancy. A reentrant function can be used
simultaneously in one or more parts of an application without corrupting data. If the function is not written to be reentrant, simultaneous calls may corrupt the function's internal data, with
unpredictable results in the application. For example, if an application has a non-reentrant printf() function and it is called both
from main loop code (i.e. the background) and also from within an
ISR (i.e. the foreground), there's an excellent chance that every
once in a while the resultant output of a call to
printf("Here we are in the main loop.\n");

from within the main loop and a call to
printf("Now we are servicing an interrupt.\n");

from within an ISR at the same time might be
Here we aNow we are servicing an interrupt.
Listing 3: Reentrancy Errors with printf()

Salvo User Manual

Chapter 2 • RTOS Fundamentals

15

This is clearly in error. What has happened is that the first instance
of printf() (called from within the main loop) got as far as printing the first 9 characters ("Here we a") of its string argument before being interrupted. The ISR also included a call to printf(),
which re-initialized its local variables and succeeded in printing its
entire 36-character string ("Now we … interrupt.\n"). After the
ISR finished, the main-loop printf() resumed where it had left
off, but its internal variables reflected having successfully written
to the end of a string argument, and no further output appeared
necessary, so it simply returned and the main loop continued executing.

Note Calling non-reentrant functions as if they were reentrant
rarely results in such benign behavior.
Various techniques can be employed to avoid this problem of a
non-reentrant printf(). One is to disable interrupts before calling
a non-reentrant function and to re-enable them thereafter. Another
is to rewrite printf() to only use local variables (i.e. variables
that are kept on the function's stack). The stack plays a very important role in reentrant functions.

Resources
A resource is something within your program that can be used by
other parts of the program. A resource might be a register, a variable or a data structure, or it might be something physical like an
LCD or a beeper. A shared resource is a resource that may be used
by more than one part of your program. If two separate parts of a
program are contending for the same resource, you'll need to manage this by mutual exclusion. Whenever a part of your program
wants to use the resource it must obtain exclusive access to it in
order to avoid corrupting it.

Multitasking and Context Switching
Many advantages can be realized by splitting a foreground / background application into one with multiple, independent tasks. In
order to multitask, such that all tasks appear to run concurrently,
some mechanism must exist to pass control of the processor and its
resources from one task to another. This is the job of the scheduler,
part of the kernel that (among its other duties) suspends one task
and resumes another when certain conditions are met. It does this
by storing the program counter for one task and restoring the pro-

16

Chapter 2 • RTOS Fundamentals

Salvo User Manual

gram counter for another. The faster the scheduler is able to switch
tasks, the better the performance of the overall application, since
the time spent switching tasks is time spent without any tasks running.
A context switch must appear transparent to the task itself. The
task's "world view" before the context switch that suspends it and
after the context switch that resumes it must be the same. This
way, task A can be interrupted at any time to allow the scheduler to
run a higher-priority task, task B. Once task B is finished, task A
resumes where it left off. The only effect of the context switch on
task A is that it was suspended for a potentially long time as a result of the context switch. Hence tasks that have time-critical operations must prevent context switches from occurring during those
critical periods.
From a task's perspective, a context switch can be "out of the
blue", in the sense that the context switch was forced upon it for
reasons external to the task, or it can be intentional due to the programmer's desire to temporarily suspend the task to do other
things.
Most processors support general-purpose stacks and have multiple
registers. Just restoring the appropriate program counter will not be
enough to guarantee the continuity of a task's execution. That's because the stack and the register values will be unique to that task at
the moment of the context switch. A context switch saves the entire task's context (e.g. program counter, registers, stack contents).
Most processor architectures require that memory must be allocated to each task to support context switching.

Tasks and Interrupts
As is the case with foreground / background systems, multitasking
systems often make extensive use of interrupts. Tasks must be protected from the effects of interrupts, ISRs should be as fast as possible, and interrupts should be enabled most of the time. Interrupts
and tasks coexist simultaneously – an interrupt may occur right in
the middle of a task. The disabling of interrupts during a task
should be minimized, yet interrupts will have to be controlled to
avoid conflicts between tasks and interrupts when shared resources
are accessed by both.

Salvo User Manual

Chapter 2 • RTOS Fundamentals

17

3

9

ISR
8
7

high-priority
task

low-priority
task

2

10
11

4

1

5

6

time
Figure 2: Interrupts Can Occur While Tasks Are Running

In Figure 2 a low-priority task is running [1] when an interrupt occurs [2]. In this example, interrupts are always enabled. The interrupt [3] runs to completion [4], whereupon the low-priority task [5]
resumes its execution. A context switch occurs [6] and the highpriority task [7] begins executing. The context switch is handled by
the scheduler (not shown). The high-priority task is also interrupted [8-10] before continuing [11].
Interrupt latency is defined as the maximum amount of time that
interrupts are disabled, plus the time it takes to execute the first
instruction of an ISR. In other words, it's the worst-case delay between when an interrupt occurs and when the corresponding ISR
begins to execute.

Preemptive vs. Cooperative Scheduling
There are two types of schedulers: preemptive and cooperative. A
preemptive scheduler can cause the current task (i.e. the task that's
currently running) to be preempted by another one. Preemption
occurs when a task with higher priority than the current task becomes eligible to run. Because it can occur at any time, preemption
requires the use of interrupts and stack management to guarantee
the correctness of the context switch. By temporarily disabling
preemption, programmers can prevent unwanted disruptions in
their programs during critical sections of code.

18

Chapter 2 • RTOS Fundamentals

Salvo User Manual

3 4
ISR
8

high-priority
task
low-priority
task

2
1

5

12
7
6

9
10 11

scheduler
time
Figure 3: Preemptive Scheduling

Preemptive Scheduling
Figure 3 illustrates the workings of a preemptive scheduler. A lowpriority task [1] is running when an external event occurs [2] that
triggers an interrupt. The task's context and some other information
for the scheduler are first saved [3] in the ISR, and the interrupt is
serviced [4]. In this example the high-priority task is waiting for
this particular event and should run as soon as possible after the
event occurs. When the ISR is finished [5], it proceeds to the
scheduler [6], which starts [7] the high-priority task [8]. When it is
finished, control returns to the scheduler [9, 10], which then restores the low-priority task's context and allows it to resume where
it was interrupted [11, 12].
Preemptive scheduling is very stack-intensive. The scheduler
maintains a separate stack for each task so that when a task resumes execution after a context switch, all the stack values that are
unique to the task are properly in place. These would normally be
return addresses from subroutine calls, and parameters and local
variables (for a language like C). The scheduler may also save a
suspended task's context on the stack, since it may be convenient
to do so.
Preemptive schedulers are generally quite complex because of the
myriad of issues that must be addressed to properly support context
switching at any time. This is especially true with regard to the
handling of interrupts. Also, as can be seen in Figure 3, a certain

Salvo User Manual

Chapter 2 • RTOS Fundamentals

19

time lag exists between when an interrupt happens and when the
corresponding ISR can run. This, plus the interrupt latency, is the
interrupt response time (t4 - t2). The time between the end of the
ISR and the resumption of task execution is the interrupt recovery
time (t7 – t5). The system's event response time is shown as (t7 - t2).

Cooperative Scheduling
A cooperative scheduler is likely to be simpler than its preemptive
counterpart. Since the tasks must all cooperate for context switching to occur, the scheduler is less dependent on interrupts and can
be smaller and potentially faster. Also, the programmer knows exactly when context switches will occur, and can protect critical regions of code simply by keeping a context-switching call out of
that part of the code. With their relative simplicity and control over
context switching, cooperative schedulers have certain advantages.
3
ISR
9

high-priority
task
low-priority
task

2
1

4
5
8

10

6 7

11

scheduler
time
Figure 4: Cooperative Scheduling

Figure 4 illustrates the workings of a cooperative scheduler. As in
the previous example, the high-priority task will run after the interrupt-driven event occurs. The event occurs while the low-priority
task is running [1, 5]. The ISR is serviced [2-4] and the scheduler
is informed of the event, but no context switch occurs until the
low-priority task explicitly allows it [6]. Once the scheduler has a
chance to run [7], it starts and runs the high-priority task to completion [8-10]. The scheduler [11] will then start whichever eligible
task has the highest priority.

20

Chapter 2 • RTOS Fundamentals

Salvo User Manual

In comparison to the preemptive scheduling, cooperative scheduling has the advantage of shorter interrupt response and recovery
times and greater overall simplicity. However, the responsiveness
is worse because a high-priority eligible task cannot run until a
lower-priority one has relinquished control of the processor via an
explicit context switch.

More on Multitasking
You can think of tasks as little programs that run within a bigger
program (your application). In fact, by using a multitasking RTOS
your application can be viewed as a framework to define tasks and
to control how and when they run. When your application is running, it means that a bunch of little programs (the tasks) are all
running in a manner that makes it appear as if they execute simultaneously. Of course only one task can actually run at a particular
instant. In order to take full advantage of the multitasking abilities
of the RTOS, you want to define your tasks such that at any particular time, the processor is making the best use of its processing
power by running whichever task is most important. Once your
task priorities are correctly defined, the scheduler will take care of
the rest.

Task Structure
What does a task in a multitasking application actually look like?
A task is generally an operation that needs to occur over and over
again in your application. The structure is really very simple, and
consists of an optional initialization, and then a main loop that is
repeated unconditionally. When used with a preemptive scheduler,
a task might look like this:
Initialize();
while (1) {
...
}
Listing 4: Task Structure for Preemptive Multitasking

because a preemptive scheduler can interrupt a task at any time.
With a cooperative scheduler a task might look like this:
Initialize();
while (1) {
...
TaskSwitch();
...

Salvo User Manual

Chapter 2 • RTOS Fundamentals

21

}
Listing 5: Task Structure for Cooperative Multitasking

The only difference between the two versions is the need to explicitly call out the context switch in the cooperative version. In cooperative multitasking it's up to each task to declare when it is willing
to potentially relinquish control of the processor to another task.
Such context switches are usually unconditional – a trip through
the scheduler may be required even if the current task is the only
task eligible to run. In preemptive multitasking this would never
occur, as the scheduler would force a context switch only when a
higher-priority task had become eligible to run.

Note Context switches can occur multiple times inside a task,
both in preemptive and cooperative multitasking systems.

Simple Multitasking
The simplest form of multitasking involves "sharing" the processor
equally between two or more tasks. Each task runs, in turn, for
some period of time. The tasks round-robin, or execute one after
the other, indefinitely.
This has limited utility, and suffers from the problems of a superloop architecture. That's because all tasks have equal, unweighted
access to the processor, and their sequence of execution is likely to
be fixed.

Priority-based Multitasking
Adding priorities to the tasks changes the situation dramatically.
That's because by assigning task priorities you can guarantee that
at any instant, your processor is running the most important task in
your system.
Priorities can be static or dynamic. Static priorities are priorities
assigned to tasks at compile time that do not change while the application is running. With dynamic priorities a task can change its
priority during runtime.
Is should be apparent that if the highest-priority task were allowed
to run continuously, then the system would no longer be multitasking. How can multiple tasks with different priorities coexist in a
multitasking system? The answer lies in how tasks actually behave
– they're not always running! Instead, what a certain task is doing

22

Chapter 2 • RTOS Fundamentals

Salvo User Manual

at any particular time depends on its state and on other factors, like
events.

Task States
An RTOS maintains each task in one of a number of task states.
Figure 5 illustrates the different states a task can be in, and the allowed transitions between states. Running is only one of several
exclusive task states. A task can also be eligible to run, it can be
delayed, it can be stopped or even destroyed / uninitialized, and it
can be waiting for an event. These are explained below.

eligible
running
delayed

stopped

destroyed

waiting

Figure 5: Task States

Before a task is created, it is in the uninitialized state. It returns to
that state when and if it is destroyed. There's not much you can do
with a destroyed task, other than create another one in its place, or
recreate the same task again. A task transitions from the destroyed
state to the stopped state when it is created via a call to the RTOS
service that creates a task.
An eligible task is one that is ready to run, but can't because it's not
the task with the highest priority. It will remain in this state until
the scheduler determines that it is the highest-priority eligible task
and makes it run. Stopped, delayed and/or waiting tasks can become eligible via calls to the corresponding RTOS services.
A running task will return to the eligible state after a simple context switch. However, it may transition to a different state if either
the task calls an RTOS service that destroys, stops, delays or waits
the task, or the task is forced into one of these states via a call to an
RTOS service from elsewhere in your application.

Salvo User Manual

Chapter 2 • RTOS Fundamentals

23

A delayed task is one that was previously running but is now suspended and is waiting for a delay timer to expire. Once the timer
has expired, the RTOS timer makes the task eligible again.
A stopped task was previously running, and was then suspended
indefinitely. It will not run again unless it is (re-)started via a call
to the RTOS service that starts a task.
A waiting task is suspended and will remain that way until the
event it is waiting for occurs (See "Event-driven Multitasking" below).
It's typical for a multitasking application to have its various tasks
in many different states at any particular instant. Periodic tasks are
likely to be delayed at any particular instant. Low-priority tasks
may be eligible but unable to run because a higher-priority task is
already running. Some tasks are likely to be waiting for an event.
Tasks may even be destroyed or stopped. It's up to the scheduler to
manage all these tasks and guarantee that each tasks runs when it
should. The scheduler and other parts of the RTOS ensure that
tasks transition from one state to the next properly.

Note The heart of a priority-based multitasking application, the
scheduler, is concerned with only one thing – running the highestpriority task that's eligible to run. Generally speaking, the scheduler interacts only with the running task and tasks that are eligible
to run.
An RTOS is likely to treat all tasks in a particular state in the same
manner, and thereby improve the performance of your application.
For example, it shouldn't expend any processor cycles on tasks that
are stopped or destroyed. After all, they're just "sitting there" and
will remain so indefinitely, or until your program makes them eligible to run.

Delays and the Timer
Most embedded programmers are familiar with the simple delay
loop construct, e.g.:
…
for (i = 0; i < 100; i++ )
asm("nop"); /* do nothing for 100 somethings */
…
Listing 6: Delay Loop

24

Chapter 2 • RTOS Fundamentals

Salvo User Manual

The trouble with doing delays like the one shown in Listing 6 is
that your application can't do any useful background processing
while the loop is running. Sure, interrupts can occur in the foreground, but wouldn't it be nice to be able to do something else during the delay?
Another problem with the code in Listing 6 is that it is compiler-,
processor- and speed-dependent. The compiler may or may not
optimize the assembly instructions that make up this loop, leading
to variations in the actual delay time. Changing the processor may
change the delay time, too. And if you increase the processor
clock, the delay will decrease accordingly. In order to circumvent
these problems delay loops are often coded in assembly language,
which severely limits code portability.
An RTOS provides a mechanism for tracking elapsed time through
a system timer. This timer is often called in your application via a
periodic interrupt. Each time it is called, the timer increments a
counter that holds the number of elapsed system ticks. The current
value of the system ticks is usually readable, and perhaps writeable
too, in order to reset it.
The rate at which the timer is called is chosen to yield enough
resolution to make it useful for time-based services, e.g. to delay a
task or track elapsed time. A fluid level monitor can probably
make do with a system tick rate of 1Hz (i.e. 1s system ticks),
whereas a keypad reader might need a system tick rate of 100Hz
(i.e. 10ms system ticks) in order to specify delays for the key debounce algorithm. An unduly fast system tick rate will result in
substantial overhead and less processing power left over for your
application, and should be avoided.
There must also be enough storage allocated to the system ticks
counter to ensure that it will not overflow during the longest time
period that you expect to use it. For example, a one-byte timer and
a 10ms system tick period will provide a maximum specifiable task
delay of 2.55s. In this example, attempting to calculate an elapsed
time via the timer will result in erroneous results if successive
reads are more than 2.55s apart. Task delays fall under similar restrictions. For example, a system with 10ms system ticks and support for 32-bit delays can delay a task up to a whopping 497 days!
Since the use of delays is common, an RTOS may provide built-in
delay services, optimized to keep overhead to a minimum and to
boost performance. By putting the desired delay inside a task, we
can suspend the task while the delay is counting down, and then

Salvo User Manual

Chapter 2 • RTOS Fundamentals

25

resume the task once the delay has expired. Specifying the delay as
a real amount of time will greatly improve our code's portability,
too. The code for delaying a task via the RTOS looks quite different than that of Listing 6:
…
OS_Delay(100); /* delay for 100 ticks @ 50Hz */
…
Listing 7: Delaying via the RTOS

In Listing 7, the call to the RTOS service OS_Delay() changes the
state of the task from running to delayed. Since the task is no
longer running, nor is it even eligible to run (remember, it's delayed), a context switch also occurs, and the highest-priority eligible task (if there is one) starts running.
In Listing 7 OS_Delay() also specifies a delay of 100 system ticks.
If the system in has a system tick rate of 50Hz, then the task will
be delayed for (100 ticks x 20ms) – two full seconds – before
resuming execution once it becomes the highest-priority eligible
task. Imagine how much processing other eligible tasks can do in
two full seconds!
An RTOS can support multiple, simultaneously delayed tasks. It's
up to the RTOS designer to maximize performance – i.e. minimize
the overhead associated with the execution of the timer – regardless of how many tasks are delayed at any time. This timer overhead cannot be eliminated; it can only be minimized.
The resolution and accuracy of the system timer may be important
to your application. In a simple RTOS, the resolution and the accuracy of the timer both equal the system tick period. For example,
delaying a task by n system ticks will result in a delay ranging
from just over n-1 to just under n system ticks of real time (e.g.
milliseconds). This is due to the asynchronous nature of the system
timer – if you delay a task immediately after the (interrupt) call to
the timer, the first delay tick will last nearly a full system tick. If,
on the other hand, you delay a task immediately prior to a system
tick, the first delay tick will be very short indeed.

Event-driven Multitasking
You may have noticed that a delayed task is actually waiting for
something – it's waiting for its delay timer to expire. The expiration of a delay timer is an example of an event, and events may

26

Chapter 2 • RTOS Fundamentals

Salvo User Manual

cause a task to change state. Therefore events are used to control
task execution. Examples of events include:
•
•
•
•
•
•
•
•
•

• an interrupt,
• an error occurring,
• a timer timing out,
• a periodic interrupt,
• a resource being freed,
• an I/O pin changing state,
• a key on a keypad being pressed,
• an RS-232 character being received or
transmitted and
• information being passed from one part of
your application to another.
Listing 8: Examples of Events

In short, an event can be any action that occurs either internal or
external to your processor. You associate an event with the rest of
your application (primarily tasks, but also ISRs and background
code) through the RTOS event services. The interaction between
events and tasks follows certain simple rules:
•

•

•

Salvo User Manual

• Creating an event makes it available to the rest
of your system. You cannot signal an event, nor
can any task(s) wait on the event, until it has
been created. Events can be created with
different initial values.
• Once an event has been created, it can be
signaled. When an event is signaled, we say that
the event has occurred. Events can be signaled
from within a task or other background code, or
from within an ISR. What happens next is
dependent on whether there are one or more
tasks waiting on the event.
• Once an event has been created, one or more
tasks can wait it. A task can only wait one event
at a time, but any number of tasks can all wait
the same event. If one or more tasks are waiting
an event and the event is signaled, the highestpriority task or the first task to wait the event
will become eligible to run, depending on how
the RTOS implements this feature. If multiple
waiting tasks share the same priority, the RTOS

Chapter 2 • RTOS Fundamentals

27

will have a well-defined scheme12 to control
which task becomes eligible.
One reason for running tasks in direct response to events is to guarantee that at any time the system can respond as quickly as possible to an event. That's because waiting tasks consume no13
processing power – they'll remaining waiting indefinitely, until the
event they're waiting on finally occurs. Furthermore, you can tailor
when the system acts on the event (i.e. run the associated task)
based on its relative importance, i.e. based on the priority of the
task(s) associated with the event.
The key to understanding multitasking's utility is to know how to
structure the tasks in your application. If you're used to superloop
programming, this may be difficult at first. That's because a common mode of thinking goes along the lines of "First I need to do
this, then that, then the other thing, etc. And I must do it over and
over again, checking to see if or when certain events have happened." In other words, the superloop system monitors events in a
sequential manner and acts accordingly.
For event-driven multitasking programming, you may want to
think along these lines: "What events are happening in my system,
both internally and externally, and what actions do I take to deal
with each event?" The difference here is that the system is purely
event-driven. Events can occur repetitively or unpredictably. Tasks
run in response to events, and a task's access to the processor is a
function of its priority.14 A task can react to an event as soon as
there are no higher-priority tasks running.

Note Priorities are associated with tasks, not events.
In order to use events in your multitasking application, you must
first ask yourself:
•
•
•
•
•
12
13
14

28

• what does my system do?
• how do I divide up its actions into separate
tasks?
• what does each task do?
• when is each task done?
• what are the events?

Generally LIFO or FIFO, i.e. the most recent task or the first task,
respectively, to wait the event will become eligible when the event is signaled.
Unless they’re waiting with a timeout, which requires the timer.
Task priorities are easily incorporated into event-based multitasking.

Chapter 2 • RTOS Fundamentals

Salvo User Manual

•

• which event(s) cause each task to run?

Note Events need not be associated with tasks one-to-one. Tasks
can interact with multiple events, and vice versa. Also, tasks that
do not interact with any events are easily incorporated – but they
are usually assigned low priorities, so that they only run when the
system has nothing else to do.

Events and Intertask Communications
An RTOS will support a variety of ways to communicate with
tasks. In event-based multitasking, for a task to react to an event,
the event must trigger some sort of communication with the task.
Tasks may also wish to communicate with each other. Semaphores, messages and message queues are used for intertask communication and are explained below.
Common to all these intertask communications are two actions:
that of signaling (also called posting or sending) and waiting (also
called pending or receiving). Each communication also requires an
initialization (creating).

Note All operations involving semaphores, messages and message queues are handled through calls to the operating system.

Semaphores
There are two types of semaphores: binary semaphores and counting semaphores. A binary semaphore can take on only two values,
0 or 1. A counting semaphore can take on a range of values based
on its size – for example, an 8-bit counting semaphore's value can
range from 0 to 255. Counting semaphores can also be 16-bit or
32-bit. Figure 6 illustrates how we will represent semaphores and
their values:
Sem

Sem

0

b

b

Sem

Sem

0

1

,
, ... ,

n

Figure 6: Binary and Counting Semaphores

Salvo User Manual

Chapter 2 • RTOS Fundamentals

29

Before it is used, a semaphore must be created with an initial
value. The appropriate value will depend on how the semaphore is
used.

Event Flags
Event flags are one such use for binary semaphores – they indicate
the occurrence of an event. If a semaphore is initialized to 0, it
means that the event has not yet occurred. When the event occurs,
the semaphore is set to 1 by signaling the semaphore.
Sem

0

1

b

Sem

1

b

time
Figure 7: Signaling a Binary Semaphore

Figure 7 shows an ISR, task or other background code signaling
[1] a binary semaphore. Once a semaphore (binary or counting)
has reached its maximum value, further signaling is in error.
In addition to signaling a semaphore, a task can also wait the
semaphore. Only tasks can wait semaphores – ISRs and other
background code cannot. Figure 8 illustrates the case of an event
having already occurred when the task waits the semaphore.

Sem

0

b

2

Sem

1

b

4

Sem

0

b

1
running

running

3

5

time
Figure 8: Waiting a Binary Semaphore When the Event
Has Already Occurred

In Figure 8, the binary semaphore is initialized to 0 [1]. Some time
later, the event occurs, signaling the semaphore [2]. When the task
finally runs [3] and waits the semaphore, the semaphore will be
reset [4] so that it can be signaled again and the task will continue
running [5].

30

Chapter 2 • RTOS Fundamentals

Salvo User Manual

Note A semaphores is always initialized without any waiting
tasks.
If the event has not yet occurred when the task waits the semaphore, then the task will be blocked. It will remain so (i.e. in the
waiting state) until the event occurs. This is shown in Figure 9.

Sem

0

4

b

Sem

0

b

1
running

3

waiting

5

eligible

6

eligible

2

time
Figure 9: Signaling a Binary Semaphore When a Task is
Waiting for the Corresponding Event

In Figure 9, an event has not yet been signaled [1] when a running
task [2] waits the binary semaphore. Since the semaphore is not
set, the task is blocked and must wait [3] indefinitely. The operating system knows that this task is blocked because it is waiting for
a particular event. When the semaphore is eventually signaled from
outside the task [4], the operating system makes the task eligible
again [5] and it will run when it becomes the most eligible task [6].
The semaphore remains cleared because a task was waiting for it
when it was signaled. Contrast this to Figure 7, where a semaphore
is signaled with no tasks waiting for it.
It is also possible to combine event flags using the conjunctive
(logical AND) or disjunctive (logical OR) combination of the
event flags. The event is signaled when all (AND) or at least one
(OR) of the event flags are set.

Note One or more tasks can concurrently wait an event. Which
task becomes eligible depends on the operating system. For example, some operating systems may make the first task to wait the
event eligible (FIFO), and others may make the highest-priority
task eligible. Some operating systems are configurable to choose
one scheme over the other.

Task Synchronization
Since tasks can be made to wait on an event before continuing, binary semaphores can be used as a means of synchronizing program

Salvo User Manual

Chapter 2 • RTOS Fundamentals

31

execution. Multitask synchronization is also possible – Figure 10
shows two tasks synchronizing their execution via two separate
binary semaphores.

6

waiting semaphore #1

signal semaphore #2

running

7

1
running

2

signal semaphore #1

4

waiting semaphore #2

5

3

time
Figure 10: Synchronizing Two Tasks with Event Flags

In Figure 10, binary semaphores #1 and #2 are initialized to 0 and
1, respectively. The upper task begins by waiting semaphore #1,
and is blocked [1]. The lower task begins running [2], and when it
is ready to wait for the upper task it signals semaphore #1 [3] and
then waits semaphore #2 [4], and is blocked [5] since it was initialized to 0. The upper task then begins running [6] since semaphore
#1 was signaled, and when it is ready to wait for the lower task it
signals semaphore #2 [7] and then waits semaphore #1, and is
blocked [1]. This continues indefinitely. Listing 9 shows the pseudocode for this example.
initialize binary semaphore #1 to 0;
initialize binary semaphore #2 to 1;
UpperTask()
{
while (1) {
/* wait for LowerTask() */
wait binary semaphore #1;
do stuff;
signal binary semaphore #2;
}
}
LowerTask()
{
while (1) {
do stuff;
signal binary semaphore #1;
/* wait for UpperTask() */
wait binary semaphore #2;
}
}
Listing 9: Task Synchronization with Binary
Semaphores

32

Chapter 2 • RTOS Fundamentals

Salvo User Manual

Resources
Semaphores can also be used to manage resources via mutual exclusion. The resource is available if the binary semaphore is 1, and
is not available if it is 0. A task that wishes to use the resources
must acquire it by successfully waiting the binary semaphore.
Once it has acquired the resource, the binary semaphore is 0, and
therefore any other tasks wishing to use the resource must wait until it is released (by signaling the binary semaphore) by the task
that has acquired the resource.
initialize binary semaphore to 1;
TaskUpdateTimeDate()
{
while (1) {
…
prepare time & date string;
wait binary semaphore;
write time & date string to display;
signal binary semaphore;
…
}
}
TaskShowAlert()
{
while (1) {
wait binary semaphore;
write alert string to display;
signal binary semaphore;
}
}
Listing 10: Using a Binary Semaphore to Control Access
to a Resource

In Listing 10 a binary semaphore is used to control access to a
shared resource, a display (e.g. an LCD). In order to enable access
to it, the semaphore must be initialized to 1. A task wishing to
write to the display must acquire the resource by waiting the semaphore. If the resource is not available, the task will be blocked until
the resource is released. After acquiring the resource and writing to
the display, the task must then release the semaphore by signaling
it.
Resources can also be controlled with counting semaphores. In this
case, the value of the counting semaphore represents how many of
the resources are available for use. A common example is that of a
ring buffer. A ring buffer has space for m elements, and elements
are added to and removed from it by different parts of an application. Figure 11 shows a scheme to transmit character strings via

Salvo User Manual

Chapter 2 • RTOS Fundamentals

33

RS-232 using a counting semaphore to control access to a ring
buffer.

from
application

Tx ISR
signals
semaphore

Tx ring buffer
of size m
2

task
waits
semaphore

t
head
tail

n
\0 g

3

i

e
s
t

4

to RS-232
transmitter
Sem

m

1

Figure 11: Using a Counting Semaphore to Implement a
Ring Buffer

In Figure 11 a counting semaphore is initialized to m [1] to represent the number of spaces available in the empty ring buffer [2].
The ring buffer is filled at its tail15 by the task [3] and emptied
from its head by the ISR [4]. Before adding a character to the
buffer the task must wait the semaphore. If it is blocked, it means
that the buffer is full and cannot accept any more characters. If the
buffer is not full, the semaphore is decremented, the task places the
character at the tail of the buffer and increments the tail pointer.
Once there are characters in the buffer16, for each character the Tx
ISR will remove it from the buffer, transmit it and increment the
semaphore by signaling it. The corresponding pseudocode is
shown17 in Listing 11.
initialize counting semaphore to m;
TaskFillTxBuffer()
{
while (1) {
wait semaphore;
place char at TxBuff[tail pointer];
increment tail pointer;
}
}
ISRTxChar()
{
send char at TxBuff[head pointer] out RS-232;
increment head pointer;
signal semaphore;

15

16
17

34

The tail pointer points to the next available free space for insertion into the
ring buffer. The head pointer points to the first available element for removal
from the ring buffer.
This is usually signified by enabling transmit interrupts.
The control of Tx interrupts, which varies based on transmitter configurations,
is not shown.

Chapter 2 • RTOS Fundamentals

Salvo User Manual

}
Listing 11: Using a Counting Semaphore to Control
Access to a Resource

By using a task to fill the ring buffer, the application need not poll
the buffer's status at regular intervals to determine when to insert
new characters. Nor does the application need to wait in a loop for
room to insert characters into the buffer. If only part of a string is
inserted before the task is blocked (i.e. the string is larger than the
available room in the buffer), the task will automatically resume
inserting additional characters each time the ISR signals the counting semaphore. If the application sends strings infrequently, a low
task priority will probably suffice. Otherwise a high task priority
may be necessary.

Note The RAM required for the semaphore that is used to manage a resource is separate from the RAM allocated to the resource
itself. The RTOS allocates memory for the semaphore – the user
must allocate memory for the resource. In this example, 8-bit
counting semaphores limit the size of the ring buffer to 256 characters. The semaphore will require one byte of RAM irrespective of
the actual (user-declared) size of the ring buffer itself.

Messages
Messages provide a means of sending arbitrary information to a
task. The information might be a number, a string, an array, a function, a pointer or anything else. Every message in a system can be
different, as long as both the sender and the recipient of the particular message understand its contents. Even the type of message
can even change from one message to the next, as long as the
sender and recipient are aware of this! As with semaphores, the
operating system provides the means to create, signal and wait
messages.
In order to provide general-purpose message contents, when a
message is sent and received, the actual content of the message is
not the information itself, but rather a pointer to the information. A
pointer is another word for the address (or location) of the information, i.e. it tells where to find the information. The message's
recipient then uses the pointer to obtain the information contained
in the message. This is called dereferencing the pointer.18
18

Salvo User Manual

In C, & is the address of operator, and * is the unary operator for indirection.
Therefore if var is a variable and p points to it, then p=&var and *p is equal
to var.

Chapter 2 • RTOS Fundamentals

35

If a message is initialized to be empty, it contains a null pointer. A
null pointer is a pointer with a value of 0. By convention, a null
pointer doesn't point to anything; therefore it carries no other information with it. A null pointer cannot be dereferenced.
Signaling (i.e. sending) a message is more complex than signaling
a semaphore. That's because the operating system's messagesignaling function requires a message pointer as an argument. The
pointer passed to the function must correctly point to the information you wish to send in the message. This pointer is normally nonzero, and is illustrated in Figure 12.
39Ah
39Bh
39Ch
39Dh
39Eh
39Fh
3A0h
3A1h

t
e
s
t
i
n
g
\0

Msg

39Ah
1

2

Figure 12: Signaling a Message with a Pointer to the
Message's Contents

In Figure 12, a C-language character string19 [1] is sent in a message [2] by signaling the message with a pointer. The string resides
at a particular physical address. The message does not contain the
first character of the string – it contains the address of the first
character of the string (i.e. a pointer to the string), and the pointer's
value is 39Ah. The pseudocode for sending this message is shown
in Listing 12.
string[] = "testing";
p = address(string);
signal message with p;
Listing 12: Signaling a Message with a Pointer

To receive a message's contents, a task must wait the message. The
task will be blocked until the message arrives. The task then extracts the contents of the message (i.e. the pointer) and uses the
pointer in whatever manner it chooses. In Listing 13, the receiving
task capitalizes the string that the message points to.
TaskCaps()
{
while (1) {
wait message containing string pointer p;

19

36

In C, character strings end with the NUL character ('\0').

Chapter 2 • RTOS Fundamentals

Salvo User Manual

while ((p) is not null) { 20
if ('a' <= (p) <= 'z')
(p) = (p) - 32;
increment p;
}
}
}
Listing 13: Receiving a Message and Operating on its
Contents

A message can contain at most one item of information (i.e. a
pointer) at a time. If the message is empty, it can be signaled. If it's
full, the message cannot be sent.
Messages can be used like binary semaphores. A message containing a null pointer is equivalent to a binary semaphore of value 0,
and a message containing a non-zero pointer is equivalent to a binary semaphore of value 1. This is useful if binary semaphores are
not explicitly supported by the RTOS.

Message Queues
Message queues are an extension of messages. A message queue
can contain multiple messages (up to a predetermined number) at
any time. Sending messages can continue until the message mailbox is full. A task that waits the message queue will receive messages until the message queue is empty.
An RTOS will need to allocate some additional RAM to manage
each message queue. This RAM will be used to keep track of the
number of messages in the message queue, and the order in which
the messages exist in the message queue.

Summary of Task and Event Interaction
Here is a summary of the rules governing the interaction of tasks
and events (i.e. semaphores, messages and message queues).
•
•
•
•
20

Salvo User Manual

• An events must be initialized. It is initialized
without any waiting tasks.
• A task cannot wait an event until the event has
been initialized.
• Multiple tasks can wait a single event.
• A task can only wait one event at a time.

"(pointer)" is pseudocode for "what is pointed to by the pointer."

Chapter 2 • RTOS Fundamentals

37

•
•
•
•
•
•
•
•

• A semaphore's value can range from 0 to its
maximum value, depending on its size.
• A message contains a pointer to some
information.
• Message queues can hold multiple messages at
once.
• An ISR, a task or other background code can
signal an event.
• Only a task can wait an event.
• A task will be blocked (i.e. it will change to
the waiting state) if the event it waits is not
available.
• Which waiting task becomes eligible when an
event is signaled is dependent on how the
operating system implements event services.
• If an event has already been signaled, no task
is waiting it, and it is signaled again, then either
an error has occurred or the signaling task can
be blocked. This is dependent on how the
operating system implements event services.

Conflicts
A variety of conflicts may occur within a multitasking environment. They are described below.

Deadlock
Deadlock occurs with two or more tasks when each task is waiting
for a resource controlled by another task. Since all of the affected
tasks are waiting, there is no opportunity for any of the resources
to become available. Therefore all the tasks will be deadlocked, i.e.
they will wait indefinitely.
The solution is for all tasks wishing to acquire the resources to
•
•
•

• always acquire the resources in a
predetermined order,
• acquire all the resources before continuing,
and
• release the resources in the opposite order.

Alternatively, by using a timeout one can break a deadlock. When
attempting to acquire the resource, an optional time period can be

38

Chapter 2 • RTOS Fundamentals

Salvo User Manual

specified. If the resource is not acquired within that time period,
the task continues, but with an error code that indicates that it
timed out waiting for the resource. Special error handling may then
be invoked.

Priority Inversions
Priority inversions occur when a high-priority task is waiting for a
resource controlled by a low-priority task. The high-priority task
must wait until the low-priority task releases the resource, whereupon it can continue. As a result, the priority of the high-priority
task is effectively reduced to that of the low-priority task.
There are a variety of ways to avoid this problem (e.g. priority inheritance), most of which involve dynamically changing the priority of a task that controls a resource based on the priority of tasks
wishing to acquire the resource.

RTOS Performance
The code to implement a multitasking RTOS may be larger than
what's required in a superloop implementation. That's because each
task requires a few extra instructions to be compatible with the
scheduler. Even so, a multitasking application is likely to have
much better performance and be more responsive than one with a
superloop. That's because a well-written RTOS can take advantage
of the fact that tasks that are not running often need not consume
any processing power at all. This means that instead of spending
instruction cycles testing flags, checking counters and polling for
events, your multitasking application makes the most of the processor's power by using it directly where you need it most – on the
highest-priority task that's eligible to run.

A Real-World Example
Let's look at an interesting example application – the controller for
a remote soda-can vending machine. It must indicate (via LEDs on
the buttons) if any selections are empty, handle the insertion of
coins and bills, properly interpret the customer's selection, release
the right item to the customer, and make change properly. A modern, microprocessor-controlled vending machine might also regulate internal temperatures (e.g. for soda cans), be connected to a
network to relay out-of-stock information to a remote location, and
be tied into a security system to deter vandalism. And of course all

Salvo User Manual

Chapter 2 • RTOS Fundamentals

39

of this has to be done without error regardless of how many unpredictable things the customer does in the quest to quench his or her
hunger or thirst.

The Conventional Superloop Approach
The refrigerated, vandal-resistant vending machine in our example
has a user interface consisting of an array of item-selection buttons
and slots for bills and coins. The main loop for a pseudo-code version of a traditional superloop implementation might look like this:
Initialize();
do forever
{
ControlTemps();
ShowEmpties();
AcceptCurrency();
flagSelectionGood = FALSE;
ReadButtons();
if (flagSelectionGood) {
ReleaseItem();
MakeChange();
}
if (Tilt()) {
CallPolice();
}
}
Listing 14: Vending Machine Superloop

where some ISRs (not shown) are employed to do things like debounce the button presses. Listing 14 also shows neither the individual functions (e.g. ReleaseItem()) nor the global variables
required to pass information between the functions, e.g. between
ReadButtons() and ReleaseItem().
Let's examine Listing 14 in more detail. In the superloop we call
ControlTemps() once each time through the loop. On an 8-bit,
8MHz processor likely to be used in such an application, we might
expect ControlTemps() to be called once every 200 microseconds
when there's no user activity. This is a huge waste of processing
power, as we know that we really only need to call it once a minute. We're calling ControlTemps() 5,000 times more often than
necessary! While this may be acceptable in a vending machine, it's
unlikely to be in a more demanding application.

40

Chapter 2 • RTOS Fundamentals

Salvo User Manual

One approach to fixing this would be to dedicate a periodic interrupt to set a globally visible bit every second. Then we could check
this bit and call ControlTemps() when the bit is set high. This approach isn't too clever, because we're still doing an operation (testing the bit) every 200 microseconds. Another approach would be
to move ControlTemps() completely into an ISR that's called
every second, but that's ill-advised, especially if ControlTemps()
is a large and complex function.
In our example, ReleaseItem() will run only when money's in the
machine and a button has been pressed. In other words, it's waiting
for an event – an event characterized by the presence of the proper
amount of money AND a valid selection button being pressed.
As illustrated in Listing 14, foreground / background superloop
software designs puts most of the required processing in a single
main loop that the processor executes over and over again. External events and time-critical processing are handled in the foreground via ISRs. Note that no single operation in the superloop has
priority over any other. The execution of the functions proceeds in
a rigidly serial manner, with the use of many hierarchical loops.
When adding more functionality to a system like this, the main
loop is likely to grow larger and slower, perhaps more ISRs will be
needed, and system complexity will increase in your attempt to
keep everything working as a whole.
For instance, in the above example there's no way for the customer
to cancel a purchase. How would you modify the code to handle
this additional requirement? You could write an expanded state
machine to handle various scenarios, or use lots of timer interrupts
to control how often various functions can run. But do you think
someone else would understand what you wrote? Or even you, two
years from now?

The Event-Driven RTOS Approach
If we start to talk about understanding, modifying and maintaining
foreground / background code of moderate to severe complexity, it
loses its appeal. That's because there are no clear relationships
among the various functions in the superloop, nor between the
functions and the flag variables, nor between the ISRs and the super loop. Let's try a different, task- and event-based approach.
Here's a list of tasks we can identify from the example above:

Salvo User Manual

Chapter 2 • RTOS Fundamentals

41

•

• Monitor and control internal temperature –
ControlTemps()

•

• Display empty bins via LEDs –
ShowEmpties()

•

• Accept or reject currency, and total it –
AcceptCurrency()

•
•
•

• Debounce and read buttons – ReadButtons()
• Make change – MakeChange()
• Release selected item to customer –
ReleaseItem()

•

• Attempt to protect the vending machine from
vandalism – CallPolice()

Let's examine each of these tasks in a little more detail. We'll look
at how important each one is, from 1 (most important) to 10 (least
important), and when each task should run.
is obviously important, as we want to keep the
sodas cool. But it probably doesn't have to run more often than,
say, once a minute, to accurately monitor and be able to control the
temperature. We'll give it a priority of 4.
ControlTemps()

isn't too important. Moreover, the status of the
empty bins only changes each time an item is released to the customer. So we'll give it a priority of 8, and we'd like it to run initially and once for every time an item is released.
ShowEmpties()

should have a reasonably high priority so that
there's no noticeable lag when the customer presses the machine's
buttons. Since button presses are completely asynchronous, we
want to test the array of buttons regularly for activity. Let's give it
a priority of 3, and run it every 40 milliseconds.
ReadButtons()

Since AcceptCurrency() is also part of the user interface, we'll
give it the same priority as ReadButtons() and we'll run it every
20 milliseconds.
The machine's manufacturer does not consider MakeChange() to
be all that important, so we'll give it a priority of 10. We'll link it to
ReleaseItem(), since change must be made only after the selected
item is delivered to the customer.
is interesting because we only need it once the
proper amount of money has been accepted and an item button is
pressed. To respond quickly we'll give it a priority of 2, and we'd
ReleaseItem()

42

Chapter 2 • RTOS Fundamentals

Salvo User Manual

like it to run when the above combination of money and button
press occurs.
The machine's manufacturer makes a big point of how vandalresistant it is. It's even capable of detecting an attack (through
built-in tilt sensors) and calling the local security service. We'll
give CallPolice() the highest priority of 1, and we'll check the
tilt sensors every 2 seconds for an attack.

Step By Step
Our vending machine example requires seven tasks with six different priorities, and a timer resolution of 20ms. To create this multitasking application from these functions, we'll need to:
•
•
•
•
•
•
•

• initialize the operating system,
• modify the structure of the tasks so as to be
compatible with the operating system and the
events,
• create prioritized tasks from the task functions,
• link the real-world events to events that the
operating system understands,
• create a system timer to keep track of elapsed
time,
• start the various tasks and
• begin multitasking.

Initializing the Operating System
Initializing the operating system is usually straightforward, e.g.
InitializeMultitasking();

This creates the necessary (empty) structures the operating system
will use to manage task execution and events. At this point all of
the system's tasks are in the uninitialized / destroyed state.

Structuring the Tasks
The tasks written for a multitasking application look similar to
those written for a superloop application. The big difference lies in
the overall program structure. The multitasking tasks are not contained in any loops or larger functions – they're all independent

Salvo User Manual

Chapter 2 • RTOS Fundamentals

43

functions. ReleaseItem(), which releases an item once a set of
conditions has been met, might look like this in pseudocode:
ReleaseItem()
{
do forever {
WaitForMessage(messageSelection, item);
Release(item);
}
}
Listing 15: Task Version of ReleaseItem()

In Listing 15 ReleaseItem() waits forever for a (particular) message and does nothing until the message arrives. While it's waiting
for the message to arrive, ReleaseItem() is in the waiting state.
When the message is sent, ReleaseItem() becomes eligible to
run, and when it runs, it extracts the contents of the message (in
this case, a code for the desired item, e.g. "B3") and releases it to
the customer. ReleaseItem() is not inside any larger loop, nor is
it called by any other functions (except indirectly by the scheduler,
below).
CallPolice()

has a similar "stand-alone" look to it:

CallPolice()
{
do forever {
Delay(1000);
if (Tilt()) {
SendMsgToPoliceHQ();
}
}
}
Listing 16: Task Version of CallPolice()

enters an infinite loop where it delays itself for
1000 x 20ms, or 2 seconds, and then sends a message to the police
headquarters if the vending machine's tilt sensors detect an attack.
It repeats this sequence indefinitely. While delayed, CallPolice()
is in the delayed state.
CallPolice()

Prioritizing the Tasks
An operating system call assigns a priority to a task, and prepares
the task for multitasking. For example,

44

Chapter 2 • RTOS Fundamentals

Salvo User Manual

CreateTask(ShowEmpties(), 8)
Listing 17: Prioritizing a Task

tells the operating system that it should give ShowEmpties()a priority of 8 and add it to the tasks whose execution it will manage.
ShowEmpties() is now in the stopped state.

Interfacing with Events
In Listing 15, ReleaseItem() is using a message to handle an
event – namely the release of an item. That message needs to be
initialized:
CreateEvent(messageSelection, empty);
Listing 18: Creating a Message Event

By initializing messageSelection to empty (i.e. no valid selection
has been made), ReleaseItem() will only release an item once the
required events (enough money inserted and appropriate button
pressed) have occurred.

Adding the System Timer
An RTOS needs some way to keep track of real time – this is usually provided via some sort of timer function that the application
must call at a regular, predefined rate. In this case that rate is 50Hz
or every 20ms. Calling the system timer is often accomplished
through an interrupt, e.g.:
InterruptEvery20ms()
{
SystemTimer();
}
Listing 19: Calling the System Timer

Starting the Tasks
Applications must create all of their tasks and events before any of
them are actually used. By providing an explicit means of starting
tasks, the RTOS enables you to manage system startup in a predictable way:
StartTask(ControlTemps());
StartTask(ShowEmpties());
StartTask(AcceptCurrency());

Salvo User Manual

Chapter 2 • RTOS Fundamentals

45

StartTask(ReadButtons());
StartTask(MakeChange());
StartTask(ReleaseItem());
StartTask(CallPolice());
Listing 20: Starting all Tasks

Since multitasking has not yet started, the order in which tasks are
started is immaterial and is not in any way dependent on their
priorities. At this point all of the tasks are in the eligible state.

Enabling Multitasking
Once everything is in place, events have been initialized and the
tasks have been started (i.e. they are all ready to execute), multitasking can begin:
StartMultitasking();
Listing 21: Multitasking Begins

The scheduler will take the eligible task with the highest priority
and run it – i.e. that task will be in the running state. From now on,
the scheduler will ensure that the highest-priority task is the only
one running at any time.

Putting It All Together
Listing 22 is a complete listing of the task- and event-driven vending machine application in pseudocode:
#include "operatingsystem.h"
extern
extern
extern
extern
extern
extern
extern
extern
extern
extern
extern

AlertPoliceHQ()
ButtonPressed()
DisplayItemCounts()
InterpretSelection()
NewCoinsOrBills()
PriceOf()
ReadDesiredTemp()
Refund()
ReleaseToCustomer()
SetActualTemp()
Tilt()

ControlTemps()
{
do forever {
Delay(500);
ReadActualTemp();
SetDesiredTemp();

46

Chapter 2 • RTOS Fundamentals

Salvo User Manual

}
}

ShowEmpties()
{
DisplayItemCounts();
do forever {
WaitForSemaphore(semaphoreItemReleased);
DisplayItemCounts();
}
}

AcceptCurrency()
{
do forever {
Delay(1);
money += NewCoinsOrBills();
}
}

ReadButtons()
{
do forever {
Delay(2);
button = ButtonPressed();
if (button) {
item = InterpretSelection(button);
SignalMessage(messageSelection, item);
}
}
}

MakeChange()
{
do forever {
WaitForMessage(messageCentsLeftOver, change);
Refund(change);
}
}

ReleaseItem()
{
CreateEvent(semaphoreItemReleased, 0);
CreateEvent(messageCentsLeftOver, empty);
do forever {
WaitForMessage(messageSelection, item);

Salvo User Manual

Chapter 2 • RTOS Fundamentals

47

if (money >= PriceOf(item)) {
ReleaseToCustomer(item);
SignalSemaphore(semaphoreItemReleased);
SignalMessage(messageCentsLeftOver,
money - PriceOf(item));
money = 0;
}
}
}

CallPolice()
{
do forever {
Delay(1000);
if (Tilt()) {
AlertPoliceHQ();
}
}
}

InterruptEvery20ms()
{
SystemTimer();
}

main()
{
money = 0;
InitializeMultitasking();
CreateTask(ControlTemps(),
4)
CreateTask(ShowEmpties(),
8)
CreateTask(AcceptCurrency(), 3)
CreateTask(ReadButtons(),
3)
CreateTask(MakeChange(),
10)
CreateTask(ReleaseItem(),
2)
CreateTask(CallPolice(),
1)
CreateEvent(messageSelection, empty);
StartTask(ControlTemps());
StartTask(ShowEmpties());
StartTask(AcceptCurrency());
StartTask(ReadButtons());
StartTask(MakeChange());
StartTask(ReleaseItem());

48

Chapter 2 • RTOS Fundamentals

Salvo User Manual

StartTask(CallPolice());
StartMultitasking();
}
Listing 22: RTOS-based Vending Machine

The RTOS Difference
The program in Listing 22 has an entirely different structure than
the superloop one in Listing 14. Several differences are immediately apparent:
•
•
•

•
•
•
•

• It's somewhat longer – this is mainly due to the
overhead of making calls to the operating
system.
• There are clearly-defined runtime priorities
associated with each task.
• The tasks themselves have simple structures
and are easy to understand. Those that
communicate with other tasks or ISRs use
obvious mechanisms (e.g. semaphores and
messages) to do so. Initialization can be taskspecific.
• The use of global variables is minimized.
• There are no delay loops.
• It's very easy to modify, add or delete a task
without affecting the others.
• The overall behavior of the application is
largely dependent on the task priorities and
intertask communication.

Perhaps most importantly, the RTOS handles the complexity of the
application automatically – tasks run on a priority basis, task
switching and state changes are handled automatically, delays require a minimum of processor resources, and the mechanisms of
intertask communications are hidden from view.
There are other differences that become more apparent during runtime. If we were to look at a timeline showing task activity, we
would see
•

Salvo User Manual

• Every 2 seconds CallPolice() wakes up to
check for tampering and then returns to the
delayed state,

Chapter 2 • RTOS Fundamentals

49

•
•
•
•

• Every second ControlTemps() wakes up to
adjust the internal temperature and then returns
to the delayed state,
• Every 40ms ReadButtons() wakes up to
debounce any button presses and then returns to
the delayed state,
• Every 20ms AcceptCurrency() wakes up to
monitor the insertion of coins and bills and then
returns to the delayed state, and
• ShowEmpties(), MakeChange() and
ReleaseItem() do nothing until a valid
selection has been made, whereupon they
briefly "come to life," deliver the selected item,
refund any change and show full/empty item
statuses, respectively, before returning to the
waiting state.

In other words, for the vast majority of the time it's running, the
vending machine's microcontroller has very little to do because the
scheduler sees only delayed and waiting tasks. If the vending machine's manufacturer wanted to promote "Internet connectivity for
enhanced stock management, remote querying and higher profits"
as an additional feature, adding an extra task to transmit sales data
(e.g. which sodas are purchased at what time and date and at what
outside temperature) and run a simple web server would be as easy
as creating another task to run in addition to the ones above and
assigning it an appropriate priority.

50

Chapter 2 • RTOS Fundamentals

Salvo User Manual

Chapter 3 • Installation

Introduction
Salvo is provided in a self-extracting executable. Each installer
will install all the files needed to build Salvo applications for the
intended target and compiler, as well as additional files like Salvo
Compiler Reference Manuals. All of the Salvo files are contained
in compressed and encrypted form within the installer.

Note This section assumes you are installing Salvo onto a PC or
PC compatible running Microsoft Windows XP. The installation
for other Windows releases is similar.

Running the Installer
1.

Launch

the

distribution-specific

lite|tiny|SE|LE|Pro-target-version.exe

salvoinstaller
on your PC. The

Welcome screen appears:

Figure 13: Welcome Screen

Salvo User Manual

51

Note Most of the installer's screens contain Next, Back and
Cancel buttons. Click on the Back button for the previous screen.
Click on the Cancel button to abort the installation.
2. After you click on the Next button, the Salvo License Agreement screen appears:

Figure 14: Salvo License Agreement Screen

This screen contains the Pumpkin Salvo License Agreement. Read
this agreement carefully. This document is included in the Salvo
folder once the installation is complete. You must accept the terms
of the License in order to continue installing Salvo. To accept the
License, click on the I Agree button. If you do not accept the License, click on the Cancel button and return the software.21
3. The Choose Components screen appears:

21

52

Instructions on returning the software are contained in the License and in the
User’s Manual.

Chapter 3 • Installation

Salvo User Manual

Figure 15: Choose Components Screen

Normally you should leave these selections at their default values.
The components typically include the Salvo core (i.e. all the compiler- and target-independent components of Salvo), as well as
compiler- and target-specific files.

Tip A description of the contents and function of each individually selectable component is available by expanding the tree and
positioning the mouse over the component of interest.
This screen provides an elegant way of restoring individual Salvo
components without doing a complete re-install. For example, if
you accidentally delete a Salvo library for a particular compiler,
you can choose to re-install just the Salvo libraries for said compiler via this screen.

Tip Installing components for toolsets you do not have installed
does not normally cause problems. Therefore it is recommended
that you leave all components selected (default).
4. The Choose Install Location screen appears:

Salvo User Manual

Chapter 3 • Installation

53

Figure 16: Choose Destination Location Screen

This screen allows you to set the directory where Salvo will be installed. The installer will place several22 directories, some with
nested subdirectories, in the destination directory. You can leave
the destination directory at its default (C:\Pumpkin\Salvo) or you
can change it by clicking on the Browse… button and selecting a
different destination directory.

Tip In order to avoid potential compiler problems with long pathnames and spaces in path names, choosing the default path name is
recommended. Choosing a deeply nested directory (e.g.
C:\Program
Files\Pumpkin\My
jects\Programming\Tools \RTOS\Salvo\v4.1.0)

Pro-

may cause
problems with some tools due to exceedingly long pathnames for
Salvo files. Also, spaces (' ') in pathnames should be avoided as
some legacy compilers do not support them.
5. After clicking on the Next button the Choose Start Menu
Folder screen appears:

22

54

See Figure 20: Typical Salvo Install Directory Contents.

Chapter 3 • Installation

Salvo User Manual

Figure 17: Choose Start Menu Folder Screen

6. Click on Install to continue. The installation begins, and ends
with the Installation Complete screen:

Figure 18: Installation Complete Screen

Salvo User Manual

Chapter 3 • Installation

55

This screen lists all the files in the Salvo distribution installed to
your PC. Individual files are marked as Extract (file was installed)
or Skipped (file was skipped because a newer destination file with
the same name was detected).

Tip The output of this screen's window is scrollable via the elevator on the right.
7. Once the installation of the files is completed, click on the Next
button. You will be greeted with the Finish screen:

Figure 19: Finish Screen

Network Installation
If you are working in a networked environment with code sharing
(e.g. for revision control) and need to install Salvo on a shared
network disk, run the installer on a Wintel PC and choose a directory on a network drive as the destination directory. You may find
it convenient to create the shortcuts in the Salvo Start Menu programs folder on each machine that is accessing Salvo over the
network.

Note Network installations must comply with the terms of the
Salvo License Agreement. See the License for more information.

56

Chapter 3 • Installation

Salvo User Manual

Installing Salvo on non-Wintel Platforms
If you are developing Salvo applications on a non-Wintel platform,
you will still need access to a Wintel machine in order to run the
installer. The installer will place all of Salvo's files into the selected destination directory (the default is C:\Pumpkin\Salvo),
with multiple subdirectories. You can then copy the entire subdirectory to another machine via a network or a mass storage device
(e.g. Zip, Jaz, tape, etc.).

Note The Salvo License Agreement allows only one copy of the
Salvo directories per installation. You must remove the entire
Salvo directory from the Wintel machine after you have transported it to your non-Wintel development environment. See the
License for more information.
Alternatively, if you are working in a networked environment with
cross-platform file sharing, you can run the installer on a Wintel
PC and select a (remote) directory on your non-Wintel platform as
the destination directory for the installation. All of the Salvo files
will be installed to the remote directory. After the installation is
complete you may want to remove the Start Menu items from the
Wintel PC if you will not be using them.

A Completed Installation
Your Salvo directory should look similar to this after a typical installation:

Figure 20: Typical Salvo Install Directory Contents
(Lib Subdirectory View)

Salvo User Manual

Chapter 3 • Installation

57

Uninstalling Salvo
The setup program automatically provides an uninstaller for each
Salvo distribution. To use the uninstaller, run the appropriate Remove Salvo item as shown below:

Figure 21: Location of the Uninstaller(s)

You will be greeted with a confirmation screen:

Figure 22: Confirming the Uninstall Operation

Click on the Yes button to begin uninstalling the specified Salvo
distribution:

58

Chapter 3 • Installation

Salvo User Manual

Figure 23: Uninstallation Complete Screen

Finally, the uninstaller will display the following screen upon successfully removing the specified Salvo distribution from your PC:

Figure 24: Uninstall Complete Screen

Click on the OK button to finish uninstalling Salvo.

Note The uninstaller will not remove any non-Salvo files in
Salvo directories, nor will it delete any non-empty directories. If,
after a Salvo uninstallation, files and/or directories still exist in the
Salvo tree, you are advised to inspect those directories and delete
the files and/or directories as required.

Uninstalling Salvo on non-Wintel Machines
If you are using Salvo on another platform (e.g. Linux), simply delete the Salvo destination directory and all of its subdirectories.

Salvo User Manual

Chapter 3 • Installation

59

Installations with Multiple Salvo Distributions
The Salvo installer is designed to support multiple Salvo distributions of different types all in one directory (usually
C:\Pumpkin\Salvo).23 For example, you could have Salvo Lite for
TI's MSP430 as well as Salvo Pro for 8051 family installed together in C:\Pumpkin\Salvo.

Installer Behavior
The Salvo installers replace files shared across all of the distributions only when the files to be installed are newer than the existing
ones. When installed, a shared file is generally made read-only.
Shared files include the target-independent Salvo header file and
source files. Files that are unique to a distribution (e.g. project files
and Salvo libraries) are always installed, i.e. overwritten by the
installer.

Installing Multiple Salvo Distributions
Normally, no extra precautions are required when installing additional Salvo distributions onto a PC containing one or more existing Salvo distributions. By virtue of the installer's behavior, only
the latest shared files should remain on the PC after each installer
has finished.

Uninstalling with Multiple Salvo Distributions
Since an uninstaller will remove shared files, it is necessary to
uninstall all of the Salvo distributions on the PC, and then re-install
the desired ones.

Copying Salvo Files
Salvo users are strongly discouraged from copying any of Salvo's
shared files to locations outside of the files' normal installation directories. Having duplicate Salvo files can lead to unpredictable
behavior, and can greatly complicate debugging.

23

60

As of Salvo v3.2.2.

Chapter 3 • Installation

Salvo User Manual

Users with revision control systems who wish to add Salvo to their
file repositories can do so by adding them in-place, and by retrieving them from a single source (e.g. a file server).

Modifying Salvo Files
Modifying Salvo's shared files can also lead to unpredictable behavior, and is therefore strongly discouraged. Generally speaking,
only Salvo Pro users should modify Salvo's shared files, and only
when a problem with the file(s) has been officially announced, and
a solution provided. Once an updated Salvo distribution is available, it should automatically replace the modified file with an updated one.

Salvo User Manual

Chapter 3 • Installation

61

62

Chapter 3 • Installation

Salvo User Manual

Chapter 4 • Tutorial

Introduction
In this chapter we'll use a two-part, step-by-step tutorial to help
you create a Salvo application from scratch. The first part is an introduction to using Salvo to write a multitasking program in C. In
the second part we'll compile it to a working application.

Part 1: Writing a Salvo Application
Let's create a multitasking Salvo application step-by-step, introducing various concepts and Salvo features as we go. We'll start
with a minimal application in C and build on it. We'll explain the
purpose and use of each new Salvo feature, and describe in-depth
what's happening in the application.

Tip Each one of the C listings below is provided as a complete
application in the Pumpkin\Salvo\Example\…\Tut\Tut5 directory of each Salvo distribution, complete with projects, source
code and executables. You may find them useful to gain more insight into their operation.

Tut1: Initializing Salvo and Starting to Multitask
Each working Salvo application is a combination of calls to Salvo
user services and application-specific code. Let's start using Salvo
by creating a multitasking application.
A minimal Salvo application is shown in Listing 23.
#include "main.h"
#include 
int main( void )
{
Init();
OSInit();
while (1) {

Salvo User Manual

63

OSSched();
}
}
Listing 23: A Minimal Salvo Application

This elementary program calls two Salvo user services whose
function prototypes are declared in salvo.h. OSInit() is called
once, and OSSched() is called over and over again from within an
infinite loop.

Tip These tutorials utilize a while

(1) { } construct in C to create an infinite loop. The for (;;) { } and do { } while (1)
constructs are functionally equivalent in terms of creating an infinite loop. All three are interchangeable for this puirpose.

Tip All user-callable Salvo functions are prefixed by "OS" or
"OS_".

Note The Init() function in main() is provided for device initialization.24 It and the header file main.h have nothing to do with
the Salvo code per se, but are provided for completeness.
OSInit()

initializes all of Salvo's data structures, pointers and
counters, and must be called before any other calls to Salvo functions. Failing to call OSInit() first before any other Salvo routines
may result in unpredictable behavior.

OSSched()

OSSched()

OSInit()

is Salvo's multitasking scheduler. Only tasks which are
in the eligible state can run, and each call to OSSched() results in
the most eligible task running until the next context switch within
that task. In order for multitasking to continue, OSSched() must be
called repeatedly.

Tip In order to make best use of your processor's call ... return
stack (whether hardware- or software-based), you should call OSSdirectly from main().

ched()

In Depth

Since there are no tasks eligible to run, the scheduler in Listing 23
has very little to do.

24

64

E.g. oscillator select and digital I/O crossbar select on Cygnal C8051F005
single-chip microcontroller.

Chapter 4 • Tutorial

Salvo User Manual

Tut2: Creating, Starting and Switching tasks
Multitasking requires eligible tasks that the scheduler can run. A
multitasking Salvo application with two tasks is shown in Listing
24.
#include "main.h"
#include 
void TaskA( void )
{
while (1) {
OS_Yield();
}
}
void TaskB( void )
{
while (1) {
OS_Yield();
}
}
int main( void )
{
Init();
OSInit();
OSCreateTask(TaskA, OSTCBP(1), 10);
OSCreateTask(TaskB, OSTCBP(2), 10);
while (1) {
OSSched();
}
}
Listing 24: A Multitasking Salvo Application with two
Tasks
TaskA() and TaskB() do nothing but run and context switch over
and over again. Since they both have the same priority (10), they
run one after the other, continuously, separated by trips through the
scheduler.

In order for multitasking to function properly, a running task must
return control to the scheduler. This occurs via a context switch (or
task switch) inside the task. Because it is designed to work without
a stack, Salvo only supports context switching at the task level.

Warning A Salvo context switch at a call ... return level below
that of the task (e.g. within a subroutine called by the task) will
cause unpredictable behavior.

Salvo User Manual

Chapter 4 • Tutorial

65

To multitask in Salvo, you must create and start tasks. Tasks are
functions that consist of an optional initialization / preamble followed by an infinite loop containing at least one context switch.
Salvo tasks cannot take any parameters. When the task is created
via OSCreateTask(), you explicitly assign an unused task control
block (tcb) to it and it is placed in the stopped state. A task can be
created in many parts of your program. Tasks are often created
prior to the start of multitasking, but they may also be created afterwards.
In order for a task to be able to run, it must be in the eligible state.
OSStartTask() can make a stopped task eligible. However, in the
interest of keeping the Salvo code size small, OSCreateTask()
automatically starts the task that it has created.25 Therefore a subsequent call to OSStartTask() is unnecessary if a Salvo task has
been created normally. Once a task is made eligible, it will run by
the scheduler as soon as it becomes the most eligible task, i.e. the
eligible task with the highest priority.

Tip When a group of eligible tasks all share the same priority,
they will execute one after the other in a round-robin fashion.
A stopped task can be started in many parts of your program.
Tasks can only be started after they are created. A task may be
started after multitasking begins.
OS_Yield()

Every task must context-switch at least once. OS_Yield() is
Salvo's unconditional context switcher. A common place to find
OS_Yield() would be at the bottom of, but still within, a task's
infinite loop.

Note All Salvo user services with conditional or unconditional
context switches are prefixed by "OS_".
OSCreateTask()

To create a task, call OSCreateTask() with a task starting address,26 a tcb pointer and a priority as parameters. The starting address is usually the start of the task, specified by the task's name.
Each task needs its own, unique tcb. The tcb contains all of the information Salvo needs to manage a task, like its start/resume address, state, priority, etc. There are OSTASKS tcbs available for use,
numbered from 1 to OSTASKS. The Salvo OSTCBP() macro is a
25
26

66

Optionally, the task can be left in the stopped state by using
OSDONT_START_TASK.
In C, this is equivalent to the name of the task (function).

Chapter 4 • Tutorial

Salvo User Manual

shorthanded27 way of specifying a pointer to a particular Salvo tcb,
e.g. OSTCBP(2) is a pointer to the second tcb. The task priority is
between 0 (highest) and 15 (lowest), and need not be unique to the
task. Once created, a task is in the stopped state.
The default behavior for OSCreateTask() is to also start the Salvo
task with the specified tcb pointer by making it eligible. It may be
a while before the task actually runs, depending on the priority of
the task, the states of any higher-priority tasks, and when the
scheduler will run again.

Tip Many Salvo services return error codes that you can use to
detect problems in your application. See Chapter 7 • Reference for
more information.
In Depth

Listing 24 illustrates some of the basic concepts of an RTOS –
tasks, task scheduling, task priorities and context switching. Tasks
are functions with a particular structure – infinite loops are commonly used. A task will run whenever it is the most eligible task,
and the scheduler decides which task is eligible based on the task
priorities. Since Salvo is a cooperative RTOS, each task must relinquish control back to the scheduler or else no other tasks will
have a chance to run. In this example, this is accomplished via
OS_Yield(). In the following examples, we'll use other context
switchers in place of OS_Yield().
While it's perhaps not immediately apparent, Listing 24 also illustrates another basic RTOS concept – that of the task state. In Salvo,
all tasks start out as destroyed – this is the state of an uninitialized
task. Creating a task changes its state to stopped, and starting a task
makes it eligible – i.e. it is now in the eligible state. When the task
is actually executing it's said to be running. In this example, after
being created and started, each task alternates between eligible and
running over and over again. And there's a short time period during
iteration of the main for() loop where neither task is running, i.e.
they're both eligible – that's when the scheduler is running.
Task scheduling in Salvo follows two very simple rules: First,
whichever task has the highest priority will run the next time the
scheduler is called. Second, all tasks with the same priority will
run in a round-robin manner as long as they are the most eligible
tasks. This means that they will run one after the other until they
have all run, and then the cycle repeats itself.

27

Salvo User Manual

&OStcbArea[n-1] is the longhanded way.

Chapter 4 • Tutorial

67

Tut3: Adding Functionality to Tasks
Listing 25 shows a multitasking application with two tasks that do
more than just context switch. We'll use more descriptive task
names this time.
#include "main.h"
#include 
unsigned int counter;
void TaskCount( void )
{
while (1) {
counter++;
OS_Yield();
}
}
void TaskShow( void )
{
InitPORT();
while (1) {
PORT = (PORT & ~0xFE)|((counter >> 8) & 0xFE);
OS_Yield();
}
}
int main( void )
{
Init();
OSInit();
OSCreateTask(TaskCount, OSTCBP(1), 10);
OSCreateTask(TaskShow, OSTCBP(2), 10);
counter = 0;
while (1) {
OSSched();
}
}
Listing 25: Multitasking with two Non-trivial Tasks

The two tasks in Listing 25 run independently of each other, and
they both access a shared global variable, a 16-bit counter. The
counter is initialized28 before multitasking begins. The first task
28

68

Strictly speaking, this initialization is unnecessary, as all ANSI compilers will
set counter to 0 before main().

Chapter 4 • Tutorial

Salvo User Manual

increments the counter every time it has a chance to run. The other
task takes the counter and outputs the upper 7 bits to an 8-bit port
(PORT) with 8 LEDs connected to it. This goes on indefinitely.

Note Since Salvo is a cooperative RTOS, only one task can access the global variable counter at a time in this example.
In Depth

In Listing 25, neither task actually runs until multitasking begins
with the call to the Salvo scheduler. Each time OSSched() is
called, it determines which task is most eligible to run, and transfers program execution to that particular task. Since both tasks
have the same priority, and are equally eligible to run, it is up to
Salvo to decide which task will run first.
In this particular example, TaskCount() will run first.29 It will
start by incrementing the counter, and will then context-switch via
OS_Yield(). This macro will make a note of where program execution is in TaskCount() (it's at the end of the for() loop), and
then return program execution to the scheduler. The scheduler then
examines TaskCount() to see if it's still eligible to continue running. In this case it is, because we made no changes to it, so it will
run again when it becomes the most eligible task.
The scheduler finishes its work, and is then called again because
it's in an infinite for() loop. This time, because Salvo roundrobins tasks of equal priority, the scheduler decides that TaskShow() is the most eligible task, and makes it run. First, PORT is
configured as an output port and initialized.30 Then TaskShow()
enters its infinite loop for the first time, PORT is initialized to 0x00
(the counter is now 0x0001), and once again OS_Yield() returns
program execution to the scheduler after noting where to "return
to" in TaskShow(). TaskShow() also remains eligible to run again.
After finishing its work, the scheduler is now called for the third
time. Once again, TaskCount() is the most eligible task, and so it
runs again. But this time it resumes execution where we last left it,
i.e. at the end of the for() loop. Since it's an infinite loop, execution resumes at the top of the loop. TaskCount() increments the
counter, and relinquishes control back to the scheduler.

29
30

Salvo User Manual

Because it was started first, and both tasks have the same priority.
In this example, each pin on I/O port PORT can be configured as an input or as
an output. At power-up, all pins are configured as inputs, hence the need to
configure them as outputs via InitPORT().InitPORT() also sets the 8-bit
I/O port's initial value to 0x00.

Chapter 4 • Tutorial

69

The next time the scheduler is called, TaskShow() resumes where
it left off, goes to the top of its for() loop, writes to PORT, and
yields back to the scheduler. This entire process of resuming a task
where it left off, running the task, and returning control back to the
scheduler is repeated indefinitely, with each task running alternately with every call to the scheduler.
When the program in Listing 25 runs, it gives the outward appearance of two separate things occurring simultaneously. Both tasks
are free-running, i.e. the faster the processor, the faster they'll run.
A counter appears to be incremented and sent to a port simultaneously. Yet we know that two separate tasks are involved, so we
refer to this program as a multitasking application. It's not very
powerful yet, and its functionality could be duplicated in many
other ways. But as we add to this application we'll see that using
Salvo will allow us to manage an increasingly sophisticated system
with a minimal coding effort, and we'll be able to maximize the
system's performance, too.

Tut4: Using Events for Better Performance
The previous example did not use one of an RTOS' most powerful
tools – intertask communications. It's also wasting processing
power, since TaskShow() runs continuously, but PORT changes
only once in every 512 calls to TaskCount(). Let's use intertask
communication to make more efficient use of our processing
power.
Listing 26 is shown below. We've used some #define preprocessor directives to improve legibility.
#include "main.h"
#include 
#define
#define
#define
#define
#define
#1 */

TASK_COUNT_P
OSTCBP(1) /* task #1 */
TASK_SHOW_P
OSTCBP(2) /* task #2 */
PRIO_COUNT
10 /* task priorities*/
PRIO_SHOW
10 /* ""
*/
BINSEM_UPDATE_PORT_P OSECBP(1) /* binsem

unsigned int counter;
void TaskCount( void )
{
while (1) {
counter++;
if (!(counter & 0x01FF)) {

70

Chapter 4 • Tutorial

Salvo User Manual

OSSignalBinSem(BINSEM_UPDATE_PORT_P);
}
OS_Yield();
}
}
void TaskShow( void )
{
InitPORT();
while (1) {
OS_WaitBinSem(BINSEM_UPDATE_PORT_P,
OSNO_TIMEOUT);
PORT = (PORT & ~0xFE)|((counter >> 8) & 0xFE);
}
}
int main( void )
{
Init();
OSInit();
OSCreateTask(TaskCount,
TASK_COUNT_P, PRIO_COUNT);
OSCreateTask(TaskShow,
TASK_SHOW_P, PRIO_SHOW);
OSCreateBinSem(BINSEM_UPDATE_PORT_P, 0);
counter = 0;
while (1) {
OSSched();
}
}
Listing 26: Multitasking with an Event

In Listing 26 we communicate between two tasks in order to update the port only when an update is required. We'll use a binary
semaphore to represent this event. We initialize it to 0, meaning
the event has not yet occurred. TaskCount() signals the binary
semaphore whenever the upper 7 bits of the counter change. TaskShow() waits for the event to occur, and then copies the upper 7
bits of the counter to PORT.
OSCreateBinSem()

Salvo User Manual

creates a binary semaphore with the specified
ecb pointer and initial value. A binary semaphore is created without any tasks waiting for it. A binary semaphore must be created
before it can be signaled or waited.
OSCreateBinSem()

Chapter 4 • Tutorial

71

OSSignalBinSem()

A binary semaphore is signaled via OSSignalBinSem(). If no task
is waiting the binary semaphore, then it is simply incremented. If
one or more tasks are waiting the binary semaphore, then the highest-priority waiting task is made eligible after signaling the binary
semaphore.

OS_WaitBinSem()

A task will wait a binary semaphore until the binary semaphore is
signaled. If the binary semaphore is zero when the tasks waits it,
then the task switches to the waiting state and returns through the
scheduler. It will keep waiting for the binary semaphore until the
binary semaphore is signaled and the task is the highest-priority
task waiting for the binary semaphore. That's because more than
one task can wait for a particular event.
If, on the other hand, the binary semaphore is 1 when the task
waits it, then the binary semaphore is reset to 0 and the task continues its execution without context switching.

Tip The "OS_" prefix in

OS_WaitBinSem() should remind you
that a context switch will unconditionally occur in every call to
OS_WaitBinSem(), regardless of the value of the binary semaphore. If the binSem is set (i.e. equal to 1) and the task is the highest-priority eligible task, then execution will continue in the task. If
not, execution in the task will resume at a later time when both of
these conditions are met.

Tip You must always specify a timeout31 when waiting a binary
semaphore via OS_WaitBinSem(). If you want the task to wait forever for the binary semaphore to be signaled, use the predefined
value OSNO_TIMEOUT.

Note In this example,

OS_WaitBinSem() is used in place of
OS_Yield(). In fact, the macro OS_WaitBinSem() includes a call
to OS_Yield(). You do not need to call OS_Yield() when using a
conditional context switcher like OS_WaitBinSem() – it does it for

you.
In Depth

In order to improve the performance of our application, we'd like
to update PORT only when the counter's upper 7 bits change. To do
this we will use a signaling mechanism between the two tasks,
called a binary semaphore. Here, the binary semaphore is a flag
31

72

The timeout parameter is required regardless of whether or not your
application is built with Salvo code (source files or libraries) that supports
timeouts. This makes it possible to rebuild applications for timeouts without
any user source code changes.

Chapter 4 • Tutorial

Salvo User Manual

that's initialized to zero to mean that there's no need to update the
port. When the binary semaphore is signaled, i.e. it is set to a value
of 1, it means that a PORT update is required.
Inter-task communication is achieved by using the binary semaphore to alert the waiting task (in this case, TaskShow()) that a
PORT update is required. This is done in TaskCount() by calling
OSSignalBinSem() with the parameter being a pointer to the binary semaphore, and by having TaskShow() wait the binary semaphore.
TaskCount() does not know which task(s) is(are) waiting
on the binary semaphore, and TaskShow() does not know how the
binary semaphore is signaled.

Note

The first time TaskShow() runs through the scheduler it calls
OS_WaitBinSem(). Since the binary semaphore was initialized to
zero, TaskShow() yields control back to the scheduler and changes
its state from eligible to waiting. Now there is only one eligible
task, TaskCount(), and the scheduler runs it repeatedly.
When TaskCount() finally signals the binary semaphore, TaskShow() is made eligible again and will run once TaskCount() returns through the scheduler. After all, since the counter's upper 7
bits change only every 512 calls to TaskCount(), there's no point
in running it more often than that. By using a binary semaphore,
TaskShow() runs only when it needs to update PORT. The rest of
the time, it is waiting and does not consume any processing power
(instruction cycles).
The performance of this application is roughly twice as good (i.e.
the counter increments at twice the speed) as that of Listing 25.
That's because a waiting task consumes no processor power whatsoever while it waits – recall that the scheduler only runs tasks that
are eligible. Since TaskShow() is waiting for the binary semaphore
over 97% of the time,32 it runs only on the rare occasion that the
counter's upper byte has changed. The rest of the time, the scheduler is running TaskCount().
It should be apparent that the calls to OS_WaitBinSem() and OSSignalBinSem() above implement some powerful functionality.
In this example, these Salvo event services control when TaskShow() will run by using a binary semaphore for intertask communications. Here the binary semaphore is a simple flag (1 bit of
32

Salvo User Manual

Measured on Test System A.

Chapter 4 • Tutorial

73

information). Salvo supports the use of binary and counting semaphores, as well as other mechanisms, to pass more information
(e.g. a count, or a pointer) from one task to another.
Listing 26 is a complete Salvo program – nothing is missing.
There's nothing "running in the background", nothing checking to
see if a waiting task should be made eligible, etc. In other words,
there's no polling going on – all of Salvo's actions are event-driven,
which contributes to its high performance. TaskShow() goes from
waiting to eligible in the call to OSSignalBinSem(), and from
running to waiting via OS_WaitBinSem(). With Salvo, you have
complete control over what the processor is doing at any one time,
and so you can optimize your program's performance without unwanted interference from the RTOS.

Tut5: Delaying a Task
One thing missing from the previous example is any notion of realtime performance – it just runs “open loop”. If we add other tasks
of equal or higher priority to the application, the rate at which the
counter increments will decline. Let's look at how an RTOS can
provide real-time performance by adding a task that runs at 2Hz,
regardless of what the rest of the system is doing. We'll do this by
repetitively delaying a task.
Being able to delay a task for a specified time period can be a very
useful feature. A task will remain in the delayed state, ineligible to
run, until the delay time specified has expired. It's up to the kernel
to monitor delays and return a delayed task to the eligible state.
The application in Listing 27 blinks the LED on the least significant bit of PORT at 1Hz by creating and running a task which delays itself 500ms after toggling the port bit, and does this
Pumprepeatedly.
This
program
is
located
in
kin\Salvo\Example\…\Tut\Tut5 in every Salvo distribution.
#include "main.h"
#include 
#define
#define
#define
#define
#define
#define
#define
#1 */

74

TASK_COUNT_P
OSTCBP(1) /* task #1 */
TASK_SHOW_P
OSTCBP(2) /* "" #2 */
TASK_BLINK_P
OSTCBP(3) /* "" #3 */
PRIO_COUNT
10 /* task priorities*/
PRIO_SHOW
10 /* ""
*/
PRIO_BLINK
2
/* ""
*/
BINSEM_UPDATE_PORT_P OSECBP(1) /* binSem

Chapter 4 • Tutorial

Salvo User Manual

unsigned int counter;
void TaskCount( void )
{
while (1) {
counter++;
if (!(counter & 0x01FF)) {
OSSignalBinSem(BINSEM_UPDATE_PORT_P);
}
OS_Yield();
}
}
void TaskShow( void )
{
while (1) {
OS_WaitBinSem(BINSEM_UPDATE_PORT_P,
OSNO_TIMEOUT);
PORT = (PORT & ~0xFE)|((counter >> 8) & 0xFE);
}
}
void TaskBlink( void )
{
InitPORT();
while (1) {
PORT ^= 0x01;
OS_Delay(50);
}
}
void main( void )
{
Init();
OSInit();
OSCreateTask(TaskCount,
TASK_COUNT_P, PRIO_COUNT);
OSCreateTask(TaskShow,
TASK_SHOW_P, PRIO_SHOW);
OSCreateTask(TaskBlink,
TASK_BLINK_P, PRIO_BLINK);
OSCreateBinSem(BINSEM_UPDATE_PORT_P, 0);
counter = 0;
enable_interrupts();
while (1) {
OSSched();

Salvo User Manual

Chapter 4 • Tutorial

75

}
}
Listing 27: Multitasking with a Delay

Additionally, interrupts are required to call OSTimer() at the desired system tick rate of 100Hz. The code to do this is located in
the source file tut5_sr.c that accompanies the project. An example for the PIC16 is shown below:33
#include 
#define TMR0_RELOAD 156 /* for 100Hz ints @ 4MHz
*/
void interrupt IntVector( void )
{
if (T0IE && T0IF) {
T0IF = 0;
TMR0 -= TMR0_RELOAD;
OSTimer();
}
}
Listing 28: Calling OSTimer() at the System Tick Rate

In order to use delays in a Salvo application, you must add the
Salvo system timer to it. In the above example we've added a 10ms
system timer by calling OSTimer() at a periodic rate of approximately 100Hz. The periodic rate is derived by a timer overflow,
which causes an interrupt. Interrupts must be enabled in order for
OSTimer() to be called – hence the call to enable_interrupts()
just prior to starting multitasking. Since delays are specified in
units of the system tick rate, the blink task is delayed by 50*10ms,
or 500ms.
OSTimer()

In order to use Salvo delay services, you must call OSTimer() at a
regular rate. This is usually done with a periodic interrupt. The rate
at which your application calls OSTimer() will determine the resolution of delays. If the periodic interrupt occurs every 10ms, by
calling OSTimer() from within the ISR you will have a system tick
period of 10ms, or a rate of 100Hz. With a tick rate defined, you
can specify delays to a resolution of one timer tick period, e.g. delays of 10ms, 20ms, ... 1s, 2s, ... are possible.

Note Salvo's timer features are highly configurable, with delays
of up to 32 bits of system ticks, and with an optional prescalar.
33

76

IntVector() is also used in tutorial # 6, below. IntVector() (and hence
the contents of tut5_isr.c) are target- and compiler-specific.

Chapter 4 • Tutorial

Salvo User Manual

Consult Chapter 5 • Configuration and Chapter 6 • Frequently
Asked Questions (FAQ) for more information.
OS_Delay()

With OSTimer() in place and called repetitively at the system tick
rate, you can now delay a task by replacing OS_Yield() with a call
to OS_Delay(), which will force the context switch and delay the
task for the number of system ticks specified. The task will automatically become eligible once the specified delay has expired.

In Depth

In Listing 27, each time TaskBlink() runs, it delays itself by
500ms and enters the delayed state upon returning to the scheduler.
When TaskBlink()'s delay expires 500ms later it is automatically
made eligible again, and will run after the current (running) task
context-switches. That's because TaskBlink() has a higher priority than either TaskCount() or TaskShow(). By making TaskBlink() the highest-priority task in our application, we are
guaranteed a minimum of delay (latency) between the expiration of
the delay timer and when TaskBlink() toggles bit 0 of PORT.
Therefore TaskBlink() will run every 500ms with minimal latency, irrespective of what the other tasks are doing.

Tip If

TaskBlink() had the same priority as TaskCount() and
TaskShow(), it would occasionally remain eligible (and would not
run) while both TaskCount() and TaskShow() ran before it. Its
maximum latency would increase. If TaskBlink() had a lower

priority, it would never run at all.
The initialization of PORT was moved to TaskBlink() because of
TaskBlink()'s priority. It will be the first task to run, and therefore PORT will be initialized as an output before TaskShow() runs
for the first time.
Salvo monitors delayed tasks once per call to OSTimer(), and the
overhead is independent of the number of delayed tasks.34
This illustrates that the system timer is useful for a variety of reasons. A single processor resource (e.g. a periodic interrupt) can be
used in conjunction with OSTimer() to delay an unlimited number
of tasks. More importantly, delayed tasks consume only a very
small amount of processing power while they are delayed, much
less than running tasks.

34

Salvo User Manual

Except when one or more task delays expire simultaneously.

Chapter 4 • Tutorial

77

Signaling from Multiple Tasks
A multitasking approach to programming delivers real benefits
when priorities are put to good use and program functionality is
clearly delineated along task lines.
Review the code in Listing 29 to see what happens when we lower
the priority of the always-running task, TaskCount(), and have
TaskShow() handle all writes to PORT. This program is located in
Pumpkin\Salvo\tut\tu6\main.c.
#include "main.h"
#include 
#define
#define
#define
#define
#define
#define
#define

TASK_COUNT_P
TASK_SHOW_P
TASK_BLINK_P
PRIO_COUNT
PRIO_SHOW
PRIO_BLINK
MSG_UPDATE_PORT_P

OSTCBP(1) /* task #1 */
OSTCBP(2) /* "" #2 */
OSTCBP(3) /* "" #3 */
12 /* task priorities*/
10 /* ""
*/
2
/* ""
*/
OSECBP(1) /* sem #1 */

unsigned int counter;
char CODE_B = 'B';
char CODE_C = 'C';
void TaskCount( void )
{
counter = 0;
while (1) {
counter++;
if (!(counter & 0x01FF)) {
OSSignalMsg(MSG_UPDATE_PORT_P,
(OStypeMsgP) &CODE_C);
}
OS_Yield();
}
}
void TaskShow( void )
{
OStypeMsgP msgP;
InitPORT();
while (1) {
OS_WaitMsg(MSG_UPDATE_PORT_P, &msgP,
OSNO_TIMEOUT);
if (*(char *)msgP == CODE_C) {
PORT = (PORT & ~0xFE)|((counter >> 8)&0xFE);

78

Chapter 4 • Tutorial

Salvo User Manual

}
else {
PORT ^= 0x01;
}
}
}
void TaskBlink( void )
{
OStypeErr err;
while (1) {
OS_Delay(50);
err = OSSignalMsg(MSG_UPDATE_PORT_P,
(OStypeMsgP) &CODE_B);
if (err == OSERR_EVENT_FULL) {
OS_SetPrio(PRIO_SHOW+1);
OSSignalMsg(MSG_UPDATE_PORT_P,
(OStypeMsgP) &CODE_B);
OSSetPrio(PRIO_BLINK);
}
}
}
void main( void )
{
Init();
OSInit();
OSCreateTask(TaskCount,
TASK_COUNT_P, PRIO_COUNT);
OSCreateTask(TaskShow,
TASK_SHOW_P, PRIO_SHOW);
OSCreateTask(TaskBlink,
TASK_BLINK,
PRIO_BLINK);
OSCreateMsg(MSG_UPDATE_PORT_P, (OStypeMsgP) 0);
enable_interrupts();
while (1) {
OSSched();
}
}
Listing 29: Signaling from Multiple Tasks

In Listing 29 we've made two changes to the previous program.
First, TaskShow() now handles all writes to PORT. Both TaskCount() and TaskBlink() send a unique message to TaskShow()
(the character ‘C' for "count" or ‘B' for "blink", respectively)
which it then interprets to either show the counter on the port or

Salvo User Manual

Chapter 4 • Tutorial

79

toggle the least significant bit of the port. Second, we've lowered
the priority of TaskCount() by creating it with a lower priority.
OSCreateMsg()

is used to initialize a message. Salvo has a defined
type for messages, and requires that you initialize the message
properly. A message is created without any tasks waiting for it. A
message must be created before it can be signaled or waited.
OSCreateMsg()

Note Salvo services require that you interface your code using
predefined types, e.g. OStypeMsgP for message pointers. You
should use Salvo's predefined types wherever possible. See
Chapter 7 • Reference for more information on Salvo's predefined
types.
OSSignalMsg()

In order to signal a message with OSSignalMsg(), you must specify both a ecb pointer and a pointer to the message contents. If no
task is waiting the message, then the message gets the pointer,
unless the message is already defined, in which case an error has
occurred. If one or more tasks are waiting the message, then the
highest-priority waiting task is made eligible. You must correctly
typecast the message pointer so that it can be dereferenced properly by whichever tasks wait the message.

OS_WaitMsg()

A task waits a message via OS_WaitMsg(). The message is returned to the task through a message pointer. In order to extract the
contents of the message, you must dereference the pointer with a
typecast matching what the message pointer is pointing to.

OS_SetPrio()

A task can change its priority and context-switch immediately
thereafter using OS_SetPrio().

OSSetPrio()

A task can change its priority using OSSetPrio(). The new priority will take effect as soon as the task yields to the scheduler.

In Depth

TaskShow() is now the only task writing to PORT. A single message is all that is required to pass unique information from two different tasks (which run at entirely different rates) to TaskShow().
In this case, the message is a pointer to a 1-byte constant. Since
messages contain pointers, casting and proper dereferencing are
required to send and receive the intended information in the message.

In Listing 29, the following scenario is possible: Immediately after
TaskCount() signals the message, TaskBlink()'s delay expires
and TaskBlink() is made eligible to run. Since TaskBlink() has
the highest priority, the message will still be present when Task-

80

Chapter 4 • Tutorial

Salvo User Manual

signals the message. Therefore OSSignalMsg() will return an error. The LED's PORT pin will fail to toggle …
Blink()

This example illustrates the use of return values for Salvo services.
By testing for the abovementioned error condition, we can guarantee the proper results by temporarily lowering TaskBlink()'s priority and yielding to the scheduler before signaling the message
again. TaskShow() will temporarily be the highest-priority task,
and it will "claim" the message. As long as TaskCount() does not
signal messages faster than once every three context switches, this
solution remains a robust one.35
In a more sophisticated application, e.g. a car's electronics, one can
imagine TaskShow() being replaced with a task that drives a
dashboard display divided into distinct regions. Four tasks would
monitor information (e.g. rpm, speed, oil pressure and water temperature) and would pass it on by signaling a message whenever a
parameter changed. TaskShow() would wait for this message.
Each message would indicate where to display the parameter, what
color(s) to use (e.g. red on overtemperature) and the parameter's
new value. Since visual displays generally have low refresh rates,
TaskShow() could run at a lower priority than the sending tasks.
These tasks would run at higher priority so as to process the information they are sampling without undue interference from the slow
display task. For example, the oil-pressure-monitoring task might
run at the highest priority, since a loss of oil pressure means certain
engine destruction. By having the display functionality in a task
instead of in a callable function, you can fine-tune the performance
of your program by assigning an appropriate priority to each of the
tasks involved.
By lowering TaskCount()'s priority we've changed the behavior of
our application. PORT updates now take precedence over the
counter incrementing. This means that PORT updates will occur
sooner after the message is signaled. The counter now increments
only when there's nothing else to do. You can dramatically and
predictably alter the behavior of your program by changing just the
priority when creating a task.

Wrapping Up
As a Salvo user you do not have to worry about scheduling, tasks
states, event management or intertask communication. Salvo han35

Salvo User Manual

An alternative solution to this problem would be to use a message queue with
room for two messages in it.

Chapter 4 • Tutorial

81

dles all of that for you automatically and efficiently. You need only
create and use the tasks and events in the proper manner to get all
of this functionality, and more.
Chapter 7 • Reference contains working examples with
commented C source code for every Salvo user service. Refer to
them for more information on how to use tasks and events.

Note

Food For Thought
Now that you're writing code with task- and event-based structures
like the ones Salvo provides, you may find it useful or even necessary to change the way you approach new programs. Instead of
worrying about how many processor resources, ISRs, global variables and clock cycles your application will require, focus instead
on the tasks at hand, their priorities and purposes, your application's timing requirements and what events drive its overall behavior. Then put it all together with properly prioritized tasks that use
events to control their execution and to communicate inside your
program.

Part 2: Building a Salvo Application
Note If you have not done so already, please follow the instructions in Chapter 3 • Installation to install all of Salvo's components
onto your computer. You may also find it useful to refer to
Chapter 5 • Configuration and Chapter 7 • Reference for more information on some of the topics mentioned below. Lastly, you
should review the Salvo Application Note that covers building applications with your compiler. Refer to your compiler's Salvo
Compiler Reference Manual for particulars.
Now that you are familiar with how to write a Salvo application,
it's time to build an executable program. Below you will find general instructions on building a Salvo application.

Working Environment
Salvo is distributed as a collection of source code files, object files,
library files and other support files. Since all source code is provided in Salvo Pro, Salvo can be compiled on many development
platforms. You will need to be proficient with your editor / com-

82

Chapter 4 • Tutorial

Salvo User Manual

piler / integrated development environment (IDE) in order to successfully compile a Salvo application.
You should be familiar with the concepts of including a file inside
another file, compiling a file, linking one or more files, working
with libraries, creating an executable program, viewing the debugging output of your compiler, and placing your program into memory.
Please refer to your editor's / compiler's / IDE's documentation on
how to include files into source code, compile source code, link to
separate object modules, and compile and link to libraries.
Many IDEs support an automatic make-type utility. You will
probably find this very useful when working with Salvo. If you do
not have a make utility, you may want to investigate obtaining one.
Both commercial and freeware / shareware make utilities exist, for
command-line hosts (e.g. DOS) and Windows 95 / 98 / 2000 / NT.

Creating a Project Directory
In creating an application with Salvo you'll include Salvo source
files in your own source code, and you'll probably also link to
Salvo object files or Salvo libraries. We strongly recommend that
you do not modify any Salvo files directly,36 nor should you duplicate any Salvo files unnecessarily. Unless you intend to make
changes to the Salvo source code, you should not change any of
Salvo's files.
By creating a working directory for each new Salvo application
you write, you'll be able to:
•
•
•
•

• minimize hard disk usage,
• manage your files better,
• make changes to one application without
affecting any others, and
• compile unique versions of Salvo libraries for
different projects.

•
Note Complete projects for certain tutorial programs can be
found in Pumpkin\Salvo\Tut.

36

Salvo User Manual

Salvo source files are installed as read-only.

Chapter 4 • Tutorial

83

Including salvo.h
Salvo's main header file, salvo.h, must be included in each of
your source files that use Salvo. You can do this by inserting
#include 

into each of your source files that calls Salvo services. You may
also need to configure your development tools to add Salvo's home
directory (usually C:\Pumpkin\Salvo) to your tools' system include path – see Setting Search Paths, below.

Note Using
#include "salvo.h"

is not recommended.

Tip If you include a project header file (e.g. myproject.h) in all
of your source files, you may want to include salvo.h in it.
Including salvo.h will automatically include your projectspecific version of salvocfg.h (see Setting Configuration Options,
below). You should not include salvocfg.h in any of your source
files – just including salvo.h is enough.

Note salvo.h has a built-in "include guard" which will prevent
problems when multiple references to include salvo.h are contained in a single source file.

Configuring your Compiler
In order to successfully compile your Salvo application you must
configure your compiler for use with the Salvo source files and
libraries. You have several options available to you when combining your code with the Salvo source code in order to build an application.

Setting Search Paths
First, you must specify the appropriate search paths so that the
compiler can find the necessary Salvo include (*.h) and source
(*.c) files.

84

Chapter 4 • Tutorial

Salvo User Manual

Tip All of Salvo's supported compilers support explicit search
paths. Therefore you should never copy Salvo files from their
source directories to your project directory in order to have the
compiler find them by virtue of the fact that it's in the current directory.
At the very least, your compiler will need to know where to find
the following files:
•
•

• salvo.h, located in Pumpkin\Salvo\inc
• salvocfg.h, located in your current project
directory

You may also need to specify the Salvo source file directory
(Pumpkin\Salvo\Src) if you have Salvo Pro and plan to include
Salvo source files in your own source files (see below).

Using Libraries vs. Using Source Files
Different methods for incorporating Salvo into your application are
outlined below. Linking to Salvo libraries is the simplest method,
but has limitations. Including the Salvo source files in your project
is the most flexible method, but isn't as simple, and requires Salvo
Pro. Creating custom Salvo libraries from the source files is for
advanced Salvo Pro users.

Tip You may find

Figure 25: Salvo Library Build Overview and
Figure 26: Salvo Source-Code Build Overview useful in

understanding the process of building a Salvo application.

Using Libraries
Just like a C compiler's library functions – e.g. rand() in the standard library (stdlib.h) or printf() in the standard I/O library
(stdio.h) – Salvo has functions (called user services) contained in
libraries. Unlike a compiler's library functions, Salvo's user services are highly configurable – i.e. their behavior can be controlled
based on the functionality you desire in your application. Each
Salvo library contains user functions compiled for a particular set
of configuration options. There are many different Salvo libraries.

Note Configuration options are compile-time tools used to configure Salvo's source code and generate libraries. Therefore the

Salvo User Manual

Chapter 4 • Tutorial

85

functionality of a precompiled library cannot be changed through
configuration options. To change a library's functionality, it must
be regenerated (i.e. re-compiled) with Salvo Pro and new configuration options.
In order to facilitate getting started, all Salvo distributions contain
libraries with most of Salvo's functionality already included. As a
beginner, you should start by using the libraries to build your applications. This way, you don't have to concern yourself with the
myriad of configuration options.

Tip The easiest and quickest way to create a working application
is to link your source code to the appropriate Salvo library. The
compiler-specific Salvo Application Notes describe in detail how
to create applications for each compiler.
Complete library-based projects for all the tutorial programs can be
found in Pumpkin\Salvo\tut\tu1-tu6. See Appendix C • File
and Program Descriptions for more information.

Using Source Files
Salvo is configurable primarily to minimize the size of the user
services and thus conserve ROM. Also, its configurability aids in
minimizing RAM usage. Without it, Salvo's user services and variables might be too large to be of any use in many applications. All
of this has its advantages and disadvantages – on the one hand, you
can fine-tune Salvo to use just the right amount of ROM and RAM
in your application. On the other hand, it can be a challenge learning how all the different configuration options work.
There are some instances where it's better to create your application by adding the Salvo source files as nodes to your project.
When you use this method, you can change configuration options
and re-build the application to have those changes take effect in the
Salvo source code. Only Salvo Pro includes source files. The rest
of this chapter covers this approach.

Setting Configuration Options
Salvo is highly configurable. You'll need to create and use a configuration file, salvocfg.h, for each new application you write.
This simple text file is used to select Salvo's compile-time configuration options, which affect things like how many tasks and events

86

Chapter 4 • Tutorial

Salvo User Manual

your application can use. All configuration options have default
values – most of them may be acceptable to your application.

Note Whenever you redefine a configuration option in

sal-

vocfg.h,

you must recompile all of the Salvo source files in your
application.
The examples below assume that you are creating and editing salvocfg.h via a text editor. Each configuration option is set via a Clanguage #define statement. For example, to configure Salvo to
support 16-bit delays, you would add
#define OSBYTES_OF_DELAYS 2

•
to your project's salvocfg.h file. Without this particular line, this
configuration option would be automatically set to its default (in
this case, 8-bit delays).

Note The name and value of the configuration option are casesensitive. If you type the name incorrectly, the intended option will
be overridden by the Salvo default.

Identifying the Compiler and Target Processor

Normally, Salvo automatically detects which compiler and target
processor you are using. It does this by detecting the presence of
certain predefined symbols provided by the compiler.

Specifying the Number of Tasks

Memory for Salvo's internal task structures is allocated at compile
time. You must specify in salvocfg.h how many tasks you would
like supported in your application, e.g.:
#define OSTASKS 4

You do not need to use all the tasks that you allocate memory for,
nor must you use their respective tcb pointers (numbered from
OSTCBP(1) to OSTCBP(OSTASKS)) consecutively. If you attempt to
reference a task for which no memory was allocated, the Salvo
user service will return a warning code.

Tip Tasks are referred to in Salvo by their tcb pointers. It's recommended that you use descriptive designations in your code to

Salvo User Manual

Chapter 4 • Tutorial

87

refer to your tasks. This is most easily done by using the #define
statement in your project's main header (.h) file, e.g.:
#define TASK_CHECK_TEMP_P37 OSTCBP(1)
#define TASK_MEAS_SPEED_P OSTCBP(2)
#define TASK_DISP_RPM_P
OSTCBP(3)

Your program will be easier to understand when calling Salvo task
services with meaningful names like these.

Specifying the Number of Events

Memory for Salvo's internal event structures is also allocated at
compile time. You must specify in salvocfg.h how many events
you would like supported in your application, e.g.:
#define OSEVENTS 3

Events include semaphores (binary and counting), messages and
message queues.
You do not need to use all the events that you allocate memory for,
nor must you use their respective ecb pointers (numbered from
OSECBP(1) to OSECBP(OSEVENTS)) consecutively. If you attempt
to reference an event for which no memory was allocated, the
Salvo user service will return a warning code.
If your application does not use events, leave OSEVENTS undefined
in your salvocfg.h, or set it to 0.

Tip You should use descriptive names for events, too. See the tip
above on how to do this.

Specifying other Configuration Options

You may also need to specify other configuration options, depending on which of Salvo's features you plan to use in your application. Many of Salvo's features are not available until they are
enabled via a configuration option. This is done to minimize the
size of the code that Salvo adds to your application. For small projects, a small salvocfg.h may be adequate. For larger projects and
more complex applications, you will need to select the appropriate

37

88

The P suffix is there to remind you that the object is a Pointer to something.

Chapter 4 • Tutorial

Salvo User Manual

configuration option(s) for all the features you wish to use. Other
configuration options include:
•
•
•

• the size of delays, counters, etc. in bytes,
• the size of semaphores and message pointers,
and
• memory-locating directives specific to the
compiler.

Tip If you attempt to use a Salvo feature by calling a Salvo function and your compiler issues an error message suggesting that it
can't find the function, this may be because the function has not
been enabled via a configuration option.
In a sophisticated application, some of the additional configuration
options might be:
#define OSBYTES_OF_DELAYS
#define OSTIMER_PRESCALAR
#define OSLOC_ECB

3
20
bank3

The values for the options will either be numeric constants, predefined constants (e.g. TRUE and FALSE), or definitions provided for
the compiler in use (e.g. bank3, used by the HI-TECH PICC compiler to locate variables in a particular bank of memory).

salvocfg.h Example – Salvo's Tut5 Application

Because the tutorial program is relatively simple, only a few configuration options need to be defined in salvocfg.h. By starting
with an empty salvocfg.h, we begin with all configurations at
their default values.
For three tasks and one event, we'll need the following #define
directives.
#define OSTASKS 3
#define OSEVENTS 1

Next, Pumpkin\Salvo\Tut\Tut5 uses binary semaphores as a
means of intertask communications. Binary Semaphore code is
disabled by default, so we enable it with:
#define OSENABLE_BINARY_SEMAPHORES TRUE

Salvo User Manual

Chapter 4 • Tutorial

89

Lastly, because we're using delays, we need to specify the size of
possible delays.
#define OSBYTES_OF_DELAYS 1

This configuration option must be specified because Salvo defaults
to no support for delays, which keeps RAM requirements to a
minimum. Since TaskBlink() delays itself for 50 system ticks, a
single byte is all that is required. With a byte for delays, each task
could delay itself for up to 255 system ticks with a single call to
OS_Delay().

Note The #defines in salvocfg.h may appear in any order.
This four-line salvocfg.h is typical for small- to medium-sized
programs of moderate complexity. The complete Salvo configuration file for this program can be found in Pumpkin\Salvo\Tut\Tut5. It is shown (with C comments removed38)
in Listing 30.
#define
#define
#define
#define

OSBYTES_OF_DELAYS
OSENABLE_BINARY_SEMAPHORES
OSEVENTS
OSTASKS

1
TRUE
1
3

Listing 30: salvocfg.h for Tutorial Program

Linking to Salvo Object Files
You can create an application by compiling and then linking your
application to some or all of Salvo's *.c source files. This method
is recommended for most applications, and is compatible with
make utilities. It is relatively straightforward, but has the disadvantage that your final executable may contain all of the Salvo functionality contained in the linked files, regardless of whether your
application uses them or not.

Note Some compilers are capable of "smart linking" whereby
functions that are linked but not used do not make it into the final
executable. In this situation there is no downside to linking your
application to all of Salvo's source files.

38

90

And without the additional configuration options that match those of the
associated freeware library.

Chapter 4 • Tutorial

Salvo User Manual

Chapter 7 • Reference contains descriptions of all the Salvo user

services, and the Salvo source files that contain them. As soon as
you use a service in your code, you'll also need to link to the appropriate source file. This is usually done in the compiler's IDE by
adding the Salvo source files to your project. If you use the service
without adding the file, you will get a link error when you make
your project.
The size of each compiled object module is highly dependent on
the configuration options you choose. Also, you can judiciously
choose which modules to compile and link to – for example, if
don't plan on using dynamic task priorities in your application, you
can modify salvocfg.h appropriately and leave out prio.c, for a
reduction in code size.

Tip The compiler-specific

Salvo Application Notes describe in

detail how to create applications for each compiler.
Complete source-code-based projects for certain tutorial programs
can be found in Pumpkin\Salvo\Tut. See Appendix C • File and
Program Descriptions for more information.

Salvo User Manual

Chapter 4 • Tutorial

91

92

Chapter 4 • Tutorial

Salvo User Manual

Chapter 5 • Configuration

Introduction
The Salvo source code contains configuration options that you can
use to tailor its linkable object code to the specific needs of your
application. These options are used to identify the compiler you're
using and the processor you're compiling for, to configure Salvo
for the number of tasks and events your application will require,
and to enable or disable support for certain services. By selecting
various configuration options you can fine-tune Salvo's abilities
and performance to best match your application.

Note All configuration options are in the form of C preprocessor
statements. They are therefore compile-time options.
This means that they will not take effect until / unless you recompile each Salvo source code file that is affected by the configuration option.
#define

The Salvo Build Process
Salvo applications are typically built in one of two ways – as a library build, or as a source-code build. Understanding Salvo's build
process will aid in your understanding of how Salvo's configuration options are applied.

Note See your compiler's Salvo Compiler Reference Manual and
the associated Salvo Application Note(s) for detailed information
on creating and building Salvo projects.

Library Builds
Source-code builds are available in all Salvo distributions.
In a library build, a Salvo application is built from user source
code (C and Assembly), from a precompiled Salvo library and
from Salvo's salvomem.c. The user C source code makes calls to
Salvo services that are contained in the Salvo library. Additionally,
Salvo's global objects (i.e. its task control blocks, etc.) are in Pump-

Salvo User Manual

93

kin\Salvo\Src\salvomem.c.

Since the size of these objects is
dependent on the application's numbers of tasks, events, etc., it
must be re-compiled each time the project's Salvo configuration –
defined in the project's salvocfg.h file – is changed.
Figure 25 presents an overview of the Salvo library build process.
In a library build, the configuration options in the project's salvocfg.h can only affect the user C source files and Salvo's salvomem.c. None of the Salvo services – contained in the Salvo library
– are affected by the configuration options in salvocfg.h.
It is essential that the configuration options used to build the Salvo
library match those applied to the user's C source files and to salvomem.c. Therefore part of the salvocfg.h for a library build
(OSUSE_LIBRARY, OSLIBRARY_XYZ) is used to recreate the entire set
of Salvo configuration options in place when the library was compiled. This is done automatically for the user by defining configuration options in salvolib.h based on the salvocfg.h settings,
and by setting any undefined configuration options to their default
values in salvo.h. The remaining configuration options in salvocfg.h simply set the sizes of Salvo's various global objects (e.g.
the number of task control blocks). salvoclcN.h is included in the
mix if a custom library is used.
For a successful library build, the chosen library must match the
library options specified in salvocfg.h. See Chapter 8 • Libraries
and your compiler's Salvo Compiler Reference Manual for more
information on salvocfg.h for library builds.

94

Chapter 5 • Configuration

Salvo User Manual

User C Source
Files
salvo.h
salvocfg.h
salvolib.h
salvoclcN.h

Salvo's
salvomem.c
salvo.h
salvocfg.h
salvolib.h
salvoclcN.h

User Assembly
Files

salvoXyz.h
salvomem.c

main.c, ...

C Preprocessor

adc.asm, ...

C Compiler

Object Files

main.obj, ...

Assembler

Object File

Object Files

salvomem.obj

adc.obj, ...

Salvo Library
File

salvoXyz.lib

Linker

User Salvo
Configuration
File

Salvo Application
File

OSUSE_LIBRARY
OSLIBRARY_TYPE
OSLIBRARY_CONFIG
...
OSTASKS
OSEVENTS
...

salvocfg.h

main.hex

Figure 25: Salvo Library Build Overview

Salvo User Manual

Chapter 5 • Configuration

95

Source-Code Builds
Source-code builds are only available in Salvo Pro distributions.
In a source-code build, a Salvo application is built from user
source code (C and Assembly) and from Salvo source code (C and
Assembly, where applicable), including Salvo's salvomem.c. The
user C source code makes calls to Salvo services that are contained
in the Salvo source code. Again, Salvo's global objects (i.e. its task
control blocks, etc.) are in \Pumpkin\Salvo\Src\salvomem.c. In
a source-code build, all of Salvo's source-code modules must be recompiled each time the project's Salvo configuration – defined in
the project's salvocfg.h file – is changed.
Figure 26 presents an overview of the Salvo source-code build
process.
In a source-code build, the configuration options in the project's
affect the user C source files and all of Salvo's C
source files, where the desired user services are contained.
salvocfg.h

Each configuration option that the user wishes to set to a nondefault value must be defined in salvocfg.h. All other configuration options are automatically set to their default values in
salvo.h. As in a library build, certain configuration options (e.g.
OSTASKS) set the sizes of Salvo's various global objects (e.g. the
number of task control blocks).

96

Chapter 5 • Configuration

Salvo User Manual

User C Source
Files

Salvo C Source
Files

salvo.h
salvocfg.h

salvo.h
salvocfg.h

salvomem.c,
salvosched.c,
...

main.c,
...

C Preprocessor

adc.asm, ...

C Compiler

Object Files

main.obj, ...

User Assembly
Files

Salvo Assembly
File

salvoportXyz.asm

Assembler

Object Files

salvomem.obj,
salvosched.obj,
...

Object Files

adc.obj, ...

Object File

salvoportXyz.obj

Linker

User Salvo
Configuration
File

Salvo Application
File

OSTASKS
OSEVENTS
OSBYTES_OF_DELAYS
OSENABLE_...
OSDISABLE_...
OSUSE...
...

salvocfg.h

main.hex

Figure 26: Salvo Source-Code Build Overview

Salvo User Manual

Chapter 5 • Configuration

97

Benefits of Different Build Types
Library builds have the advantage that all of the Salvo services are
available in the library, and the linker will add only those necessary when building the application. The disadvantage is that if a
different library configuration is required, both the salvocfg.h
and the project file must be edited to ensure a match between the
desired library and the library that linker sees.
With a source-code build, Salvo can be completely reconfigured
just by simply adding or changing entries in salvocfg.h, and by
adding the required Salvo source files to the project.

Note Salvo Pro is required for source-code builds.
Another benefit of library builds is that rebuilding a project within
a Makefile-driven system is faster, since the library need not be
rebuilt when allowable changes (e.g. changing the number of
tasks) are made to salvocfg.h.

Configuration Option Overview
This section describes the Salvo configuration options. Each description includes information on:
•
•
•
•
•
•
•
•
•

• the name of the configuration option,
• the purpose of the configuration option,
• the allowed values for the configuration
option,
• the default value for the configuration option,
• the compile-time action that results from the
configuration option,
• related configuration options,
• which user services are enabled by the
configuration option,
• how it affects memory requirements39 and
• notes particular to the configuration option.

You can fine-tune Salvo's capabilities, performance and size by
choosing configuration options appropriate to your application.

39

98

ROM requirements are described as small (e.g. a few lines of code in a single
function) to considerable (e.g. a few lines of code in nearly every function).

Chapter 5 • Configuration

Salvo User Manual

Note All configuration options are contained in the user file
salvocfg.h,
salvocfg.h

and should not be placed in any other file(s).
should be located in the same directory as your
application's source files. See Chapter 4 • Tutorial for more
information on salvocfg.h.

Caution Whenever a configuration option is changed in
salvocfg.h,

you must recompile all of the Salvo files in your
application. Failing to do so may result in unpredictable behavior
or erroneous results.

Configuration Options for all Distributions
The configuration options described in this section can be used
with:
•
•
•
•
•
•

• Salvo Lite
• Salvo tiny
• Salvo SE
• Salvo LE
• Salvo Pro
• Salvo Developer

and are listed in alphabetical order.
These configuration options affect the Salvo header (*.h) files, as
well as salvomem.c.

Salvo User Manual

Chapter 5 • Configuration

99

OSCOMPILER: Identify Compiler in Use
Name:
Purpose:
Allowed Values:
Default Value:
Action:
Related:
Enables:
Memory Required:
Notes

OSCOMPILER

To identify the compiler you're using to
generate your Salvo application.
see salvo.h
OSUNDEF, or automatically defined for certain compilers.
Configures Salvo source code for use with
the selected compiler.
OSTARGET

–
n/a

This configuration option is used within the Salvo source code
primarily to implement non-ANSI C directives like in-line assembly instructions and #pragma directives.
Salvo automatically detects the presence of nearly all of Salvo's
supported compilers, and sets OSCOMPILER accordingly.40 Therefore it is usually unnecessary to define OSCOMPILER in salvocfg.h.
If you are working with an as-yet-unsupported compiler, use
OSUNDEF and refer to Chapter 10 • Porting for further instructions.

40

100

OSCOMPILER can be overridden by setting it in salvocfg.h.

Chapter 5 • Configuration

Salvo User Manual

OSEVENTS: Set Maximum Number of Events
Name:
Purpose:

Allowed Values:
Default Value:
Action:
Related:

Enables:
Memory Required:

Notes

OSEVENTS

To allocate memory at compile time for
event control blocks (ecbs), and to set an
upper limit on the number of supported
events.
0 or greater.
0

Configures Salvo source code to support
the desired number of events.
OSENABLE_BINARY_SEMAPHORES,
OSENABLE_EVENT_FLAGS,
OSENABLE_EVENTS, OSENABLE_MESSAGES,
OSENABLE_MESSAGE_QUEUES,
OSENABLE_SEMAPHORES, OSEVENT_FLAGS,
OSTASKS, OSMESSAGE_QUEUES
event-related services
When non-zero, requires a configurationdependent amount of RAM for each ecb.

Events (event flags, all semaphores, messages and message
queues) are numbered from 1 to OSEVENTS.
Since event memory is allocated at compile time, the ecb memory
will be used whether or not the event is actually created via OSCreateBinSem|Eflag|Msg|MsgQ|Sem().
On a typical 8-bit processor, the amount of memory required by
each event is 2-4 bytes41 depending on which configuration options
are enabled.

41

Salvo User Manual

For the purposes of these size estimates, pointers to ROM memory are
assumed to be 16 bits, and pointers to RAM memory are assumed to be 8 bits.
This is the situation for the PIC16 and PIC17 family of processors.

Chapter 5 • Configuration

101

OSEVENT_FLAGS: Set Maximum Number of Event Flags
Name:
Purpose:

Allowed Values:
Default Value:
Action:
Related:
Enables:
Memory Required:

Notes

OSEVENT_FLAGS

To allocate memory at compile time for
event flag control blocks (efcbs), and to
set an upper limit on the number of supported event flags.
1 or greater.
1 if OSENABLE_EVENT_FLAGS is TRUE, 0
otherwise
Configures Salvo source code to support
the desired number of event flags.
OSENABLE_EVENT_FLAGS, OSLOC_EFCB,
-

When non-zero, requires a configurationdependent amount of RAM for each efcb.

This configuration parameter allocates RAM for event flag control
blocks. Event flags require no other additional memory.
Event flags are numbered from 1 to OSEVENT_FLAGS.
Since event flag memory is allocated at compile time, the efcb
memory will be used whether or not the event flag is actually created via OSCreateEFlag().
On a typical 8-bit processor, the amount of memory required by
each event flag control block is represented by
OSBYTES_OF_EVENT_FLAGS.

102

Chapter 5 • Configuration

Salvo User Manual

OSLIBRARY_CONFIG: Specify Precompiled Library
Configuration
Name:
Purpose:

Allowed Values:
Default Value:
Action:
Related:
Enables:
Memory Required:
Notes

OSLIBRARY_CONFIG

To guarantee that an application's source
files are compiled using the same salvocfg.h as was used to create the specified precompiled library.
OSA, OSD, OSE, OSM, OSS, OST, OSY
not defined
Sets the configuration options inside salvolib.h to match those used to generate
the library specified.
OSLIBRARY_TYPE, OSLIBRARY_GLOBALS,
OSLIBRARY_VARIANT, OSUSE_LIBRARY
–
n/a

OSLIBRARY_CONFIG
is
used
in
OSLIBRARY_GLOBALS,
OSLIBRARY_OPTION,
OSLIBRARY_VARIANT and OSUSE_LIBRARY to

conjunction

with

OSLIBRARY_TYPE,

properly specify the
precompiled Salvo library you're linking to your project.
Library configurations might refer to, for example, whether the
library is configured to support delays and/or events.
Please see your compiler's Salvo Compiler Reference Manual and
Chapter 8 • Libraries for complete instructions on the use of
OSLIBRARY_CONFIG.
See Also

Salvo User Manual

OSUSE_LIBRARY.

Chapter 5 • Configuration

103

OSLIBRARY_GLOBALS: Specify Memory Type for Global
Salvo Objects in Precompiled Library
Name:
Purpose:

Allowed Values:
Default Value:
Action:
Related:
Enables:
Memory Required:
Notes

OSLIBRARY_GLOBALS
OSLIBRARY_CONFIG,
OSLIBRARY_VARIANT

OSLIBRARY_GLOBALS

To guarantee that an application's source
files are compiled using the same salvocfg.h as was used to create the specified precompiled library.
OSA …
not defined
Sets the configuration options inside salvolib.h to match those used to generate
the library specified.
OSLIBRARY_TYPE, OSLIBRARY_CONFIG,
OSLIBRARY_VARIANT, OSUSE_LIBRARY
–
n/a
is

used

in

OSLIBRARY_OPTION,
and OSUSE_LIBRARY to

conjunction

with

OSLIBRARY_TYPE,

properly specify the
precompiled Salvo library you're linking to your project.
Library globals might refer to, for example, whether the library
expects Salvo's global objects to be placed in internal or external
RAM.
Please see your compiler's Salvo Compiler Reference Manual and
Chapter 8 • Libraries for complete instructions on the use of
OSLIBRARY_GLOBALS.
See Also

104

OSUSE_LIBRARY.

Chapter 5 • Configuration

Salvo User Manual

OSLIBRARY_OPTION: Specify Precompiled Library
Option
Name:
Purpose:

Allowed Values:
Default Value:
Action:
Related:
Enables:
Memory Required:
Notes

OSLIBRARY_OPTION
OSLIBRARY_CONFIG,
OSLIBRARY_VARIANT

OSLIBRARY_OPTION

To guarantee that an application's source
files are compiled using the same salvocfg.h as was used to create the specified precompiled library.
OSA … or OSNONE
not defined
Sets the configuration options inside salvolib.h to match those used to generate
the library specified.
OSLIBRARY_CONFIG, OSLIBRARY_GLOBALS,
OSLIBRARY_VARIANT, OSUSE_LIBRARY
–
n/a
is

used

in

OSLIBRARY_GLOBALS,
and OSUSE_LIBRARY to

conjunction

with

OSLIBRARY_TYPE,

properly specify the
precompiled Salvo library you're linking to your project.
Library options might refer to, for example, whether the library
contains and/or supports embedded debugging information.
Please see your compiler's Salvo Compiler Reference Manual and
Chapter 8 • Libraries for complete instructions on the use of
OSLIBRARY_OPTION.
See Also

Salvo User Manual

OSUSE_LIBRARY.

Chapter 5 • Configuration

105

OSLIBRARY_TYPE: Specify Precompiled Library Type
Name:
Purpose:

Allowed Values:
Default Value:
Action:
Related:
Enables:
Memory Required:
Notes

OSLIBRARY_TYPE

To guarantee that an application's source
files are compiled using the same salvocfg.h as was used to create the specified precompiled library.
OSF or OSL
not defined
Sets the configuration options inside salvolib.h to match those used to generate
the library specified.
OSLIBRARY_CONFIG, OSLIBRARY_GLOBALS,
OSLIBRARY_VARIANT, OSUSE_LIBRARY
–
n/a

OSLIBRARY_TYPE is used in conjunction with OSLIBRARY_CONFIG,
OSLIBRARY_GLOBALS, OSLIBRARY_OPTION, OSLIBRARY_VARIANT
and OSUSE_LIBRARY to properly specify the precompiled Salvo li-

brary you're linking to your project.
Library types normally refer to whether the library is a freeware
library (OSF) or a standard library (OSL).
Please see your compiler's Salvo Compiler Reference Manual and
Chapter 8 • Libraries for complete instructions on the use of
OSLIBRARY_TYPE.
See Also

106

OSUSE_LIBRARY.

Chapter 5 • Configuration

Salvo User Manual

OSLIBRARY_VARIANT: Specify Precompiled Library
Variant
Name:
Purpose:

Allowed Values:
Default Value:
Action:
Related:
Enables:
Memory Required:
Notes

OSLIBRARY_VARIANT

To guarantee that an application's source
files are compiled using the same salvocfg.h as was used to create the specified precompiled library.
OSA … and OSNONE
not defined
Sets the configuration options inside salvolib.h to match those used to generate
the library specified.
OSLIBRARY_CONFIG, OSLIBRARY_GLOBALS,
OSLIBRARY_TYPE, OSUSE_LIBRARY
–
n/a

OSLIBRARY_VARIANT must be used in conjunction with
OSLIBRARY_CONFIG, OSLIBRARY_GLOBALS, OSLIBRARY_OPTION,
OSLIBRARY_TYPE and OSUSE_LIBRARY to properly specify the pre-

compiled Salvo library you're linking to your project.
Library variants might refer to, for example, whether the library
supports signaling events from within ISRs.
Not all libraries have variants. If a variant does not exist, set
to OSNONE.

OSLIBRARY_VARIANT

Please see your compiler's Salvo Compiler Reference Manual and
Chapter 8 • Libraries for complete instructions on the use of
OSLIBRARY_VARIANT.

See Also

Salvo User Manual

OSUSE_LIBRARY.

Chapter 5 • Configuration

107

OSMESSAGE_QUEUES: Set Maximum Number of
Message Queues
Name:
Purpose:

Allowed Values:
Default Value:
Action:
Related:

OSMESSAGE_QUEUES

To allocate memory at compile time for
message queue control blocks (mqcbs),
and to set an upper limit on the number of
supported message queues.
1 or greater.
1 if OSENABLE_MESSAGE_QUEUES is TRUE,
0 otherwise
Configures Salvo source code to support
the desired number of message queues.
OSENABLE_MESSAGE_QUEUES, OSLOC_MQCB,
OSLOC_MSGQ

Enables:
Memory Required:

Notes

message-queue-related services
When non-zero, requires a configurationdependent amount of RAM for each
mqcb.

This configuration parameter only allocates RAM for message
queue control blocks. It does not allocate RAM for the message
queues themselves – you must do that explicitly.
Message queues are numbered from 1 to OSMESSAGE_QUEUES.
Since message queue memory is allocated at compile time, the
mqcb memory will be used whether or not the message queue is
actually created via OSCreateMsgQ().
On a typical 8-bit processor, the amount of memory required by
each message queue control block is 6 bytes.

108

Chapter 5 • Configuration

Salvo User Manual

OSTARGET: Identify Target Processor
Name:
Purpose:
Allowed Values:
Default Value:
Action:
Related:
Enables:
Memory Required:
Notes

OSTARGET

To identify the processor you're using in
your Salvo application.
see salvo.h
NONE

Configures Salvo source code for the target processor.
OSCOMPILER

–
n/a

This configuration option is used within the Salvo source code
primarily to implement non-ANSI C directives like in-line assembly instructions and #pragma directives.
Nearly all of Salvo's supported compilers automatically override
your settings and define OSTARGET based on the command-line arguments passed to the compiler to identify the processor. Therefore
it is usually unnecessary to define OSTARGET in salvocfg.h.
If you are working with an as-yet-unsupported compiler, choose
OSUNDEF. See Chapter 10 • Porting for more information.

Salvo User Manual

Chapter 5 • Configuration

109

OSTASKS: Set Maximum Number of Tasks and Cyclic
Timers
Name:
Purpose:

Allowed Values:
Default Value:
Action:
Related:
Enables:
Memory Required:

Notes

OSTASKS

To allocate memory at compile time for
task control blocks (tcbs), and to set an
upper limit on the number of supported
tasks and cyclic timers.
1 or greater.
0

Configures Salvo source code to support
the desired number of tasks and cyclic
timers.
OSEVENTS

general and task-related services
When non-zero, requires a configurationdependent amount of RAM for each tcb,
and RAM for two tcb pointers.

Tasks and cyclic timers are numbered from 1 to OSTASKS. Each
task and each cyclic timer requires one tcb.
Since task and cyclic timer memory is allocated and fixed at compile time, the tcb memory will be used whether or not the task is
actually created via OSCreateTask() or the cyclic timer is created
via OSCreateCycTmr().
The amount of memory required by each task is dependent on several configuration options, and will range from a minimum of 4 to
a maximum 12 bytes per task.42

42

110

For the purposes of these size estimates, pointers to ROM memory are
assumed to be 16 bits, and pointers to RAM memory are assumed to be 8 bits.
This is the situation for the PIC16 and PIC17 family of processors.

Chapter 5 • Configuration

Salvo User Manual

OSUSE_LIBRARY: Use Precompiled Library
Name:
Purpose:
Allowed Values:

Default Value:
Action:
Related:

OSUSE_LIBRARY

To simplify linking to a precompiled Salvo
library.
FALSE: you are not linking to a precompiled Salvo library.
TRUE: you are linking to a precompiled
Salvo library.
FALSE

If TRUE, the proper configuration options
for the specified library will be used to
build the application.
OSLIBRARY_CONFIG, OSLIBRARY_GLOBALS,
OSLIBRARY_OPTION, OSLIBRARY_TYPE,
OSLIBRARY_VARIANT

Enables:
Memory Required:
Notes

–
n/a

Salvo's configuration options are compile-time options. When linking to a precompiled library of Salvo services, the settings for your
own application must match those originally used when the library
was generated. OSUSE_LIBRARY, and the related OSLIBRARY_XYZ
configuration options, take the guesswork out of creating a salvocfg.h header file for library builds.

Warning Failure to have matching configuration options may
lead to compile- and link-time errors that can be difficult to interpret. Because of the large number of configuration options and
their interrelationships, you must use OSUSE_LIBRARY and
OSLIBRARY_XYZ when linking to precompiled Salvo libraries.
Configuration options used to create precompiled Salvo libraries
differ from library to library. Please see your compiler's Salvo
Compiler Reference Manual and Chapter 8 • Libraries for complete instructions on the use of OSUSE_LIBRARY and
OSLIBRARY_XYZ.

Salvo User Manual

Chapter 5 • Configuration

111

Configuration Options for Source Code
Distributions
The configuration options described in this section can only be
used Salvo Pro and are listed in alphabetical order.
These configuration options affect the Salvo header (*.h) and
source (*.c) files.

112

Chapter 5 • Configuration

Salvo User Manual

OSBIG_SEMAPHORES: Use 16-bit Semaphores
Name:
Purpose:
Allowed Values:

Default Value:
Action:
Related:
Enables:
Memory Required:

Notes

OSBIG_SEMAPHORES

To select 8- or 16-bit counting semaphores.
FALSE: Counting semaphores range from 0
to 255.
TRUE: Counting semaphores range from 0
to 32,767.
FALSE

Changes the defined type OStypeSem from
8- to 16-bit unsigned integer.
–
-

When TRUE, requires an additional byte of
RAM for each ecb.

This configuration option can be used to minimize the size of ecbs.
Make OSBIG_SEMAPHORES TRUE only if your application requires
16-bit counting semaphores.
OSBIG_SEMAPHORES,

when TRUE, will usually enlarge the size of
ecbs by one byte on 8-bit targets.

Salvo User Manual

Chapter 5 • Configuration

113

OSBYTES_OF_COUNTS: Set Size of Counters
Name:
Purpose:

Allowed Values:
Default Value:
Action:

Related:
Enables:
Memory Required:

Notes

114

OSBYTES_OF_COUNTS

To allocate the RAM needed to hold the
maximum possible value for counters
used in Salvo, and to enable the code to
run the counters.
0, 1, 2, 4
0

If zero, disables all counters. If non-zero,
enables the counters OSctxSws and OSidleCtxSws, and sets the defined type
OStypeCount to be 8-, 16-, or 32-bit unsigned integer.
OSGATHER_STATISTICS

–
When non-zero, requires RAM for all enabled counters.

Salvo uses simple counters to keep track of context switches and
notable occurrences. Once a counter reaches its maximum value it
remains at that value.

Chapter 5 • Configuration

Salvo User Manual

OSBYTES_OF_DELAYS: Set Length of Delays
Name:
Purpose:

Allowed Values:
Default Value:
Action:

Related:
Enables:
Memory Required:

Notes

OSBYTES_OF_DELAYS

To enable delays and timeout services and
to allocate the RAM needed to hold the
maximum specified value (in system
ticks) for delays and timeouts.
0, 1, 2, 4
0

If zero, disables all delay and timeout services. If non-zero, enables the delay and
timeout services, and sets the defined type
OStypeDelay to be 8-, 16- or 32-bit unsigned integer.
OSTIMER_PRESCALAR
OS_Delay(), OSTimer()

When non-zero, requires 1, 2 or 4 additional bytes of RAM for each tcb and 1
tcb pointer in RAM.

Disabling delays and timeouts will reduce the size of the Salvo
code considerably. It will also reduce the size of the tcbs by 2 to 6
bytes per tcb.
in
conjunction
with
very long delays and timeouts while minimizing tcb memory requirements.

Use

OSTIMER_PRESCALAR
of
OSBYTES_OF_DELAYS can provide for

Salvo User Manual

Chapter 5 • Configuration

115

OSBYTES_OF_EVENT_FLAGS: Set Size of Event Flags
Name:
Purpose:
Allowed Values:
Default Value:
Action:
Related:
Enables:
Memory Required:

Notes

OSBYTES_OF_EVENT_FLAGS

To select 8-, 16- or 32-bit event flags.
1, 2, 4
1

Sets the defined type OStypeEFlag to 8-,
16- or 32-bit unsigned integer.
OSENABLE_EVENT_FLAGS

–
When event flags are enabled, requires 1, 2
or 4 bytes of RAM for each event flag control block (efcb) and additional ROM
(code) dependent on the target processor.

You can tailor the size of event flags in your Salvo application via
this configuration parameter.
Since each bit is independent of the others, it may be to your advantage to have a single, large event flag instead of multiple,
smaller ones. For example, the RAM requirements for two 8-bit
event flags will exceed those for a single 16-bit event flag since the
former requires two event control blocks, whereas the latter needs
only one.

116

Chapter 5 • Configuration

Salvo User Manual

OSBYTES_OF_TICKS: Set Maximum System Tick Count
Name:
Purpose:
Allowed Values:
Default Value:
Action:

Notes

OSBYTES_OF_TICKS

To enable elapsed time services and to allocate the RAM needed to hold the
maximum specified system ticks value.
0, 1, 2, 4
0

If zero, disables all elapsed time services.
If non-zero, enables the services , and sets
the defined type OStypeTick to be 8-, 16or 32-bit unsigned integer.

Related:
Enables:

OSTIMER_PRESCALAR

Memory Required:

When non-zero, requires RAM for the system tick counter.

OSGetTicks(), OSSetTicks(),
OSTimer()

Salvo uses a simple counter to keep track of system ticks. After it
reaches its maximum value the counter rolls over to 0.
Elapsed time services based on the system tick are obtained
through OSGetTicks() and OSSetTicks().
OSBYTES_OF_TICKS
OSBYTES_OF_DELAYS.

Salvo User Manual

must

be

Chapter 5 • Configuration

greater

or

equal

to

117

OSCALL_OSCREATEEVENT: Manage Interrupts when
Creating Events
Name:
Purpose:

Allowed Values:

Default Value:
Action:
Related:

OSCALL_OSCREATEEVENT

For use on target processors without software stacks in order to manage for interrupts when calling event-creating
services.
OSFROM_BACKGROUND: Your application
creates events only in mainline code.
OSFROM_FOREGROUND: Your application
creates events only within interrupts.
OSFROM_ANYWHERE Your application creates events both in mainline code and
within interrupts. You must explicitly
control interrupts around
OSCALL_OSCREATEEVENT (see below).
OSFROM_BACKGROUND

Configures the interrupt control for all
Salvo event-creating services.
OSCALL_OSSIGNALEVENT,
OSCALL_OSRETURNEVENT

Enables:
Memory Required:

Notes

–
Small variations in ROM depending on its
value.

is required only when using a compiler
that does not maintain function parameters and auto variables on a
software stack or in registers. Therefore this configuration parameter and all similar ones are only needed when using certain target
processors and compilers.

OSCALL_OSCREATEEVENT

Compilers that maintain function parameters and auto variables in
a dedicated area of RAM usually do so because a software stack
and stack pointers do not exist on the target processor. In order to
minimize RAM usage, these compilers43 overlay the parameter and
variable areas of multiple functions as long as the functions do not
occupy the same call graph. This is all done transparently – no
user involvement is required.
The issue is complicated by wanting to call Salvo services from
both mainline (background) and interrupt (foreground) code. In
this case, each service needs its own parameter and auto variable
43

118

E.g. the HI-TECH PICC and V8C compilers.

Chapter 5 • Configuration

Salvo User Manual

area separate from that of mainline-only services, and the user
must "wrap" each mainline service with calls to disable and then
re-enable interrupts44 in order to avoid data corruption. See the examples below.
The control of interrupts in each event-creating service like OSCreateBinSem() depends on where it is called in your application. In
Figure 27 interrupts will be disabled and re-enabled inside OSCreateBinSem(). This is referred to as protecting a critical region of
code, and is typical of RTOS services. In this situation,
OSCALL_OSCREATEEVENT must be set to OSFROM_BACKGROUND.
int main( void )
{
…
OSCreateBinSem(BINSEM1_P);
…
}
Figure 27: How to call OSCreateBinSem() when
OSCALL_OSCREATEEVENT is set to
OSFROM_BACKGROUND

In Figure 28 OSCreateBinSem() must not change the processor's
interrupt status, because re-enabling interrupts within an ISR can
cause unwanted nested interrupts. In this situation, set
OSCALL_OSCREATEEVENT to OSFROM_FOREGROUND.
interrupt myISR( void )
{
…
if (some_condition) {
OSCreateBinSem(BINSEM2_P);
}
…
}
Figure 28: How to call OSCreateBinSem() when
OSCALL_OSCREATEBINSEM is set to
OSFROM_FOREGROUND

In Figure 29, OSCreateBinSem() is called from the background as
well as the foreground. In this situation, OSCALL_OSCREATEEVENT
must be set to OSFROM_ANYWHERE and OSCreateBinSem() must be
preceded by OSProtect() and followed by OSUnprotect() wherever it's called in mainline (background) code.
int main( void )
{
…
44

Salvo User Manual

See "Interrupt Levels" in the HI-TECH PICC and PICC-18 User's Guide.

Chapter 5 • Configuration

119

OSProtect();
OSCreateBinSem(BINSEM1_P);
OSUnprotect();
…
OSProtect();
OSCreateBinSem(BINSEM2_P);
OSUnprotect();
…
}
interrupt myISR( void )
{
…
if (some_condition) {
OSCreateBinSem(BINSEM2_P);
}
…
}
Figure 29: How to call OSCreateBinSem() when
OSCALL_CREATEBINSEM is set to
OSFROM_ANYWHERE

Failing to set OSCALL_OSCREATEEVENT properly to reflect where
you are calling OSCreateBinSem() in your application may cause
unpredictable results, and may also result in compiler errors.
With

compilers
(e.g.
HI-TECH
PICC),
OSCALL_OSCREATEEVENT also automatically enables certain special
directives45 in the Salvo source code to ensure proper compilation.

45

120

some

E.g. #pragma interrupt_level 0, to allow a function to be called both
from mainline code and from an interrupt. In this situation a function has
"multiple call graphs."

Chapter 5 • Configuration

Salvo User Manual

OSCALL_OSGETPRIOTASK: Manage Interrupts when
Returning a Task's Priority
OSCALL_OSGETPRIOTASK manages how
OSGetPrio() and OSGetPrioTask().

interrupts are controlled in

See OSCALL_OSCREATEEVENT for more information on interrupt
control for services that can be called from the foreground.

OSCALL_OSGETSTATETASK: Manage Interrupts when
Returning a Task's State
OSCALL_OSGETSTATETASK manages how interrupts
in OSGetState() and OSGetStateTask().

are controlled

See OSCALL_OSCREATEEVENT for more information on interrupt
control for services that can be called from the foreground.

OSCALL_OSMSGQCOUNT: Manage Interrupts when
Returning Number of Messages in Message Queue
OSCALL_OSMSGQCOUNT
OSMsgQCount().

manages how interrupts are controlled in

See OSCALL_OSCREATEEVENT for more information on interrupt
control for services that can be called from the foreground.

OSCALL_OSMSGQEMPTY: Manage Interrupts when
Checking if Message Queue is Empty
OSCALL_OSMSGQEMPTY
OSMsgQEmpty().

manages how interrupts are controlled in

See OSCALL_OSCREATEEVENT for more information on interrupt
control for services that can be called from the foreground.

Salvo User Manual

Chapter 5 • Configuration

121

OSCALL_OSRETURNEVENT: Manage Interrupts when
Reading and/or Trying Events
manages how interrupts are controlled in
event-reading and event-trying services (e.g. OSReadEFlag() and
OSTrySem(), respectively).
OSCALL_OSRETURNEVENT

See OSCALL_OSCREATEEVENT for more information on interrupt
control for event-reading and event-trying services.

OSCALL_OSSIGNALEVENT: Manage Interrupts when
Signaling Events and Manipulating Event Flags
manages how interrupts are controlled in
event-signaling services (e.g. OSSignalMsg()), OSClrEFlag() and
OSSetEFlag().
OSCALL_OSSIGNALEVENT

See OSCALL_OSCREATEEVENT for more information on interrupt
control for event-signaling services.

OSCALL_OSSTARTTASK: Manage Interrupts when
Starting Tasks
OSCALL_OSSTARTTASK
OSStartTask().

manages how interrupts are controlled in

See OSCALL_OSCREATEEVENT for more information on interrupt
control for event-signaling services.

122

Chapter 5 • Configuration

Salvo User Manual

OSCLEAR_GLOBALS: Explicitly Clear all Global
Parameters
Name:
Purpose:
Allowed Values:
Default Value:
Action:
Related:
Enables:
Memory Required:

OSCLEAR_GLOBALS

To guarantee that all global variables used
by Salvo are explicitly initialized to zero.
FALSE, TRUE
TRUE

If TRUE, configures OSInit() to explicitly
fill all global variables (e.g. queue pointers, tcbs, ecbs, etc.) with 0.
OSENABLE_EVENTS,
OSENABLE_STACK_CHECKING
OSInitTcb() and OSInitEcb() for some
values of OSCOMPILER.
When TRUE, requires a small amount of

ROM.
Notes

All ANSI C compilers must initialize global variables to zero. OSInit() clears Salvo's variables by default. For those applications
where ROM memory is extremely precious, this configuration option can be disabled, and your application may shrink somewhat as
a result.

Caution If you disable this configuration option you must be
absolutely sure that your compiler explicitly initializes all of
Salvo's global variables to zero. Otherwise your application may
not work properly. Even if your compiler does zero all global variables, keep in mind that OSInit() will no longer (re-)zero the
global variables, and you will not be able to re-initialize Salvo via
a call to OSInit().

Salvo User Manual

Chapter 5 • Configuration

123

OSCLEAR_UNUSED_POINTERS: Reset Unused Tcb and
Ecb Pointers
Name:
Purpose:
Allowed Values:

OSCLEAR_UNUSED_POINTERS

Default Value:
Action:

FALSE

Related:
Enables:
Memory Required:

Notes

To aid in debugging Salvo activity.
Salvo makes no attempt to reset
no-longer used pointers in tcbs and ecbs.
TRUE: Salvo resets all unused tcb and ecb
pointers to NULL.
FALSE:

When TRUE, enables code to null unused
tcb and ecb pointers.
OSBYTES_OF_DELAYS, OSENABLE_TIMEOUTS,
–
When TRUE, requires a small amount of
ROM.

This configuration option is primarily of use to you if you are interested in viewing or debugging Salvo internals. It is much easier
to understand the status of the queues, tasks and events if the unused pointers are NULLed.
Enabling this configuration option will add a few instructions to
certain Salvo services.

124

Chapter 5 • Configuration

Salvo User Manual

OSCOLLECT_LOST_TICKS: Configure Timer System For
Maximum Versatility
Name:
Purpose:
Allowed Values:
Default Value:
Action:

Related:

OSCOLLECT_LOST_TICKS

To avoid delay- and timeout-related tick
errors due to poor task yielding behavior.
FALSE, TRUE
TRUE

Configures Salvo source code to log up to
a maximum number of ticks in the timer
for later delay and timeout processing in
the scheduler.
OSBYTES_OF_DELAYS, OSBYTES_OF_TICKS,
OSENABLE_TIMEOUTS

Enables:
Memory Required:

Notes

–
Target- and compiler-dependent. In most
cases, should reduce ROM requirements
slightly.

When OSCOLLECT_LOST_TICKS is FALSE, OSTimer() can log only
a single tick per call for eventual processing in the scheduler OSSched(). If, for example, an application has tasks that fail to yield
back to the scheduler within 2 system ticks, any tasks delayed or
waiting with a timeout during this period will appear to have their
delays or timeouts lengthened by the amount of time the poorlybehaved task(s) fails to yield to the scheduler.
When OSCOLLECT_LOST_TICKS is TRUE, OSTimer() can log up to
255 ticks for eventual processing in the scheduler. In the above example, the error in the delays or timeouts of simultaneously delayed or waiting tasks will be minimized.
OSCOLLECT_LOST_TICKS has no effect on the system's free-running
system tick counter OStimerTicks, which is accessed via OSGetTicks() and OSSetTicks().

Salvo User Manual

Chapter 5 • Configuration

125

OSCOMBINE_EVENT_SERVICES: Combine Common
Event Service Code
Name:
Purpose:
Allowed Values:

Default Value:
Action:
Related:
Enables:
Memory Required:

Notes

OSCOMBINE_EVENT_SERVICES

To minimize code size with multiple event
types enabled.
FALSE: All event services are implemented
as separate, independent functions.
TRUE: Event services use common code
where possible.
FALSE

Changes the structure of the Salvo source
code to produce minimum aggregate or
individual size of event services.
–
–
When TRUE, reduces ROM requirements
when event services for two or more
event types are used.

The services for creating, signaling and waiting events contain
common source code. When OSCOMBINE_EVENT_SERVICES is
TRUE, event services use that common code, e.g. OSCreateBinSem() and OSCreateMsgQ() use the same underlying function.
This means that the incremental increase in size of the object code
is relatively small when another event type is enabled via
OSENABLE_XYZ.
When OSCOMBINE_EVENT_SERVICES is FALSE, each event service
is implemented as a separate, independent function, and some code
is therefore duplicated. This is used when generating the Salvo
freeware libraries for maximum versatility.
When creating an application using two or more event types, the
aggregate size of all of the event services will be smaller when
OSCOMBINE_EVENT_SERVICES is TRUE.
The C language va_arg() and related functions are required when
OSCOMBINE_EVENT_SERVICES is TRUE.
Setting OSCOMBINE_EVENT_SERVICES to TRUE with HI-TECH
8051C and the small or medium memory models will prevent you
from calling any allowed event services (e.g. OSSignalMsg())
from an ISR. This restriction is lifted in the large model.

126

Chapter 5 • Configuration

Salvo User Manual

OSCTXSW_METHOD: Identify Context-Switching
Methodology in Use
Name:
Purpose:
Allowed Values:

Default Value:
Action:
Related:
Enables:
Memory Required:

Notes

OSCTXSW_METHOD

To configure the inner workings of the
Salvo context switcher.
OSRTNADDR_IS_PARAM: OSSaveRtnAddr()
is passed the task's return address as a parameter.
OSRTNADDR_IS_VAR: OSSaveRtnAddr()
reads the tasks's return address through a
global variable.
OSVIA_OSCTXSW: OSCtxSw() is used to return to the scheduler.
OSVIA_OSDISPATCH: OSCtxSw() is used in
conjunction with OSDispatch().
Defined for each compiler and target in
portXyz.h. If left undefined, default is
OSRTNADDR_IS_PARAM.
Configures Salvo source code for use with
the selected compiler and target processor.
OSRTNADDR_OFFSET

–
When set to OSRTNADDR_IS_VAR, requires
a small amount of RAM. ROM requirements vary.

This configuration option is used within the Salvo source code to
implement part of the context switcher OS_Yield().

Warning Unless you are porting Salvo to an as-yet-unsupported
compiler, do not override the value of OSCTXSW_METHOD in the
porting file salvoportXyz.h appropriate for your compiler. Unpredictable results will occur.
If you are working with an as-yet-unsupported compiler, refer to
the Salvo source code and Chapter 10 • Porting for further instructions.

Salvo User Manual

Chapter 5 • Configuration

127

OSCUSTOM_LIBRARY_CONFIG: Select Custom Library
Configuration File
Name:
Purpose:
Allowed Values:
Default Value:
Action:
Related:
Enables:
Memory Required:
Notes

OSCUSTOM_LIBRARY_CONFIG

To simply the generation and use of custom Salvo libraries.
0, 1 through 2046
0 (i.e. no custom library is selected)
Configures Salvo source code to include
the specified custom library configuration
file.
salvoclc1.h through salvoclc20.h
–
n/a

is used to ensure that the Salvo configuration for projects built with custom libraries matches the configuration that was in effect when the library was generated.
OSCUSTOM_LIBRARY_CONFIG

This configuration option need only be used when creating and using custom user libraries. There is no need to use
OSCUSTOM_LIBRARY_CONFIG when the freeware or standard libraries supplied in a Salvo distribution are used.
See Chapter 8 • Libraries for detailed information on using
OSCUSTOM_LIBRARY_CONFIG.

46

128

Values in excess of 20 will result in an error message when building a Salvo
library or application. Can be extended to larger values if need be – see
salvo/inc/salvolib.h.

Chapter 5 • Configuration

Salvo User Manual

OSDISABLE_ERROR_CHECKING: Disable Runtime Error
Checking
Name:
Purpose:
Allowed Values:
Default Value:
Action:
Related:
Enables:
Memory Required:

Notes

OSDISABLE_ERROR_CHECKING

To turn off runtime error checking.
FALSE: Error checking is enabled.
TRUE: Error checking is disabled.
FALSE

Disables certain error checking in some
Salvo user services.
–
–
When FALSE, requires ROM for errorchecking.

By default, Salvo performs run-time error checking on certain parameters passed to user services, like task priorities.
This error checking can be costly in terms of code space (ROM)
used. It can be disabled by setting OSDISABLE_ERROR_CHECKING to
TRUE. However, this is never recommended.

Caution Disabling error checking is strongly discouraged. It
should only be used as a last resort in an attempt to shrink code
size, with the attendant knowledge that any run-time error that
goes unchecked may result in unpredictable behavior.

Salvo User Manual

Chapter 5 • Configuration

129

OSDISABLE_FAST_SCHEDULING: Configure RoundRobin Scheduling
Name:
Purpose:
Allowed Values:
Default Value:
Action:
Related:
Enables:
Memory Required:

Notes

OSDISABLE_FAST_SCHEDULING

To alter execution sequence of tasks running in a round-robin manner.
FALSE: Fast scheduling is used.
TRUE: Fast scheduling is not used.
FALSE

Changes the way in which eligible tasks
returning to the scheduler are re-enqueued
into the eligible queue.
–
–
When TRUE, requires a small amount of
additional ROM.

By default, the Salvo scheduler immediately re-enqueues the current task upon its return to the scheduler if it is still eligible. This
has a side effect on round-robin scheduling that is best illustrated
by example.
If OSDISABLE_FAST_SCHEDULING is FALSE and the current task
signals an event upon which another task of equal priority is waiting, then the scheduler will run the signaling task again before the
waiting
task.47
On
the
other
hand,
if
OSDISABLE_FAST_SCHEDULING is TRUE in this situation, then the
scheduler will run the waiting task before the signaling task. In
other words, the round-robin sequence of task execution matches
the order in which the tasks are made eligible if
OSDISABLE_FAST_SCHEDULING is set to TRUE.
Setting OSDISABLE_FAST_SCHEDULING to TRUE will have a small
but significant negative impact on the context-switching speed of
your application.

47

130

This is indirectly related to the minimal stack depth required by
OSSignalXyz() services.

Chapter 5 • Configuration

Salvo User Manual

OSDISABLE_TASK_PRIORITIES: Force All Tasks to Same
Priority
Name:
Purpose:
Allowed Values:
Default Value:
Action:
Related:
Enables:
Memory Required:

Notes

Salvo User Manual

OSDISABLE_TASK_PRIORITIES

To reduce code (ROM) size when an application does not require prioritized
tasks.
FALSE: Tasks can have assigned priorities.
TRUE: All tasks have same priority (0).
FALSE

Removes priority-setting and prioritydependent code from Salvo services.
–
–
When FALSE, requires ROM for management of task priorities.

By default, Salvo schedules task execution based on task priorities.
Some savings in ROM size can be realized by disabling Salvo's
priority-specific code. When OSDISABLE_TASK_PRIORITIES is set
to TRUE, all tasks run at the same priority and round-robin.

Chapter 5 • Configuration

131

OSENABLE_BINARY_SEMAPHORES: Enable Support for
Binary Semaphores
Name:
Purpose:
Allowed Values:
Default Value:
Action:

Notes

132

OSENABLE_BINARY_SEMAPHORES

To control compilation of binary semaphore code via the preprocessor.
FALSE, TRUE
FALSE

If FALSE, binary semaphore services are
not available. If TRUE, OSCreateBinSem(), OSSignalBinSem()and
OS_WaitBinSem() are available.

Related:

OSENABLE_EVENT_FLAGS,
OSENABLE_MESSAGES,
OSENABLE_MESSAGE_QUEUES,
OSENABLE_SEMAPHORES, OSEVENTS

Enables:
Memory Required:

–
When TRUE, requires ROM for binary
semaphore services.

This configuration option is useful when controlling which parts of
Salvo are to be included in an application. If you are including or
linking to salvobinsem.c in your source code, you can control its
compilation solely via this configuration option in salvocfg.h.
This may be more convenient than, say, editing your source code
or modifying your project.

Chapter 5 • Configuration

Salvo User Manual

OSENABLE_BOUNDS_CHECKING: Enable Runtime
Pointer Bounds Checking
Name:
Purpose:
Allowed Values:
Default Value:
Action:

Related:

OSENABLE_BOUNDS_CHECKING

To check for out-of-range pointer arguments.
FALSE, TRUE
FALSE

If FALSE, pointer arguments are not
bounds-checked. If TRUE, some services
return an error if the pointer argument is
out-of-bounds.
OSDISABLE_ERROR_CHECKING,
OSSET_LIMITS

Enables:
Memory Required:

Notes

–
When TRUE, requires ROM for pointer
bounds checking.

The result of passing an incorrect pointer to a service is unpredictable. Some protection can be achieved by bounds-checking the
pointer to ensure that it is within a valid range of pointer values
appropriate for the service. This can be useful when debugging an
application that uses variables as placeholders for pointers instead
of constants.
The utility of runtime pointer bounds checking is limited. Since
valid pointers do not have successive addresses, the allowed range
includes not only the valid pointer values but also all the other values within that range. Therefore runtime pointer bounds checking
will only detect a small subset of invalid pointer arguments.
OSENABLE_BOUNDS_CHECKING
OSSET_LIMITS is set to TRUE.

Salvo User Manual

is overridden (i.e. set to TRUE) when

Chapter 5 • Configuration

133

OSENABLE_CYCLIC_TIMERS: Enable Cyclic Timers
Name:
Purpose:
Allowed Values:
Default Value:
Action:
Related:
Enables:
Memory Required:

Notes

134

OSENABLE_CYCLIC_TIMERS

To control compilation of cyclic timer
code via the preprocessor.
FALSE, TRUE
FALSE

If FALSE, cyclic timer services are not
available. If TRUE, cyclic timer services
are available.
–
–
When TRUE, requires ROM and in some
cases, tcb RAM.

This configuration option is useful when controlling which parts of
Salvo are to be included in an application. If you are including or
linking to any of the salvocyclicN.c source files in your source
code, you can control their compilation solely via this configuration option in salvocfg.h. This may be more convenient than,
say, editing your source code or modifying your project.

Chapter 5 • Configuration

Salvo User Manual

OSENABLE_EVENT_FLAGS: Enable Support for Event
Flags
Name:
Purpose:
Allowed Values:
Default Value:
Action:

Notes

OSENABLE_EVENT_FLAGS

To control compilation of event flag code
via the preprocessor.
FALSE, TRUE
FALSE

If FALSE, event flag services are not available. If TRUE, OSCreateEFlag(), OSClrEFlag(), OSSetEFlag()and
OS_WaitEFlag() are available.

Related:

OSBYTES_OF_EVENT_FLAGS,
OSENABLE_BINARY_SEMAPHORES,
OSENABLE_MESSAGES,
OSENABLE_MESSAGE_QUEUES,
OSENABLE_SEMAPHORES,
OSEVENTS, OSEVENT_FLAGS

Enables:
Memory Required:

–
When TRUE, requires ROM for event flag
services.

This configuration option is useful when controlling which parts of
Salvo are to be included in an application. If you are including or
linking to salvoeflag.c in your source code, you can control its
compilation solely via this configuration option in salvocfg.h.
This may be more convenient than, say, editing your source code
or modifying your project.
A value of 0 for OSEVENT_FLAGS automatically resets (overrides)
OSENABLE_EVENT_FLAGS to FALSE.

Salvo User Manual

Chapter 5 • Configuration

135

OSENABLE_EVENT_READING: Enable Support for Event
Reading
Name:
Purpose:
Allowed Values:
Default Value:
Action:

Related:

OSENABLE_EVENT_READING

To control compilation of event-reading
code via the preprocessor.
FALSE, TRUE
FALSE

If FALSE, event-reading services are not
available. If TRUE, OSReadBinSem(), OSReadEFlag(), OSReadMsg(), OSReadMsgQ()and OSReadSem() are
available.
OSCALL_OSRETURNEVENT,
OSENABLE_EVENT_TRYING

Enables:
Memory Required:

Notes

–
When TRUE, requires ROM for eventreading services.

If you use any event-reading services (e.g. OSReadMsg()), you
must set OSENABLE_EVENT_READING to TRUE in salvocfg.h. If
you do not use any event-reading services, leave it at is default
value of FALSE.
This configuration option is useful when controlling which parts of
Salvo are to be included in an application. If you are including
Salvo event source code in your project, you can keep unused
event-reading services out of your final object file solely via this
configuration option in salvocfg.h. This may be more convenient
than, say, editing your source code or modifying your project.
A value of TRUE for OSENABLE_EVENT_TRYING automatically sets
(overrides) OSENABLE_EVENT_READING to TRUE.

136

Chapter 5 • Configuration

Salvo User Manual

OSENABLE_EVENT_TRYING: Enable Support for Event
Trying
Name:
Purpose:
Allowed Values:
Default Value:
Action:

Related:

OSENABLE_EVENT_TRYING

To control compilation of event-trying
code via the preprocessor.
FALSE, TRUE
FALSE

If FALSE, event-trying services are not
available. If TRUE, OSTryBinSem(),
OSTryMsg(), OSTryMsgQ()and OSTrySem() are available.
OSCALL_OSRETURNEVENT,
OSENABLE_EVENT_READING

Enables:
Memory Required:

Notes

–
When TRUE, requires ROM for eventtrying services.

If you use any event-trying services (e.g. OSTrySem()), you must
set OSENABLE_EVENT_TRYING to TRUE in salvocfg.h. If you do
not use any event-trying services, leave it at is default value of
FALSE.
This configuration option is useful when controlling which parts of
Salvo are to be included in an application. If you are including
Salvo event source code in your project, you can keep unused
event-trying services out of your final object file solely via this
configuration option in salvocfg.h. This may be more convenient
than, say, editing your source code or modifying your project.
A value of TRUE for OSENABLE_EVENT_TRYING automatically sets
(overrides) OSENABLE_EVENT_READING to TRUE.

Salvo User Manual

Chapter 5 • Configuration

137

OSENABLE_FAST_SIGNALING: Enable Fast Event
Signaling
Name:
Purpose:
Allowed Values:
Default Value:
Action:

Related:
Enables:
Memory Required:

Notes

OSENABLE_FAST_SIGNALING

To increase the rate at which events can be
signaled.
FALSE, TRUE
FALSE

If FALSE, signaled events are processed48
when the waiting task runs.
If TRUE, signaled events are processed
when the event is signaled.
–
–
When TRUE, requires a moderate amount
of additional ROM, and extra tcb RAM
for messages and message queues.

With OSENABLE_FAST_SIGNALING set to FALSE, when an event is
signaled and a task was waiting the event, the event remains signaled until the waiting task runs. For example, when a binary
semaphore is signaled with TaskA() waiting, OSSignalBinSem()
will return OSERR_EVENT_FULL if called again before TaskA()
runs. When TaskA() runs, the binary semaphore is reset to 0, and a
subsequent call to OSSignalBinSem() will succeed. On the other
hand, if OSENABLE_FAST_SIGNALING is TRUE, the binary semaphore will immediately return to zero when TaskA() is made eligible by OSSignalBinSem(), and thereafter the binary semaphore
can be signaled again without error.
Fast signaling is useful when multiple tasks are waiting an event,
or the same event is signaled in rapid succession. In these situations, OSSignalXyz() will succeed until no tasks are waiting the
event and the event has been signaled.

48

138

E.g. a semaphore is decremented.

Chapter 5 • Configuration

Salvo User Manual

OSENABLE_IDLE_COUNTER: Track Scheduler Idling
Name:
Purpose:

OSENABLE_IDLE_COUNTER

Allowed Values:

Default Value:
Action:
Related:

To count how many times the scheduler
has been idle.
FALSE: Salvo does not keep track of how
often the scheduler OSSched() is idle.
TRUE: The OSidleCtxSw counter is incremented each time the scheduler is called
with no eligible tasks, i.e. the system is
idle.
FALSE

If TRUE, configures Salvo to track scheduler idling.
OSGATHER_STATISTICS,
OSENABLE_IDLING_HOOK

Enables:
Memory Required:

Notes

–
When TRUE, requires a small amount of
ROM, plus one byte of RAM.

If

OSGATHER_STATISTICS,
OSENABLE_COUNTS
and
OSENABLE_IDLE_COUNTER are all TRUE, and Salvo's idling hook
function is enabled via OSENABLE_IDLING_HOOK, then the OSidleCtxSws counter will be incremented each time the scheduler is

called and there are no tasks eligible to run. The percentage of time
your application is spending idle can be obtained by:
idle time = (OSidleCtxSws / OSctxSws) x 100

Salvo User Manual

Chapter 5 • Configuration

139

OSENABLE_IDLING_HOOK: Call a User Function when
Idling
Name:
Purpose:
Allowed Values:

Default Value:
Action:
Related:
Enables:
Memory Required:

Notes

140

OSENABLE_IDLING_HOOK

To provide a simple way of calling a user
function when idling.
FALSE: No function is called when idling.
TRUE: An external user hook function
named OSIdlingHook() is called when
idling.
FALSE

If TRUE, OSSched() calls OSIdlingHook() when no tasks are eligible to run.
–
–
When TRUE, requires a small amount of
ROM.

When you enable this both configuration, you must also define an
external function void OSIdlingHook(void). It will be called
automatically when your Salvo application is idling.

Chapter 5 • Configuration

Salvo User Manual

OSENABLE_MESSAGES: Enable Support for Messages
Name:
Purpose:
Allowed Values:
Default Value:
Action:

Related:

Enables:
Memory Required:

Notes

Salvo User Manual

OSENABLE_MESSAGES

To control compilation of message code
via the preprocessor.
FALSE, TRUE
FALSE

If FALSE, message services are not available. If TRUE, OSCreateMsg(), OSSignalMsg() and OS_WaitMsg() are
available.
OSENABLE_BINARY_SEMAPHORES,
OSENABLE_EVENT_FLAGS,
OSENABLE_MESSAGE_QUEUES,
OSENABLE_SEMAPHORES, OSEVENTS
–
When TRUE, requires ROM for message
services.

This configuration option is useful when controlling which parts of
Salvo are to be included in an application. If you are including or
linking to salvomsg.c in your source code, you can control its
compilation solely via this configuration option in salvocfg.h.
This may be more convenient than, say, editing your source code
or modifying your project.

Chapter 5 • Configuration

141

OSENABLE_MESSAGE_QUEUES: Enable Support for
Message Queues
Name:
Purpose:
Allowed Values:
Default Value:
Action:

Related:

OSENABLE_MESSAGE_QUEUES

To control compilation of message queue
code via the preprocessor.
FALSE, TRUE
FALSE

If FALSE, message services are not available. If TRUE, OSCreateMsgQ(), OSSignalMsgQ() and OS_WaitMsgQ() are
available.
OSENABLE_BINARY_SEMAPHORES,
OSENABLE_EVENT_FLAGS, OSENABLE_MESSAGES, OSENABLE_SEMAPHORES,
OSEVENTS,
OSMESSAGE_QUEUES

Enables:
Memory Required:

Notes

–
When TRUE, requires ROM for message
queue services.

This configuration option is useful when controlling which parts of
Salvo are to be included in an application. If you are including or
linking to salvomsgq.c in your source code, you can control its
compilation solely via this configuration option in salvocfg.h.
This may be more convenient than, say, editing your source code
or modifying your project.
A value of 0 for OSMESSAGE_QUEUES automatically resets (overrides) OSENABLE_MESSAGE_QUEUES to FALSE.

142

Chapter 5 • Configuration

Salvo User Manual

OSENABLE_OSSCHED_DISPATCH_HOOK: Call User
Function Inside Scheduler
Name:
Purpose:
Allowed Values:

Default Value:
Action:
Related:
Enables:
Memory Required:

Notes

OSENABLE_OSSCHED_DISPATCH_HOOK

To provide a simple way of calling a user
function from inside the scheduler.
FALSE: No user function is called from
OSSched().
TRUE: An external, user-supplied function
named OSSchedDispatchHook()is called
within OSSched() immediately prior to
the task being dispatched.
FALSE

If TRUE, you must define your own function to be called automatically each time
the scheduler runs.
–
–
When TRUE, requires ROM for user function and function call.

This configuration option is provided for advanced users who want
to call a function immediately prior to the most eligible task being
dispatched by the scheduler.
Interrupts are normally disabled when OSSchedEntryHook() is
called.

Salvo User Manual

Chapter 5 • Configuration

143

OSENABLE_OSSCHED_ENTRY_HOOK: Call User
Function Inside Scheduler
Name:
Purpose:
Allowed Values:

Default Value:
Action:
Related:
Enables:
Memory Required:

Notes

OSENABLE_OSSCHED_ENTRY_HOOK

To provide a simple way of calling a user
function from inside the scheduler.
FALSE: No user function is called from
OSSched().
TRUE: An external, user-supplied function
named OSSchedEntryHook()is called
within OSSched() immediately upon entry.
FALSE

If TRUE, you must define your own function to be called automatically each time
the scheduler runs.
–
–
When TRUE, requires ROM for user function and function call.

This configuration option is provided for advanced users who want
to call a function immediately upon entry into the scheduler.
Interrupts are normally enabled when OSSchedDispatchHook() is
called.

144

Chapter 5 • Configuration

Salvo User Manual

OSENABLE_OSSCHED_RETURN_HOOK: Call User
Function Inside Scheduler
Name:
Purpose:
Allowed Values:

Default Value:
Action:
Related:
Enables:
Memory Required:

Notes

OSENABLE_OSSCHED_RETURN_HOOK

To provide a simple way of calling a user
function from inside the scheduler.
FALSE: No user function is called from
OSSched().
TRUE: An external, user-supplied function
named OSSchedReturnHook()is called
within OSSched() immediately after the
dispatched task has returned to the scheduler.
FALSE

If TRUE, you must define your own function to be called automatically each time
the scheduler runs.
–
–
When TRUE, requires ROM for user function and function call.

This configuration option is provided for advanced users who want
to call a function immediately after the most eligible task has returned to the scheduler.
Interrupts are normally enabled when OSSchedReturnHook() is
called.

Salvo User Manual

Chapter 5 • Configuration

145

OSENABLE_SEMAPHORES: Enable Support for
Semaphores
Name:
Purpose:
Allowed Values:
Default Value:
Action:

Related:

Enables:
Memory Required:

Notes

146

OSENABLE_SEMAPHORES

To control compilation of semaphore code
via the preprocessor.
FALSE, TRUE
FALSE

If FALSE, semaphore services are not
available. If TRUE, OSCreateSem(), OSSignalSem() and OS_WaitSem() are
available.
OSENABLE_BINARY_SEMAPHORES,
OSENABLE_EVENT_FLAGS,
OSENABLE_MESSAGES,
OSENABLE_MESSAGE_QUEUES, OSEVENTS
–
When TRUE, requires ROM for semaphore
services.

This configuration option is useful when controlling which parts of
Salvo are to be included in an application. If you are including or
linking to salvosem.c in your source code, you can control its
compilation solely via this configuration option in salvocfg.h.
This may be more convenient than, say, editing your source code
or modifying your project.

Chapter 5 • Configuration

Salvo User Manual

OSENABLE_STACK_CHECKING: Monitor Call ... Return
Stack Depth
Name:
Purpose:
Allowed Values:

Default Value:
Action:

Related:
Enables:
Memory Required:

Notes

OSENABLE_STACK_CHECKING

To enable the user to discern the maximum
call ... return stack depth used by Salvo
services.
FALSE: Stack depth checking is not performed.
TRUE: Maximum and current stack depth is
recorded.
FALSE

If TRUE, enables code in each function to
monitor the current call ... return stack
depth and record a maximum call ... return stack depth if it has changed.
OSGATHER_STATISTICS, OSRpt()
–
When TRUE, requires a considerable
amount of ROM, plus two bytes of RAM.

Current and maximum stack depth are tracked to a maximum call
... return depth of 255.
Current stack depth is held in OSstkDepth. Maximum stack depth
is held in OSmaxStkDepth.
Stack depth is only calculated for call ... returns within Salvo code
and is not necessarily equal to the current hardware stack depth of
your processor. However, for most applications they will be the
same since OSSched() is usually called from main().

Salvo User Manual

Chapter 5 • Configuration

147

OSENABLE_TCBEXT0|1|2|3|4|5: Enable Tcb Extensions
Name:
Purpose:
Allowed Values:
Default Value:
Action:
Related:

Enables:
Memory Required:

Notes

OSENABLE_TCBEXT0|1|2|3|4|5

To add user-definable variables to a task's
control block.
FALSE: Named tcb extension is not enabled.
TRUE: Named tcb extension is enabled.
FALSE

If TRUE, creates a user-definable and accessible object of type OStypeTcbExt0|1|2|3|4|5 within each tcb.
OSLOC_TCB,
OSTYPE_TCBEXT0|1|2|3|4|5,
OScTcbExt0|1|2|3|4|5,
OStcbExt0|1|2|3|4|5
tcbExt0|1|2|3|4|5 fields
When TRUE, requires additional RAM per
tcb.

Salvo's standard tcb fields are reserved for the management of
tasks and events. In some instances it is useful to additional variables that are unique to the particular task. Salvo's tcb extensions
are ideal for this purpose.
The default type for a tcb extension is void * (i.e. a void pointer).
A tcb extension's type can be overridden to any type49 by using the
appropriate OSTYPE_TCBEXT0|1|2|3|4|5 configuration option.
Once enabled via OSENABLE_TCBEXT0|1|2|3|4|5, a tcb extension can
be accessed through the OScTcbExt0|1|2|3|4|5 or OStcbExt0|1|2|3|4|5 macros.
controls the storage type of tcb extensions. Tcb extensions are only initialized if / when OSInitTcb() is called, or by the
compiler's startup code. Any desired mix of the tcb extensions can
be enabled.
OSLOC_TCB

Consider the case of several identical tasks, all created from a single task function, which run concurrently. Each task is responsible
for one of several identical communications channels, each with its
own I/O and buffers. Enable a tcb extension of type pointer-to49

148

Including structures, etc.

Chapter 5 • Configuration

Salvo User Manual

struct,

and initialize it uniquely for each task. At runtime each
task runs independently of the others, managing its own communications channel, defined by the struct. Since only one task function need be defined, substantial savings in code size can be
realized.
The example in Listing 31 illustrates the use of a single, unsignedchar-sized tcb extension tcbExt1 that each of four identical tasks
uses as an index into an array of offsets in the 4KB buffer the tasks
share.
…
const unsigned offset[4] = { 3072,
2048,
1024,
0
};
void TaskBuff( void )
{
while (1) {
printf("Task %d's buffer ",
OStID(OScTcbP, OSTASKS));
printf("starts at %d\n", offset[OScTcbExt1]);
…
OS_Yield();
}
}
main()
{
OSInit();
OSCreateTask(TaskBuff,
OSCreateTask(TaskBuff,
OSCreateTask(TaskBuff,
OSCreateTask(TaskBuff,

OSTCBP(2),
OSTCBP(6),
OSTCBP(7),
OSTCBP(8),

OStcbExt1(OSTCBP(2))
OStcbExt1(OSTCBP(6))
OStcbExt1(OSTCBP(7))
OStcbExt1(OSTCBP(8))

0;
1;
2;
3;

=
=
=
=

1);
1);
1);
1);

for (i = 0; i < 4; i++) {
OSSched();
}
}
Listing 31: Tcb Extension Example

Each time TaskBuff() runs, it can obtain its offset into the 4KB
buffer through OStcbExt1 for the current task, namely, itself. For
this example, OSENABLE_TCBEXT1 was set to TRUE and

Salvo User Manual

Chapter 5 • Configuration

149

OSTYPE_TCBEXT1 was set to unsigned char in the project's salvocfg.h. The resulting output is shown in Figure 30.

Figure 30: Tcb Extension Example Program Output

Tcb extensions can be used for a variety of purposes, including
•
•
•

50

150

• Passing information via a pointer to a task at
startup or during runtime.50
• Avoiding the use of task-specific global
variables accessed indirectly via OStID().
• Embedding objects of any type in a task's tcb.

This is useful because Salvo tasks must be declared as void Task ( void
), i.e. without any parameters.

Chapter 5 • Configuration

Salvo User Manual

OSENABLE_TIMEOUTS: Enable Support for Timeouts
Name:
Purpose:
Allowed Values:
Default Value:
Action:
Related:
Enables:
Memory Required:

Notes

OSENABLE_TIMEOUTS

To be able to specify an optional timeout
when waiting for an event.
FALSE: Timeouts cannot be specified.
TRUE: Timeouts can be specified.
FALSE

If TRUE, enables the passing of an extra
parameter to specify a timeout when waiting for an event..
–
OSTimedOut()

When TRUE, requires a considerable
amount of ROM, plus an additional byte
of RAM per tcb.

By specifying a timeout when waiting for an event, the waiting
task can continue if the event does not occur within the specified
time period. Use OSTimedOut() to detect if a timeout occurred.
If timeouts are enabled, you can use the defined symbol
OSNO_TIMEOUT for those calls that do not require a timeout.
See Chapter 6 • Frequently Asked Questions (FAQ) for more information on using timeouts.

Salvo User Manual

Chapter 5 • Configuration

151

OSGATHER_STATISTICS: Collect Run-time Statistics
Name:
Purpose:
Allowed Values:
Default Value:
Action:

Related:

OSGATHER_STATISTICS

To collect run-time statistics from your
application.
FALSE: Statistics are not collected.
TRUE: A variety of statistics are collected.
FALSE

If TRUE, enables Salvo code to collect runtime statistics from your application on
the number of errors, warnings, timeouts,
context switches and calls to the idle
function.
OSBYTES_OF_COUNTS,
OSENABLE_STACK_CHECKING

Enables:
Memory Required:

Notes

–
When TRUE, requires a small amount of
ROM, plus RAM for counters.

The numbers of errors, warnings and timeouts are tracked to a
maximum value of 255.
The maximum number of any counter is dependent on the value of
OSBYTES_OF_COUNTS. If OSBYTES_OF_COUNTS is not defined or is
defined to be 0, it will be redefined to 1.
Which statistics are collected is highly dependent on the related
configuration options listed above.
If enabled via OSLOGGING, error and warning logging will occur
regardless of the value of OSGATHER_STATISTICS.

152

Chapter 5 • Configuration

Salvo User Manual

OSINTERRUPT_LEVEL: Specify Interrupt Level for
Interrupt-callable Services
Name:
Purpose:

Allowed Values:
Default Value:
Action:
Related:
Enables:
Memory Required:
Notes

OSINTERRUPT_LEVEL

To specify the interrupt level used in the
Salvo source code. For use with these
compilers:
HI-TECH PICC and PICC-Lite
HI-TECH PICC-18
HI-TECH V8C
0-7 (depends on compiler)
0
OSCALL_OSXYZ

–
–

Some compilers support an interrupt level feature. With
OSINTERRUPT_LEVEL you can specify which level is used by Salvo
services called from the foreground.
All affected Salvo services use the same interrupt level.

Salvo User Manual

Chapter 5 • Configuration

153

OSLOC_ALL: Storage Type for All Salvo Objects
Name:
Purpose:
Allowed Values:
Default Value:
Action:
Related:

Enables:
Memory Required:
Notes

OSLOC_ALL

To place Salvo objects anywhere in RAM.
See Table 1.
OSLOC_DEFAULT (in portxyz.h).
Set the memory storage type for all of
Salvo's objects that aren't overridden by
OSLOC_XYZ.
OSLOC_ALL, OSLOC_COUNT, OSLOC_CTCB,
OSLOC_DEPTH, OSLOC_ECB, OSLOC_ERR,
OSLOC_LOGMSG, OSLOC_MQCB,
OSLOC_MSGQ, OSLOC_PS, OSLOC_SIGQ,
OSLOC_TCB, OSLOC_TICK
–
n/a

Many compilers support a variety of storage types (also called
memory types) for static objects. Depending on the target processor's architecture, it may be advantageous or necessary to place
Salvo's variables into RAM spaces other than the default provided
by the compiler.
when used alone, will locate all of Salvo's objects in
the specified RAM space. OSLOC_ALL overrides all other undefined
OSLOC_XYZ configuration parameters. To place all of Salvo's variables in RAM Bank 2 with the HI-TECH PICC compiler, use:
OSLOC_ALL,

#define OSLOC_ALL bank2

in salvocfg.h. To place the event control blocks (ecbs) in data
RAM, and everything else in external RAM with the Keil Cx51
compiler, use:
#define OSLOC_ALL xdata
#define OSLOC_ECB data

The storage types for all of Salvo's objects are set via OSLOC_ALL
and the remaining OSLOC_XYZ (see below) configuration parameters. Do not attempt to set storage types in any other manner –
compile- and / or run-time errors are certain to result.
Table 1 lists the allowable storage types / type qualifiers for Salvo
objects for each supported compiler (where applicable). Those on
separate lines can be combined, usually in any order.

154

Chapter 5 • Configuration

Salvo User Manual

compiler
HI-TECH PICC

storage types / type qualifiers
bank1, bank2, bank3
persistent

HI-TECH PICC-18
HI-TECH V8C
Keil Cx51

near
persistent
persistent
data, idata, far, xdata

not supported – use
Microchip MPLAB-C18

OSMPLAB_C18_LOC_ALL_NEAR

in-

stead
Table 1: Allowable Storage Types / Type Qualifiers for
Salvo Objects

See Also

Salvo User Manual

OSLOC_XYZ, Chapter 11 • Tips, Tricks and Troubleshooting

Chapter 5 • Configuration

155

OSLOC_COUNT: Storage Type for Counters
Name:
Purpose:
Allowed Values:
Default Value:
Action:
Related:
Enables:
Memory Required:
Notes

OSLOC_COUNT

To place Salvo counters anywhere in
RAM.
See Table 1.
OSLOC_DEFAULT (in portxyz.h).
Set storage type for Salvo counters.
OSLOC_ALL

–
n/a

will locate the context switch and idle context switch
counters in the specified RAM area. Memory is allocated for these
counters only when statistics are gathered.
OSLOC_COUNT

To explicitly specify RAM Bank 0 with the HI-TECH PICC compiler, use:
#define OSLOC_COUNT

in salvocfg.h.
As with all OSLOC_XYZ configuration options, multiple type qualifiers can be used with OSLOC_COUNT. For example, to prevent HITECH PICC start-up code from re-initializing Salvo's counters in
RAM bank 2, use:
#define OSLOC_COUNT bank2 persistent

See Also

156

Chapter 11 • Tips, Tricks and Troubleshooting

Chapter 5 • Configuration

Salvo User Manual

OSLOC_CTCB: Storage Type for Current Task Control
Block Pointer
will locate the current task control block pointer in the
specified RAM area. This pointer is used by OSSched().
OSLOC_CTCB

See OSLOC_COUNT for more information on setting storage types for
Salvo objects.

OSLOC_DEPTH: Storage Type for Stack Depth Counters
will locate the 8-bit call ... return stack depth and
maximum stack depth counters in the specified RAM area. Memory is allocated for these counters only when stack depth checking
is enabled.
OSLOC_DEPTH

See OSLOC_COUNT for more information on setting storage types for
Salvo objects.

See Also

OSENABLE_STACK_CHECKING

OSLOC_ECB: Storage Type for Event Control Blocks and
Queue Pointers
will locate the event control blocks, the eligible queue
pointer and the delay queue pointer in the specified RAM area.
Memory is allocated for ecbs only when events are enabled. Memory is allocated for the delay queue pointer only when delays
and/or timeouts are enabled.
OSLOC_ECB

See OSLOC_COUNT for more information on setting storage types for
Salvo objects.

See Also

OSEVENTS

OSLOC_EFCB: Storage Type for Event Flag Control
Blocks
OSLOC_EFCB will locate the event flag control blocks – declared to
be of type OSgltypeEfcb by the user – in the specified RAM area.

Salvo User Manual

Chapter 5 • Configuration

157

See OSLOC_COUNT for more information on setting storage types for
Salvo objects.

OSLOC_ERR: Storage Type for Error Counters
will locate the 8-bit error, warning and timeout counters in the specified RAM area. Memory is allocated for these
counters only when logging is enabled.
OSLOC_ERR

See OSLOC_COUNT for more information on setting storage types for
Salvo objects.

See Also

OSENABLE_TIMEOUTS, OSGATHER_STATISTICS, OS_LOGGING

OSLOC_GLSTAT: Storage Type for Global Status Bits
will locate Salvo's global status bits in the specified
RAM area. Memory is allocated for these bits whenever time functions are enabled.
OSLOC_GLSTAT

See OSLOC_COUNT for more information on setting storage types for
Salvo objects.

OSLOC_LOGMSG: Storage Type for Log Message String
will locate the character buffer used to hold log
messages in the specified RAM area. This buffer is needed to create error, warning and descriptive informational messages.
OSLOC_LOGMSG

See OSLOC_COUNT for more information on setting storage types for
Salvo objects.

See Also

OS_LOGGING, OSLOG_MESSAGES

OSLOC_LOST_TICK: Storage Type for Lost Ticks
will locate the character buffer used to hold lost
ticks in the specified RAM area. This buffer is used to avoid timing errors when the scheduler is not called rapidly enough.
OSLOC_LOST_TICK

158

Chapter 5 • Configuration

Salvo User Manual

See OSLOC_COUNT for more information on setting storage types for
Salvo objects.

See Also

OS_LOGGING, OSLOG_MESSAGES

OSLOC_MQCB: Storage Type for Message Queue Control
Blocks
will locate the message queue control blocks (mqcbs)
in the specified RAM area. Each message queue has an mqcb associated with it – however, message queues and mqcbs need not be
in the same bank.
OSLOC_MQCB

See OSLOC_COUNT for more information on setting storage types for
Salvo objects.

OSLOC_MSGQ: Storage Type for Message Queues
tells Salvo that the message queue buffers are located
in the specified RAM area. By using the predefined Salvo qualified
type OSgltypeMsgQP when declaring each buffer it will be automatically placed in the desired RAM bank.
OSLOC_MSGQ

See OSLOC_COUNT for more information on setting storage types for
Salvo objects.

See Also

OSMESSAGE_QUEUES

OSLOC_PS: Storage Type for Timer Prescalar
will locate the timer prescalar (used by OSTimer()) in
the specified RAM area.
OSLOC_PS

See OSLOC_COUNT for more information on setting storage types for
Salvo objects.

See Also

Salvo User Manual

OSENABLE_PRESCALAR

Chapter 5 • Configuration

159

OSLOC_TCB: Storage Type for Task Control Blocks
OSLOC_TCB

will locate the task control blocks in the specified

RAM area.
See OSLOC_COUNT for more information on setting storage types for
Salvo objects.

OSLOC_SIGQ: Storage Type for Signaled Events Queue
Pointers
will locate the signaled events queue pointers in the
specified RAM area. Memory is allocated for this counter only
when events are enabled.
OSLOC_SIGQ

See OSLOC_COUNT for more information on setting storage types for
Salvo objects.

OSLOC_TICK: Storage Type for System Tick Counter
will locate the system tick counter in the specified
RAM area. Memory is allocated for this counter only when ticks
are enabled.
OSLOC_TICK

See OSLOC_COUNT for more information on setting storage types for
Salvo objects.

See Also

160

OSBYTES_OF_TICKS

Chapter 5 • Configuration

Salvo User Manual

OSLOGGING: Log Runtime Errors and Warnings
Name:
Purpose:
Allowed Values:
Default Value:
Action:
Related:
Enables:
Memory Required:

Notes

OSLOGGING

To log runtime errors and warnings.
FALSE: Errors and warnings are not
logged.
TRUE: Errors and warnings are logged.
FALSE

Configures Salvo functions to log all errors
and warnings that occur when during
execution.
OSLOG_MESSAGES, OSRpt()
–
When TRUE, requires a considerable
amount of ROM, plus RAM for the error
and warning counters.

Most Salvo functions return an 8-bit error code. Additionally,
Salvo can track run-time errors and warnings through the dedicated
8-bit counters OSerrs and OSwarns.
OSRpt()
is TRUE.

will display the error and warning counters if OSLOGGING

The value of OSLOGGING has no effect on the return codes for Salvo
user services.
OSLOGGING

See Also

Salvo User Manual

is not affected by OSGATHER_STATISTICS.

OSRpt()

Chapter 5 • Configuration

161

OSLOG_MESSAGES: Configure Runtime Logging
Messages
Name:
Purpose:

OSLOG_MESSAGES

Allowed Values:

Notes

To aide in debugging your Salvo application.
OSLOG_NONE: No messages are generated.
OSLOG_ERRORS: Only error messages are
generated.
OSLOG_WARNINGS: Error and warning messages are generated.
OSLOG_ALL: Error, warning and informational messages are generated.

Default Value:
Action:

OSLOG_NONE

Related:
Enables:
Memory Required:

OSLOGGING

Configures Salvo functions to log in a
user-understandable way all errors, warnings and/or general information that occurs when each function executes.
–
When TRUE, requires a considerable
amount of ROM, plus RAM for an 80character buffer, OSlogMsg[].

Most Salvo functions return an 8-bit error code. If your application
has the ability to printf() to a console, Salvo can be configured
via this configuration option to report on errors, warnings and/or
general information with descriptive messages. If an error, warning
or general event occurs, a descriptive message with the name of the
corresponding Salvo function is output via printf(). This can be
useful when debugging your application, when modifying the
source code or when learning to use Salvo.
Applications that do not have a reentrant printf() may have
problems when reporting any errors. In these cases, set OSLOG_MESSAGES to OSLOG_NONE.
Stack depth for printf() is not tracked by Salvo – your application may have problems if there is insufficient stack depth beyond
that used by Salvo.
OSLOGGING

162

must be TRUE to use OSLOG_MESSAGES.

Chapter 5 • Configuration

Salvo User Manual

The value of OSLOG_MESSAGES has no effect on the return codes for
Salvo user services.

Salvo User Manual

Chapter 5 • Configuration

163

OS_MESSAGE_TYPE: Configure Message Pointers
Name:
Purpose:
Allowed Values:
Default Value:
Action:
Related:
Enables:
Memory Required:
Notes

OS_MESSAGE_TYPE

Enable message pointers to access any area
in memory. Compiler-dependent.
Any pointer type supported by the compiler.
void

Redefines the defined type OStypeMsg.
OSCOMPILER
-

Dependent on definition

Salvo's message pointers (of type OStypeMsgP), used by messages
and message queues, are normally defined as void pointers, i.e.
void *. A void pointer can usually point to anywhere in RAM or
ROM. This is useful, for instance, if some of your message pointers point to constant strings in ROM as well as static variables (in
RAM).
Some supported compilers require an alternate definition for message pointers in order to point to ROM and RAM together, or to
external memory, etc. By redefining OS_MESSAGE_TYPE, message
pointers can point to the memory of interest.
For example, for Salvo's message pointers to access both ROM and
RAM with the HI-TECH PICC compiler, OS_MESSAGE_TYPE must
be defined as const instead of void, because PICC's const *
pointers can access both ROM and RAM, whereas its void *
pointers can only access RAM.
Changing OS_MESSAGE_TYPE may affect the size of ecbs.

164

Chapter 5 • Configuration

Salvo User Manual

OSMPLAB_C18_LOC_ALL_NEAR: Locate all Salvo
Objects in Access Bank (MPLAB-C18 Only)
Name:
Purpose:
Allowed Values:

Default Value:
Action:
Related:
Enables:
Memory Required:

Notes

Salvo User Manual

OSMPLAB_C18_LOC_ALL_NEAR

To improve application performance by
placing Salvo's global objects in access
RAM.
FALSE: Salvo's global objects are placed in
banked RAM.
TRUE: Salvo's global objects are placed in
access RAM.
FALSE

Declares all of Salvo's global objects to be
of type near.
–
–
When TRUE, should reduce ROM requirements.

Salvo's OSLOC_XYZ configuration cannot be used with MPLABC18. Use OSMPLAB_C18_LOC_ALL_NEAR instead to place all of
Salvo's global objects in access RAM for improved run-time
performance.

Chapter 5 • Configuration

165

OSOPTIMIZE_FOR_SPEED: Optimize for Code Size or
Speed
Name:
Purpose:
Allowed Values:

Default Value:
Action:
Related:
Enables:
Memory Required:

Notes

OSOPTIMIZE_FOR_SPEED

To allow you to optimize your application
for minimum Salvo code size or maximum speed.
FALSE: Salvo source code will compile for
minimum size with existing configuration
options.
TRUE: Salvo source code will compile for
maximum speed with existing configuration options.
FALSE

Takes advantage of certain opportunities to
increase the speed of the Salvo code.
OSENABLE_DELAYS

–
When TRUE, requires small amounts of
ROM and RAM.

Opportunities exist in the Salvo source code to improve execution
speed at the cost of some additional lines of code or bytes of RAM.
This configuration option enables you to take advantage of these
opportunities.
This configuration option does not override other parameters that
may also have an effect on code size.
This configuration option is completely independent of any optimizations your compiler may perform. The interaction between it
and your compiler is of course unpredictable.
The interplay between execution speed and memory requirements
is complex and is most likely to be unique to each application. For
example, configuring Salvo for maximum speed may in some
cases both increase speed and shrink ROM size, at the expense of
some memory RAM.

166

Chapter 5 • Configuration

Salvo User Manual

OSPIC18_INTERRUPT_MASK: Configure PIC18 Interrupt
Mode
Name:
Purpose:
Allowed Values:
Default Value:
Action:
Related:
Enables:
Memory Required:
Notes

OSPIC18_INTERRUPT_MASK

To allow you to control which PIC18
PICmicro interrupts are disabled during
Salvo's critical sections.
0xC0, 0x80, 0x40, 0x00
0xC0 (all interrupts are disabled during
critical sections).
Defines the interrupt-clearing mask that
will be used in Salvo services that contain
critical regions of code.
–
–
–

is currently supported for use with the
IAR PIC18 and Microchip MPLAB-C18 compilers.
OSPIC18_INTERRUPT_MASK

Microchip PIC18 PICmicro MCUs support two distinct interrupt
modes of operation: one with two levels of interrupt priorities
(IPEN is 1), and one that is compatible with Microchip's mid-range
PICmicro devices (IPEN is 0). Depending on how your application
calls Salvo services, it may be to your advantage to change
OSPIC18_INTERRUPT_MASK to minimize interrupt latency.
When OSPIC18_INTERRUPT_MASK is set to 0xC0, all interrupts
(global / high-priority and peripheral / low-priority) are disabled
during critical regions. Therefore a value of 0xC0 is compatible
with both priority schemes and any method of calling Salvo services.
When OSPIC18_INTERRUPT_MASK is set to 0x80, only global /
high-priority interrupts are disabled during critical regions. Therefore a value of 0x80 should only be used in two cases: 1) in compatibility mode, and 2) in priority mode if Salvo services that can
be called from the foreground / ISR level are called exclusively
from high-level interrupts.
When OSPIC18_INTERRUPT_MASK is set to 0x40, only peripheral /
low-priority interrupts are disabled during critical regions. Therefore a value of 0x40 should only be used in priority mode if Salvo
services that can be called from the foreground / ISR level are

Salvo User Manual

Chapter 5 • Configuration

167

called exclusively from low-level interrupts. A value of 0x40 must
not be used in compatibility mode.
A value of 0x00 is permitted. However, it must only be used on
applications that do not use interrupts.
Failure to use the correct value of OSPIC18_INTERRUPT_MASK for
your application will lead to unpredictable runtime results.
See Microchip's PIC18 PICmicro databooks and your PIC18 compiler's Salvo Compiler Reference Manual for more information.

168

Chapter 5 • Configuration

Salvo User Manual

OSRPT_HIDE_INVALID_POINTERS: OSRpt() Won't
Display Invalid Pointers
Name:
Purpose:
Allowed Values:

Default Value:
Action:
Related:

OSRPT_HIDE_INVALID_POINTERS

To make the output of OSRpt() more legible.
FALSE: All tcb and ecb pointer values will
be displayed, regardless of whether or not
they are valid.
TRUE: Only those pointers which are valid
are shown in the monitor.
TRUE

Configures OSRpt() to show or hide invalid pointers.
OSRPT_SHOW_ONLY_ACTIVE,
OSRPT_SHOW_TOTAL_DELAY

Enables:
Memory Required:

Notes

–
When TRUE, requires a small amount of
ROM.

In some cases, the pointer fields of tcbs and ecbs are meaningless.
For example, if a task has been destroyed, the pointers in its tcb are
invalid. By making OSRPT_HIDE_INVALID_POINTERS TRUE,
OSRpt()'s output is simplified by removing unnecessary information. Invalid pointers are displayed as "n/a".
See Chapter 7 • Reference for more information on OSRpt().

Salvo User Manual

Chapter 5 • Configuration

169

OSRPT_SHOW_ONLY_ACTIVE: OSRpt() Displays Only
Active Task and Event Data
Name:
Purpose:
Allowed Values:

Default Value:
Action:
Related:

OSRPT_SHOW_ONLY_ACTIVE

To remove unnecessary information from
OSRpt()'s output.
FALSE: Show the contents of each tcb and
ecb.
TRUE: Show only the contents of each active tcb and ecb.
TRUE

Configures OSRpt() to show only tasks
which are not destroyed and events which
have already been created.
OSRPT_HIDE_INVALID_POINTERS,
OSRPT_SHOW_TOTAL_DELAY

Enables:
Memory Required:

Notes

–
When TRUE, requires a small amount of
ROM.

By showing neither the tcb contents of tasks in the destroyed state,
nor the ecb contents of events which have not yet been created,
OSRpt()'s output is simplified. However, if you wish to have all
the tasks and events displayed by OSRpt(), set this configuration
option to FALSE.
See Chapter 7 • Reference for more information on OSRpt().

170

Chapter 5 • Configuration

Salvo User Manual

OSRPT_SHOW_TOTAL_DELAY: OSRpt() Shows the Total
Delay in the Delay Queue
Name:
Purpose:
Allowed Values:

Default Value:
Action:
Related:

OSRPT_SHOW_TOTAL_DELAY

To aid in computing total delay times
when viewing OSRpt()'s output.
FALSE: Only individual task delay fields
are shown.
TRUE: The total (cumulative) delay for all
the tasks in the delay queue is computed
and shown.
TRUE

Configures OSRpt() to compute and display the total delay of all delayed tasks.
OSRPT_HIDE_INVALID_POINTERS,
OSRPT_SHOW_ONLY_ACTIVE

Enables:
Memory Required:

Notes

–
When TRUE, requires a small amount of
ROM.

Task delays are stored in the delay queue in an incremental (and
not absolute) scheme. When debugging your application it may be
useful to be able to see the total delay of all tasks in the delay
queue.
See Chapter 7 • Reference for more information on OSRpt().

Salvo User Manual

Chapter 5 • Configuration

171

OSRTNADDR_OFFSET: Offset (in bytes) for ContextSwitching Saved Return Address
Name:
Purpose:
Allowed Values:
Default Value:

Action:
Related:
Enables:
Memory Required:
Notes

OSRTNADDR_OFFSET

To configure the inner workings of the
Salvo context switcher.
Any literal.
Defined for each compiler and target in
portXyz.h whenever OSCTXSW_METHOD
is OSRTNADDR_IS_VAR. If left undefined,
default is 0.
Configures Salvo source code for use with
the selected compiler and target processor.
OSCTXSW_METHOD

–
n/a

This configuration option is used within the Salvo source code to
implement part of the context switcher OS_Yield().

Warning Unless you are porting Salvo to an as-yet-unsupported
compiler, do not override the value of OSCTXSW_METHOD in the
porting file salvoportXyz.h appropriate for your compiler. Unpredictable results will occur.
If you are working with an as-yet-unsupported compiler, refer to
the Salvo source code and Chapter 10 • Porting for further instructions.

172

Chapter 5 • Configuration

Salvo User Manual

OSSCHED_RETURN_LABEL(): Define Label within
OSSched()
Name:
Purpose:
Allowed Values:
Default Value:
Action:
Related:
Enables:
Memory Required:
Notes

OSSCHED_RETURN_LABEL

To define a globally visible label for certain Salvo context switchers.
Undefined, or defined to be the instruction(s) required to create a globally visible label.
Defined but valueless.
Creates a globally visible label for use by
the goto statement.
–
–
–

Salvo context switchers for certain compilers and/or target processors may be implemented with a goto-based approach rather than
with a call-based approach. For those circumstances, a globally
visible label within the scheduler OSSched() is required. By declaring a label via this configuration parameter, a context switcher
will be able to "return" from a task to the appropriate part of the
scheduler.
The preferred name for the label is OSSchedRtn.
For the Microchip 12-bit PICmicros (e.g. PIC16C57), which have
only a 2-level hardware call…return stack, the following is used
with the HI-TECH PICC compiler:
#define OSSCHED_RETURN_LABEL() { \
asm("global _OSSchedRtn"); \
asm("_OSSchedRtn:"); \
}

This creates a globally visible label OSSchedRtn that can be
jumped to from other parts of the program.
See the various portxyz.h compiler- and target-specific porting
files for more information.

Salvo User Manual

Chapter 5 • Configuration

173

OSSET_LIMITS: Limit Number of Runtime Salvo Objects
Name:
Purpose:
Allowed Values:

Default Value:
Action:
Related:
Enables:
Memory Required:
Notes

OSSET_LIMITS

To limit the number of permissible Salvo
objects when using the freeware libraries.
FALSE: The numbers of Salvo objects are
limited only by their definitions in salvomem.c.
TRUE: Salvo services reject operations on
Salvo objects that are outside the limits
set by the configuration parameters.
FALSE

Adds run-time bounds-checking on pointer
arguments.
OSENABLE_BOUNDS_CHECKING

Bounds-checking code sections in various
Salvo services.
When TRUE, requires some ROM.

Services involving Salvo objects (e.g. events) normally accept
pointer arguments to any valid control blocks. However, when
OSSET_LIMITS is TRUE, OSENABLE_BOUNDS_CHECKING is set to
TRUE, and these services will only accept pointers that are within
the control blocks as specified by configuration parameters (e.g.
OSEVENTS) at compile time, and otherwise return an error code.
In other words, if OSSignalXyz() is compiled with OSSET_LIMITS
as TRUE and OSEVENTS as 4, passing it an event control block
pointer (ecbP) of OSECBP(5) or higher51 will result in OSSignalXyz() returning an error code of OSERR_BAD_P.
All users should leave this option at its default value.

51

174

ecbs are numbered from 1 to OSEVENTS.

Chapter 5 • Configuration

Salvo User Manual

OSSPEEDUP_QUEUEING: Speed Up Queue Operations
Name:
Purpose:
Allowed Values:
Default Value:
Action:
Related:
Enables:
Memory Required:

Notes

OSSPEEDUP_QUEUEING

To improve queueing performance.
FALSE: Use standard queueing algorithm.
TRUE: Use fast queueing algorithm.
FALSE

Configures queueing routines for fastest
performance.
–
–
When TRUE, requires a small amount of
ROM and RAM.

It is possible to improve the speed of certain operations involving
queues approximately 25% through the use of local variables in a
few of Salvo's internal queueing routines.
Applications with minimal RAM should leave this configuration
option at its default value.
See Chapter 9 • Performance for more information on queueing.

Salvo User Manual

Chapter 5 • Configuration

175

OSTIMER_PRESCALAR: Configure Prescalar for
OSTimer()
Name:
Purpose:
Allowed Values:
Default Value:
Action:
Related:
Enables:
Memory Required:

Notes

OSTIMER_PRESCALAR

To allow you maximum flexibility in
locating OSTimer() within your application.
0, 2 to (2^32)-1.
0

If non-zero, adds code and an 8- to 32-bit
countdown timer to OSTimer() to implement a prescalar.
OSBYTES_OF_DELAYS, OSBYTES_OF_TICKS
–
When TRUE, requires a small amount of
ROM, plus RAM for the prescalar.

If your application uses delays or timeouts, OSTimer() must be
called at the desired system tick rate. This is typically every 10100ms. If your processor has limited resources, it may be unacceptable to dedicate a (relatively slow) timer resource to
OSTimer(). By using OSTIMER_PRESCALAR you can call
OSTimer() at one rate but have it actually perform its timer-related
duties at a much slower rate, as dictated by the value of
OSTIMER_PRESCALAR.
Unlike some hardware prescalars, which provide powers-of-2 prescaling (e.g. 1:2, 1:4, ...), the Salvo timer prescalar is implemented
with a simple countdown timer, and can therefore provide a prescalar rate anywhere from 1:2 to 1:(2^32)-1.
A prescalar value of 1 accomplishes nothing and should not be
used.
Whenever OSTimer() is called and its prescalar has not reached 0,
a minimum of housekeeping is performed. When the prescalar
reaches zero, OSTimer() increments the system tick count (if enabled), and the scheduler processes delayed and/or timed-out tasks.

176

Chapter 5 • Configuration

Salvo User Manual

OSTYPE_TCBEXT0|1|2|3|4|5: Set Tcb Extension Type
Name:
Purpose:
Allowed Values:
Default Value:
Action:
Related:
Enables:
Memory Required:

Notes

OSTYPE_TCBEXT0|1|2|3|4|5

To allow you to change the type of a tcb
extension.
Any valid C-language type.
void *

Redefines OStypeTcbExt0|1|2|3|4|5.
OSENABLE_TCBEXT0|1|2|3|4|5,
OScTcbExt0|1|2|3|4|5, OStcbExt0|1|2|3|4|5
–
Dependent on definition – affects size of
tcbs.

A tcb extension can be of any valid type, and can have memory
type qualifiers applied to it so long as they do not conflict with existing OSLOC_XYZ configuration options.
To use tcb extensions, the associated OSENABLE_TCBEXT0|1|2|3|4|5
must be set to TRUE.
See the example for OSENABLE_TCBEXT0|1|2|3|4|5 for more information.

Salvo User Manual

Chapter 5 • Configuration

177

OSUSE_CHAR_SIZED_BITFIELDS: Pack Bitfields into
Chars
Name:
Purpose:
Allowed Values:

Default Value:
Action:
Related:
Enables:
Memory Required:

Notes

OSUSE_CHAR_SIZED_BITFIELDS

To reduce the size of Salvo objects.
FALSE: Places Salvo bitfields into intsized objects.
TRUE: Places Salvo bitfields into charsized objects.
FALSE

Alters the typedef for OStypeBitField.
–
–
When FALSE, reduces RAM requirements
slightly.

ANSI C supports bitfields in structures. Multiple bits are combined
into a single int-sized value, e.g.:
typedef struct {
int field0:2;
int field1:1;
int field2:4;
} bitfieldStruct;

Some compilers (e.g. HI-TECH PICC, Keil C51) allow the packing of bitfields into a single char-sized value in order to save
memory. To use this feature, set OSUSE_CHAR_SIZED_BITFIELDS
to TRUE. The Salvo type OStypeBitField will be of type char.
Not all compilers support this feature. If you are having problems
compiling a Salvo application, set OSUSE_CHAR_SIZED_BITFIELDS
to FALSE. The Salvo type OStypeBitField will then be of type
int.

178

Chapter 5 • Configuration

Salvo User Manual

OSUSE_EVENT_TYPES: Check for Event Types at
Runtime
Name:
Purpose:
Allowed Values:

Default Value:
Action:

Related:
Enables:
Memory Required:

Notes

OSUSE_EVENT_TYPES

To check for correct usage of an ecb
pointer.
FALSE: Event-type error checking is not
performed.
TRUE: When using an event service (e.g.
OSSignalSem()), Salvo verifies that the
event being operated on is correct for the
service.
TRUE

If TRUE, enables code to verify that the
event type is what the service expects.
This requires additional ROM, and a byte
is added to each ecb (RAM).
–
–
When TRUE, requires a moderate amount
of ROM.

Salvo uses event control block (ecb) pointers as handles to events.
These pointers are passed as arguments to user event services (e.g.
OS_WaitMsg()). A user might inadvertently pass an ecb pointer for
one type of event (e.g. a semaphore) to a service for another type
of event (e.g. OSSignalMsg()). The result would be unpredictable.
Therefore an extra layer of error checking can be enabled to ensure
that your application is protected against this sort of error.

Caution If you disable this configuration option you must be
especially careful with event service arguments. The use of #define statements with descriptive names (e.g. SEM1_P, SEM_COM1_P,
MSG12_P) for ecb pointers is highly recommended.

Salvo User Manual

Chapter 5 • Configuration

179

OSUSE_INLINE_OSSCHED: Reduce Task Call…Return
Stack Depth
Name:
Purpose:
Allowed Values:
Default Value:
Action:

Related:
Enables:
Memory Required:

Notes

OSUSE_INSELIG_MACRO

To reduce the call…return stack depth at
which Salvo tasks run.
FALSE, TRUE
FALSE

If FALSE, OSSched() is called as a function, and Salvo tasks run at a call…return
stack depth of 1 greater than that of OSSched(). If TRUE, OSSched() is used in an
inline form (i.e. macro), which reduces its
call…return stack depth by 1.
OSUSE_INLINE_OSTIMER

–
When FALSE, a small amount of extra
ROM and one additional call…return
stack level are used by OSSched(). When
TRUE, OSSched() uses less ROM and
only one call…return stack level.

Normally, you will call Salvo's scheduler in your application like
this:
main()
{
…
OSInit();
…
while (1) {
OSSched();
}
}

Since OSSched() calls Salvo tasks indirectly via function pointers,
each task will run with two return addresses pushed onto the target
processor's call…return stack: one inside of OSSched(), and one
inside of main().52 This means that the call…return stack depth
available to your functions called from within a Salvo task is equal
to 2 less than the target processor's maximum call…return stack
depth.
52

180

This assumes that the compiler uses a goto main(), and calls all functions
inside of main() from a call…return stack level of 0. Also, interrupts would
add additional return addresses to the call…return stack.

Chapter 5 • Configuration

Salvo User Manual

If your target processor's call…return stack depth is limited, and
you make deep, nested calls from within Salvo tasks or interrupt
routines, you may want to reduce the call…return stack depth at
which Salvo tasks run. By setting OSUSE_INLINE_OSSCHED to
TRUE, and calling the scheduler like this:
main()
{
…
OSInit();
…
while (1) {
#include "salvosched.c"
}
}

you can make Salvo tasks run with one fewer return addresses on
the call…return stack, thereby freeing up one call…return stack
level for other functions.

Salvo User Manual

Chapter 5 • Configuration

181

OSUSE_INLINE_OSTIMER: Eliminate OSTimer()
Call…Return Stack Usage
Name:
Purpose:

OSUSE_INLINE_OSTIMER

Allowed Values:
Default Value:
Action:
Related:
Enables:
Memory Required:

Notes

To enhance ISR performance and reduce
Salvo's call…return stack usage.
FALSE, TRUE
FALSE

If FALSE, OSTimer() is called as a function
from an ISR. If TRUE, uses a macro to perform the same operation.
OSUSE_INLINE_OSTIMER

–
When FALSE, a small amount of extra
ROM and one call…return stack level are
used by OSTimer(). When TRUE,
OSTimer() uses less ROM and no
call…return stack levels.

Normally you might call OSTimer() like this from your Salvo application:
void interrupt PeriodicIntVector ( void )
{
…
OSTimer();
}

This works for many applications. However, there may be disadvantages that arise when calling OSTimer() from an ISR. They
include slower interrupt response time and larger code size due to
the overhead of a call…return chain of instructions through
OSTimer()and the need to save context during interrupts, and the
consumption of one call…return stack level.
You

can

avoid

all

of

these problems by setting
OSUSE_INLINE_OSTIMER to TRUE and using OSTimer() like this:
void interrupt PeriodicIntVector ( void )
{
…
{ #include "salvotimer.c" }
}

This will insert an in-line version of OSTimer() into your ISR.

182

Chapter 5 • Configuration

Salvo User Manual

OSUSE_INSELIG_MACRO: Reduce Salvo's Call Depth
Name:
Purpose:
Allowed Values:
Default Value:
Action:

Related:
Enables:
Memory Required:

Notes

Salvo User Manual

OSUSE_INSELIG_MACRO

To reduce Salvo's maximum call depth and
parameter RAM usage.
FALSE, TRUE
TRUE

If FALSE, uses a function to perform a
common operation internal to Salvo. If
TRUE, uses a macro to perform the same
operation.
–
–
When FALSE, requires a small amount of
ROM and may require extra RAM on the
stack. When TRUE, requires a moderate
amount of ROM.

If your processor is severely RAM-limited, you should leave this
configuration option at its default value. For those processors that
have a lot of RAM available (e.g. those with a general-purpose
stack), then by setting OSUSE_INSELIG_MACRO to FALSE you should
realize a reduction in code size at the expense of an additional call
level and the RAM required to pass a tcb pointer as a parameter.

Chapter 5 • Configuration

183

OSUSE_MEMSET: Use memset() (if available)
Name:
Purpose:
Allowed Values:
Default Value:
Action:

Related:
Enables:
Memory Required:
Notes

OSUSE_MEMSET

To take advantage of the presence of a
working memset() library function.
FALSE, TRUE
FALSE

If FALSE, your code will use Salvo functions to clear global Salvo variables. If
TRUE, memset() will be used to clear
global Salvo variables.
OSLOC_XYZ

–
Requires some ROM when FALSE.

Compilers will often use the standard library function memset() to
clear (zero) global variables in start-up code.
If your target processor has a linear organization for RAM, you
should probably set OSUSE_MEMSET to TRUE.
If you target processor uses banked memory, memset() may not
work correctly for certain settings of OSLOC_ECB and OSLOC_TCB.
In these cases, you should set OSUSE_MEMSET to FALSE in order to
use Salvo's explicit byte-by-byte structure clearing functions.

184

Chapter 5 • Configuration

Salvo User Manual

Organization
The configuration options are loosely organized as outlined below,
by category.
Compiler in use:
Target processor:

Salvo User Manual

OSCOMPILER
OSTARGET

Tasks and events:

OSBIG_SEMAPHORES,
OSEABLE_BINARY_SEMAPHORES,
OSENABLE_EVENT_READING,
OSENABLE_EVENT_TRYING,
OSENABLE_FAST_SIGNALING,
OSENABLE_IDLE_COUNTER,
OSENABLE_IDLING_HOOK,
OSENABLE_MESSAGES,
OSENABLE_MESSAGE_QUEUES,
OSENABLE_SEMAPHORES, OSEVENTS,
OSMESSAGE_QUEUES, OSMESSAGE_TYPE,
OSTASKS, OSTASKS

Size-specific:

OSBYTES_OF_COUNTS,
OSBYTES_OF_DELAYS,
OSBYTES_OF_EVENT_FLAGS,
OSBYTES_OF_TICKS

Time and ticks:

OSCOLLECT_LOST_TICKS,
OSENABLE_TIMEOUTS,
OSTIMER_PRESCALAR

Optimizations:

OSCLEAR_GLOBALS,
OSOPTIMIZE_FOR_SPEED,
OSSPEEDUP_QUEUEING,
OSUSE_OSINSELIGQ_MACRO

Monitor and
debugging:

OSCLEAR_UNUSED_POINTERS, OSENABLE_STACK_CHECKING, OSLOGGING,
OSLOG_MESSAGES,
OSRPT_HIDE_INVALID_POINTERS,
OSRPT_SHOW_ONLY_ACTIVE,
OSRPT_SHOW_TOTAL_DELAY

Error checking:

OSDISABLE_ERROR_CHECKING,
OSUSE_EVENT_TYPES

Statistics:

OSGATHER_STATISTICS

Chapter 5 • Configuration

185

Memory allocation
and RAM banking:

OSLOC_ALL, OSLOC_COUNT, OSLOC_CTCB,
OSLOC_DEPTH, OSLOC_ECB, OSLOC_ERR,
OSLOC_LOGMSG, OSLOC_LOST_TICK,
OSLOC_MQCB, OSLOC_MSGQ, OSLOC_PS,
OSLOC_SIGQ, OSLOC_TCB, OSLOC_TICK,
OSMPLAB_C18_LOC_ALL_NEAR,
OSUSE_CHAR_SIZED_BITFIELDS,
OSUSE_MEMSET

Interrupts:

OSCALL_OSCREATEEVENT,
OSCALL_OSMSGQCOUNT,
OSCALL_OSMSGQEMPTY,
OSCALL_OSRETURNEVENT,
OSCALL_OSSIGNALEVENT,
OSCALL_OSSTARTTASK,
OSINTERRUPT_LEVEL,
OSTIMER_PRESCALAR

Porting:

OSCTXSW_METHOD, OSRTNADDR_OFFSET

Stack depth usage:
Code compression:
Linking to libraries:

Hooks to user code:
Scheduler behavior:

OSUSE_INLINE_OSSCHED,
OSUSE_INLINE_OSTIMER
OSCOMBINE_EVENT_SERVICES
OSCUSTOM_LIBRARY_CONFIG,
OSLIBRARY_CONFIG,
OSLIBRARY_GLOBALS, OSLIBRARY_TYPE,
OSLIBRARY_VARIANT, OSUSE_LIBRARY
OSENABLE_IDLING_HOOK,
SENABLE_OSSCHED_DISPATCH_HOOK,
OSENABLE_OSSCHED_ENTRY_HOOK,
OSENABLE_OSSCHED_RETURN_HOOK
OSDISABLE_FAST_SCHEDULING

Extensions:

OSENABLE_TCBEXT0|1|2|3|4|5,
OSTYPE_TCBEXT0|1|2|3|4|5

Cyclic Timers:

OSENABLE_CYCLIC_TIMERS

Table 2: Configuration Options by Category

Choosing the Right Options for your Application
You must select a compiler and a target when configuring Salvo
for your application. Depending on how many Salvo services you
wish to use in your application, you will also need to select and/or
configure other options. Consult the table below for further information:
Multitasking:

186

OSTASKS

Chapter 5 • Configuration

Salvo User Manual

Using events:

Using multiple event
types:
Keeping unused code
out of your
application:
Delaying tasks:
Waiting on events
with a timeout:
Setting the size of
event flags:
Keeping track of
elapsed time:
Counting the number
of context switches:
Using 16-bit
semaphores:
Using ROM and
RAM pointers:
Having an idle
function:
Checking call ...
return stack depth:
Collecting statistics:
Logging descriptive
error, warning and
status messages:
Optimizing your
application:
Making the most of
limited resources:
Avoiding event-type
mismatches:
Learning how Salvo
works:

Salvo User Manual

OSENABLE_BINARY_SEMAPHORES,
OSENABLE_EVENT_FLAGS,
OSENABLE_FAST_SIGNALING,
OSENABLE_MESSAGES,
OSENABLE_MESSAGE_QUEUES,
OSENABLE_SEMAPHORES, OSEVENTS
OSCOMBINE_EVENT_SERVICES
OSENABLE_EVENT_READING,
OSENABLE_EVENT_TRYING
OSBYTES_OF_DELAYS
OSBYTES_OF_DELAYS
OSBYTES_OF_EVENT_FLAGS
OSBYTES_OF_TICKS,
OSCOLLECT_LOST_TICKS
OSBYTES_OF_COUNTS,
OSGATHER_STATISTICS
OSBIG_SEMAPHORES
OSMESSAGE_TYPE
OSENABLE_IDLING_HOOK,
OSENABLE_IDLE_COUNTER
OSENABLE_STACK_CHECKING,
OSGATHER_STATISTICS
OSGATHER_STATISTICS
OSLOGGING, OSLOG_MESSAGES
OSCLEAR_GLOBALS,
OSOPTIMIZE_FOR_SPEED,
OSSPEEDUP_QUEUEING
OSTIMER_PRESCALAR
OSUSE_EVENT_TYPES
OSCLEAR_UNUSED_POINTERS,
OSRPT_HIDE_INVALID_POINTERS,
OSRPT_SHOW_ONLY_ACTIVE,
OSRPT_SHOW_TOTAL_DELAY

Chapter 5 • Configuration

187

Porting to other
compilers and / or
target processors:
Minimizing Salvo's
call…return stack
usage:

OSCTXSW_METHOD, OSRTNADDR_OFFSET,
OSUSE_MEMSET
OSUSE_INLINE_OSSCHED,
OSUSE_INLINE_OSTIMER

Calling Salvo
services from the
background and the
foreground:

OSCALL_OSCREATEEVENT,
OSCALL_OSMSGQCOUNT,
OSCALL_OSMSGQEMPTY,
OSCALL_OSRETURNEVENT,
OSCALL_OSSIGNALEVENT,
OSCALL_OSSTARTTASK

Locating Salvo's
variables in
memory:

OSLOC_ALL, OSLOC_COUNT, OSLOC_CTCB,
OSLOC_DEPTH, OSLOC_ECB, OSLOC_ERR,
OSLOC_LOGMSG, OSLOC_LOST_TICK,
OSLOC_MQCB, OSLOC_MSGQ, OSLOC_PS,
OSLOC_SIGQ, OSLOC_TCB, OSLOC_TICK,
OSMPLAB_C18_LOC_ALL_NEAR

Building an
application with
libraries:

OSCUSTOM_LIBRARY_CONFIG,
OSLIBRARY_CONFIG,
OSLIBRARY_GLOBALS, OSLIBRARY_TYPE,
OSLIBRARY_VARIANT, OSUSE_LIBRARY

Running multiple
tasks at same
priority (roundrobin):
Minimizing memory
usage:
Extending taskspecific
functionality:
Using cyclic timers
in place of tasks:

OSDISABLE_FAST_SCHEDULING

OSUSE_CHAR_SIZED_BITFIELDS
OSENABLE_TCBEXT0|1|2|3|4|5,
OSTYPE_TCBEXT0|1|2|3|4|5
OSENABLE_CYCLIC_TIMERS

Table 3: Configuration Options by Desired Feature

Predefined Configuration Constants
Predefined symbols are listed with their values below.

188

FALSE
TRUE

0
1

OSLOG_NONE, OSLOG_ERRORS,
OSLOG_WARNINGS, OSLOG_ALL

see OSLOG_MESSAGES

Chapter 5 • Configuration

Salvo User Manual

OSUNDEF, OSNONE

0

OSPIC12, OSPIC16, OSPIC17,
OSPIC18, OSIX86, OSI8051,
OSM68HC11, OSMSP430,
OSVAV8, etc.

see OSTARGET

OSAQ_430, OSGCC, OSHT_8051C,
OSHT_PICC, OSHT_V8C,
OSIMAGECRAFT, OSMW_CW,
OSMIX_PC, OSIAR_ICC,
OSMPLAB_C18, OSKEIL_C51,

see OSCOMPILER

etc.
OSFROM_BACKGROUND,
OSFROM_FOREGROUND,
OSFROM_ANYWHERE

see OSCALL_XYZ

OSRTNADDR_IS_PARAM,
OSRTNADDR_IS_VAR,
OSVIA_OSCTXSW,
OSVIA_OSDISPATCH, etc.

see OSCTXSW_METHOD

OSALL_BITS, OSANY_BITS,
OSEXACT_BITS

see OS_WaitEFlag()

Table 4: Predefined Symbols

Obsolete Configuration Parameters
Obsolete configuration parameters – id defined – are automatically
caught during the preprocessing stage. Including them in your
salvocfg.h will result in a compile-time error message indicating
the name of the configuration option. Some error messages include
instructions on alternate, renamed or related configuration options.

Salvo User Manual

Chapter 5 • Configuration

189

190

Chapter 5 • Configuration

Salvo User Manual

Chapter 6 • Frequently Asked
Questions (FAQ)

General
What is Salvo?
Salvo is a powerful and feature-rich real-time operating system
(RTOS) for single-chip microcontrollers with limited ROM and
RAM. By imposing a few constraints on conventional RTOS programming, Salvo rewards you with the power of an RTOS without
all of the RAM requirements.
Salvo is so small that it runs where other RTOSes can't. Its RAM
requirements are minuscule, and it doesn't need much ROM, either.
Salvo is not a state machine. It is not a "a neat trick." It is not an
app note. Salvo is all the RTOS code you need and more to create
a high-performance embedded multitasking program in systems
where kilobytes of ROM are a luxury and available RAM is measured in tens of bytes.

Is there a shareware / freeware / open source version of
Salvo?
There is a freeware version called Salvo Lite.
Processor- and compiler-specific freeware libraries are provided as
part of each Salvo Lite distribution. Each freeware library supports a limited number of tasks and events. All of the default functionality is included in the freeware libraries. If you need more
tasks and/or events, or you need access to Salvo's advanced functionality, then you should consider purchasing Salvo LE or Pro.

Salvo User Manual

191

Salvo Pro includes all source code. Source code is not included53 in
Salvo Lite or LE. Salvo is not open source.

Just how small is Salvo?
On a single-chip microcontroller, a typical54 multitasking application might need around 1K ROM and around fifty bytes of RAM
for all of Salvo's code and data.

Why should I use Salvo?
If you want to:
•
• • get your embedded product to market ahead of
the competition,
• • add greater software functionality to your
existing hardware design,
• • improve the real-time performance of a
complex design,
• • not have to re-invent the wheel,
• • have a powerful framework to do multitasking
programming,
• • control the increasing complexity of your
applications,
• • minimize your hardware costs by using
smaller and cheaper processors,
• • not be left behind by the multitasking / RTOS
wave and/or
• • maximize the reliability of your complex
applications
then Salvo is for you.
Low-cost single-chip microcontrollers are capable of hosting sophisticated real-time applications, but programming them to do so
can be quite a challenge. Real-time kernels can simplify the design
of complex software. They provide proven mechanisms to accomplish a variety of well-understood operations within predictable
time frames. Unfortunately, most commercial real-time offerings
require large amounts of ROM and RAM – requirements that are
largely incompatible with these chips. Programmers of low-end
53
54

192

Except for a few specific files in certain freeware versions.
Microchip® PIC16C64 with five concurrent tasks and five events.

Chapter 6 • Frequently Asked Questions (FAQ)

Salvo User Manual

embedded processors have been at a disadvantage when developing non-trivial applications.
Salvo changes all of that. Now you can develop applications for
inexpensive one-chip microcontrollers similar to how you would
for a Pentium® in an embedded application.
Salvo will get your application up and running quickly. It provides
you with a clean and easily-understood multitasking framework
that uses a minimum of memory to get the job done.

What should I consider Salvo Pro over Salvo LE?
With Salvo Pro, you have the Salvo source code. With source code
you have complete access to all of Salvo's configurability. This
means that you can build custom Salvo libraries with Salvo Pro.
Plus, when your compiler is updated with support for new processors or with new optimizations, you can take advantage of the new
compiler features without waiting for a Salvo libraries to be rebuilt
and packaged into a new Salvo release.
Another advantage of having Salvo Pro is that it allows you to step
through the Salvo code in C when symbolically debugging your
application.
Additionally, if / when bugs are found and identified in the Salvo
code, you can make changes locally without having to wait for a
new Salvo release.
Lastly, some organizations demand access to source code for code
reviews and code maintenance.
You can upgrade from Salvo LE to Salvo Pro at anytime.

What can I do with Salvo?
You can throw out any preconceived notions on how difficult or
time-consuming embedded programming can be. You can stop
dreaming about multiple, independent processes running concurrently in your application without crashing. You can reorganize
your code and no longer worry about how a change in one area
might affect another. You can add new functionality to your existing programs and know that it will integrate seamlessly. You can
easily link external and internal events to program action.

Salvo User Manual

Chapter 6 • Frequently Asked Questions (FAQ)

193

Once you start creating applications with Salvo, you can focus on
adding functionality to and improving the performance of your application by creating tasks and events tailored specifically to it.
You can create multitasking applications where tasks pass information to other tasks and the rest of your application. You can prioritize the tasks so that your processor is spending its time doing
what's most important, instead of unnecessary housekeeping
chores. You can have events control how and when tasks run. You
can worry a lot less about interrupts. You can write powerful, efficient and reliable multitasking applications with predictable realtime performance.
And you can do all of this a lot more quickly than you'd expect.

What kind of RTOS is Salvo?
Salvo is a priority-based, event-driven, cooperative, multitasking
RTOS. It is designed to run on processors with severely limited
resources (primarily ROM and RAM).

What are Salvo's minimum requirements?
Salvo requires a full-featured ANSI-C-compliant C compiler from
a third party. Contact the factory or visit the website for a list of
tested and/or approved compilers.
If you're not already reasonably proficient in C, you will need to
review certain concepts (particularly pointers, if you plan on using
messages and message queues) before beginning with Salvo. You
don't need to be an expert C programmer to use Salvo.

What kind of processors can Salvo applications run on?
Salvo requires a processor with a hardware call…return stack of at
least 4 levels and enough memory for Salvo's code and data. ROM
and RAM requirements vary, and are controlled primarily by your
application's source code and settings in the Salvo configuration
file salvocfg.h.

194

Chapter 6 • Frequently Asked Questions (FAQ)

Salvo User Manual

My compiler doesn't implement a stack. It allocates
variables using a static overlay model. Can it be used
with Salvo?
Salvo has been implemented with this type of compiler, with conventional compilers (parameters and return addresses on the stack),
and with compilers that take an in-between approach.
Where a general-purpose stack is present, Salvo's use of it is
minimal.55 It can run on stack-less processors as well as any
processor with a stack, from a PICmicro® to a Pentium®.

How many tasks and events does Salvo support?
Salvo supports an unlimited number of tasks and events. The number of tasks and events in your application is limited only by available RAM. Salvo's default configuration supports up to 255 tasks,
255 events and 255 message queues.

How many priority levels does Salvo support?
Salvo supports 16 distinct priority levels. Tasks can share priority
levels.

What kind of events does Salvo support?
Salvo supports binary semaphores, counting semaphores, event
flags, messages and message queues. You can create ("init")
events, signal ("post", "put", "unlock", "release", "send") events
and have tasks wait ("pend", "get", "lock", "acquire", "receive") on
each event.

Is Salvo Y2K compliant?
Yes. Salvo does not provide any functions for reporting or setting
the absolute time of day and date (e.g. 10:22.36pm, Nov. 11,
1999). Therefore Salvo is by definition Y2K compliant.

55

Salvo User Manual

A stack pointer (SP) and/or PUSH and POP instructions are evidence of a
general-purpose stack.

Chapter 6 • Frequently Asked Questions (FAQ)

195

Where did Salvo come from?
Salvo 1.0 was originally developed in assembly language for use in
a low-cost, high-performance multichannel racecar data acquisition
system. Its appeal to a wider audience was quickly recognized,
whereupon it was rewritten in C for greater portability and configurability.

Getting Started
Where can I find examples of projects that use Salvo?
Every Salvo distribution has demo, tut (tutorial) and ex (example) folders. Refer to File and Program Descriptions in the Salvo
User Manual for a test system (e.g. sysa) that's similar to yours.
Then search these folders in your Salvo installation for project
files, source code (usually main.c) and configuration files (salvocfg.h).

Which compiler(s) do you recommend for use with Salvo?
As a matter of policy, we do not take any positions regarding the
compilers we have certified for use with Salvo. The fact that we've
certified a particular compiler should suggest to you that we consider it to be a production-level tool. When purchasing a compiler,
we suggest you base your decision on the quality of its output,
suitability to the task, flexibility, IDE (if included), debugging
tools, support and price.
Unless otherwise noted in the Salvo Compiler Reference Manuals,
compilers for the same target are generally interchangeable as far
as Salvo is concerned.

Is there a tutorial?
Yes. An in-depth tutorial can be found in the Salvo User Manual.

196

Chapter 6 • Frequently Asked Questions (FAQ)

Salvo User Manual

Apart from the Salvo User Manual, what other sources of
documentation are available?
The Application Notes contain information on a variety of topics.
The Salvo Compiler Reference Manuals contain compiler-specific
information.

I'm on a tight budget. Can I use Salvo?
You can use Salvo Lite, with its complete set of freeware libraries,
to create fully functioning Salvo applications. You'll be limited to
the numbers of tasks and events your application can support.

I only have an assembler. Can I use Salvo?
No. You will need a certified C compiler to use Salvo.

Performance
How can using Salvo improve the performance of my
application?
If you're used to programming within the conventional foreground
/ background loop model, converting your application to a Salvo
application may yield substantial performance benefits.
For example, it's not uncommon to write a program that polls
something (say an I/O pin) repeatedly and performs a complicated
and time-consuming action whenever the pin changes. You might
have a timer interrupt which calls a subroutine to poll a port pin
and XOR it against its previous value. If the pin changes, then you
might set a bit in a global status byte, which is then tested every
time through your main loop. If the bit is set, you disable interrupts, clear the status bit, reenable interrupts and then take an appropriate action.
The problem with this approach is that your program is consuming
processor cycles while sampling information that remains unchanged for most of the time. The more infrequently the event (in
this case, the change on I/O pin) occurs, the more inefficient your
program is.

Salvo User Manual

Chapter 6 • Frequently Asked Questions (FAQ)

197

The solution is to employ an event-based approach by using Salvo.
When a task is made to wait an event, and the event is not available (e.g. the I/O pin hasn't changed), then the task is put into a
waiting state. From this time forward, until the event occurs, not a
single processor cycle is expended on waiting for the event. Zip,
zero, nada. When the event does finally occur, the task will process
the event as soon as it is made to run by the scheduler. In other
words, it's the event that drives all the other actions directly. With
events driving your application, it can spend its time on the most
important things, as defined by you, the programmer.
It's important that you understand the distinction between polled
and event-based actions.

How do delays work under Salvo?
Salvo provides a simple means of delaying tasks. While a task is
delayed, it consumes a minimum of processor resources, and your
other (non-delayed) tasks can continue to run. The overhead to
support one or more delayed tasks is the same. You can specify
delays to the resolution of the system timer, which is under your
control.
See the Timer and Timing section in this FAQ for more information.

What's so great about having task priorities?
The point of assigning priorities to tasks is to make the most of
your processor's power by having it always doing what is most important at that particular instant in time.
For example, say you have an instrument whose primary purpose
is to generate moderate-frequency waveforms. But you'd also like
to monitor various analog voltages in the instrument to ensure no
out-of-range conditions. By assigning the waveform-generating
task a high priority, and the analog-sampling task a low priority,
the Salvo application will automatically run the sampling task
when there's no demand for the waveform to be generated. But
while the waveform is being generated, the sampling task will not
interfere.
All you have to do in Salvo is assign each task an appropriate priority, and ensure that each task context-switches often enough to
allow other tasks to run as needed.

198

Chapter 6 • Frequently Asked Questions (FAQ)

Salvo User Manual

When does the Salvo code in my application actually run?
Salvo's code runs only when you explicitly call Salvo's user services within your application. In most cases it's pretty obvious
when your processor is running Salvo code – for example, when
you start a task by calling OSCreateTask() or OSStartTask().
When the scheduler and timer actually run is perhaps a little less
obvious. The scheduler runs as part of any context switch in your
code, and it also runs when there are no tasks eligible to run. The
timer runs whenever it is called at the periodic system timer rate,
which is usually done via a periodic interrupt.

How can I perform fast, timing-critical operations under
Salvo?
In order to control critical timing under any RTOS, follow these
two rules: 1) give timing-critical tasks high priorities, and 2) use
Salvo's flexible features to prevent or delay it from doing anything
during a critical time period.
Since Salvo is a cooperative multitasking RTOS, during a timingcritical task there is only one source of potential interference – interrupts. Interrupts which might involve Salvo would be those that
signal events and / or call the system timer OSTimer(). By preventing calls to Salvo services during timing-critical operations
you can guarantee the proper operation of your system.
If, on the other hand, your application can tolerate the timing jitter
that will occur if Salvo services are invoked during a critical period, then you may not have much to worry about. This is usually
the case with operations whose frequency is much less (e.g. 1/50)
than that of the system timer.

Memory
How much will Salvo add to my application's ROM and
RAM usage?
Salvo's ROM requirements depend on how many of its functions
you call, and its RAM requirements depend on how many tasks
and resources you create. Salvo was specifically designed for processors with limited memory resources, and so it requires only a

Salvo User Manual

Chapter 6 • Frequently Asked Questions (FAQ)

199

small fraction of what a typical multitasking kernel would normally need.
The Salvo User's Manual contains specific information on memory
requirements for a variety of representative test systems.

How much RAM will an application built with the libraries
use?
Using a PIC16 library56 that supports multitasking, delays, and
events (binary and counting semaphores, as well as messages), an
application will need
•
•
•

● 10 bytes of RAM for Salvo's global
variables57
● 5 bytes of RAM per task
● 3 bytes of RAM event

The compiler will need some additional RAM to handle local variables, interrupt save and restore, etc. But the numbers above represent how little RAM Salvo needs to implement all its functionality.

Do I need to worry about running out of memory?
No. Salvo's RAM memory requirements are fixed at compile time.
They are simply:
•
•
•
•
•
•

#(tasks) x sizeof(task control block)
+ #(events) x sizeof(event control block)
+ #(tcb pointers58) x sizeof(tcb pointer)
+ #(message queues) x sizeof(message queue
control block)
+ #(message queues) x sizeof(user-defined
message queues)
+ sizeof(variables associated with configuration
options)

These requirements do not change during runtime, and are not dependent on call depth, the status of any of the tasks, the values of
56
57
58

200

sfp42Cab.lib, for the PIC16F877 for use with the HI-TECH PICC
compiler.
4 of the 10 bytes of global variables are for the 32-bit elapsed time counter,
which can be disabled by doing a source-code build (no libraries).
2 or 3, depending on the configuration.

Chapter 6 • Frequently Asked Questions (FAQ)

Salvo User Manual

any of the events or any other multitasking-related issues. Once
you define tasks and events in Salvo and your application has the
memory to support them, you can do whatever you want without
the fear of running out of memory.
Salvo cannot "run out of memory" during runtime.

If I define a task or event but never use it, is it costing me
RAM?
Yes. The RAM memory is allocated at compile time.

How much call ... return stack depth does Salvo use?
Normal stack depth is 4, and in some instances Salvo can be configured to use a maximum call…return stack depth of 3. This
means that no Salvo function will require a call-return stack more
than 4 levels deep, not including interrupts. This is accomplished
by setting the following configuration parameters in your salvocfg.h:
#define
#define
#define
#define

OSLOGGING
OSUSE_INLINE_OSSCHED
OSUSE_INLINE_OSTIMER
OSUSE_OSINSELIGQ_MACRO

FALSE
TRUE
TRUE
TRUE

and making the appropriate changes to your source code (see the
configuration options' descriptions for more information). These
options will configure Salvo to use in-line forms of various functions (thus saving one or more call…return stack levels) and to use
simple function return codes without debug messages (saving another call…return stack level).
When calling Salvo functions (e.g. OSSignalMsg()) from ISRs,
remember that ISRs are likely to run one or more stack levels deep,
depending on when the interrupt is serviced. This will affect the
maximum call ... return stack depth in your application.
By choosing OSENABLE_STACK_CHECKING Salvo will monitor the
stack depth of all of its functions and report back the maximum
stack depth reached. This is especially useful when simulating your
application by running Salvo on a PC.
Note that the numbers above are based on Salvo's inherent
call...return tree, and do not include any additional stack depth due

Salvo User Manual

Chapter 6 • Frequently Asked Questions (FAQ)

201

to how your compiler does certain things like indirect function
calls.

Why must I use pointers when working with tasks? Why
can't I use explicit task IDs?
Salvo user services originally took task, event and message queue
IDs (simple integer constants) as parameters to refer to Salvo objects. The advantage of this approach was that it was very easy for
beginners to understand, it easily accommodated run-time error
checking, and the memory requirements (mainly when passing parameters) were minimal. However, it also had several severe disadvantages, including increased code size, lack of flexibility, poor
run-time performance and increased call…return stack usage.
Salvo services now use pointers as parameters to refer to Salvo objects. Along with the attendant advantages that pointers bring with
them, Salvo's syntax is more like other, larger RTOSes. Somewhat
surprisingly, the memory requirements actually decreased for
many target processors.
With the pointer-based approach, the simplest way to refer to a
task is to use the OSTCBP() macro, which returns a pointer to the
tcb of a particular task. This is a compile-time constant (it's an address of an array element), and on many targets59 uses the same
amount of memory as an 8-bit integer constant. Similar macros
exist for events, message queues, etc. These macros allow you to
refer to Salvo objects explicitly.
An alternative approach is to use a handle, a variable that contains
a pointer to a particular task's tcb. This offers flexibility but has the
disadvantage that it consumes extra RAM. For some applications
handles can be very useful.
Using the C #define preprocessor directive for event IDs can substantially improve code legibility. For example, use:
/* pointer to display binSem. */
#define BINSEM_DISP_P OSECBP(3)
/* create display semaphore, init to 1. */
OSCreateSem(BINSEM_DISP_P, 1);
...
/* get display. */
OS_WaitSem(BINSEM_DISP_P, OSNO_TIMEOUT);
59

202

E.g. PIC16 and PIC17 series of PICmicro MCUs.

Chapter 6 • Frequently Asked Questions (FAQ)

Salvo User Manual

...
/* release display. */
OSSignalSem(BINSEM_DISP_P);

to reference the binary semaphore that is used as a resource to control access to a display in a easy-to-read manner.

How can I avoid re-initializing Salvo's variables when I
wake up from sleep on a PIC12C509 PICmicro MCU?
The PIC12C509 has a simple architecture (no interrupts, single reset vector) and always vectors to the last location in ROM when it
wakes from sleep due to the watchdog timer or wake-on-pinchange. Normally, the startup code generated by the compiler will
initialize all static and global variables immediately after any type
of reset – power-on reset (POR) or otherwise. This will reset all of
Salvo's variables to 0, equivalent to calling OSInit().
Since you'd like to preserve the state of your multitasking system
on wake-from-sleep, and not reset it, you must declare Salvo's
variables to be of type persistent. This instructs the compiler to
skip the initialization for these variables. If you are using HITECH PICC, the easiest way to declare Salvo's variables as persistent is to use the OSLOC_ALL configuration option, like this:
#define OSLOC_ALL bank1 persistent

This will place all of Salvo's variables in RAM bank 1, and will
prevent the startup code (which is executed after every type of reset, not just POR) from resetting the variables to zero. If you use
this method, you must call OSInit() after each POR (and not after other types of reset) in order to properly initialize Salvo.

Libraries
What kinds of libraries does Salvo include?
Every Salvo distribution includes the freeware Salvo libraries.
Additionally, the Salvo LE and Pro include the standard Salvo libraries. There are many different library types, depending on how
much functionality you need.

Salvo User Manual

Chapter 6 • Frequently Asked Questions (FAQ)

203

What's in each Salvo library?
Each Salvo library contains the default Salvo functionality for the
particular library type. Additionally, each library is compiled for a
default number of Salvo objects (tasks, events, etc.). Some libraries
(notably those for targets with extremely limited RAM) have a
subset of the normal functionality.

Why are there so many libraries?
Each library is generated with a particular compiler, target processor and library type in mind. As a result, a large number of libraries is required to span all the possible combinations.

Should I use the libraries or the source code when
building my application?
If you don't have Salvo Pro, you'll have to use the libraries.
With Salvo Pro, you should use the standard libraries until you
reach a situation where the configuration of the library no longer
suits your application, e.g. you want 32-bit delays and the library
supports only 8-bit delays. In that case, you can use the source
code and some configuration options to build a custom Salvo library.
Alternatively, you can build a Salvo application wholly from the
Salvo source code, bypassing the libraries altogether.

What's the difference between the freeware and standard
Salvo libraries?
There is very little difference. The freeware libraries are limited to
a maximum number of Salvo objects. The standard libraries support as many Salvo objects as you can fit in RAM.

My library-based application is using more RAM than I
can account for. Why?
The default number of Salvo objects used by each library requires
a certain amount of RAM, whether or not you use all of those objects. If your application uses fewer objects, you can reduce the

204

Chapter 6 • Frequently Asked Questions (FAQ)

Salvo User Manual

application's RAM requirements with a different set of configuration objects. See Chapter 8 • Libraries for more information.

I'm using a library. Why does my application use more
RAM than one compiled directly from source files?
Each library is created with its own default configuration. Some
configurations include Salvo features that require one or more
bytes of RAM. For example, the library may be configured to support a single message queue as well as other event types. Each
message queues requires its own message queue control block
(mqcb), and RAM has been allocated for it in the library. Therefore even if you do not use message queues in your application
when linking to a library, RAM is allocated for this (unused) message queue.
You can reduce some of the library's RAM requirements by overriding the RAM allocations. See Chapter 8 • Libraries for more
information.

I'm using a freeware library and I get the message "#error:
OSXYZ exceeds library limit – aborting." Why?
You've probably set OSXYZ to a number that exceeds the maximum
value supported by the library. Remove OSXYZ from your salvocfg.h or upgrade to Salvo LE or Pro.

Why can't I alter the functionality of a library by adding
configuration options to my salvocfg.h?
The configuration options affect a library only at compile time.
Since the libraries are precompiled, changing configuration options
in your salvocfg.h will have no effect on them. Choose a different library with the functionality you desire, or use the source
code.

Salvo User Manual

Chapter 6 • Frequently Asked Questions (FAQ)

205

The libraries are very large – much larger than the ROM
size of my target processor. Won't that affect my
application?
No. Your compiler will extract only the modules that it needs from
the library you're using. In fact, linking to libraries creates the
smallest possible Salvo applications.

I'm using a library. Can I change the bank where Salvo
variables are located?
No. On banked target processors, the locations of the Salvo variables are determined by the library. To "move" the variables to
another bank, you'll need to build a custom library, or use the
source files, set your own configuration options, and recompile.

Configuration
I'm overwhelmed by all the configuration options. Where
should I start?
Nearly all of the configuration options are for Salvo Pro users doing source-code builds, or building custom libraries.
If you're using a Salvo library, the only configuration options you
need are the ones that tell Salvo which kind of library you're using
and how many Salvo objects you want in your application. You
needn't worry too much about the others.
If you have Salvo Pro, or you want more objects than are supported by default in the standard libraries, you'll find various configuration options useful when tailoring Salvo to your application.
Start with the default configurations (no configuration options in
your salvocfg.h), which are described in Chapter 5 • Configuration. Then modify your salvocfg.h as you enable Salvo functionality that differs from the default.
Three good places to get acquainted with the configuration options
and how they're used are the tutorial, example and demonstration
programs in the standard Salvo distribution. By examining the programs and their corresponding salvocfg.h files you should be
able to develop a feel for when to use a particular configuration

206

Chapter 6 • Frequently Asked Questions (FAQ)

Salvo User Manual

option. These programs are found in \salvo\tut, \salvo\ex
and \salvo\demo.

Do I have to use all of Salvo's functionality?
You can use as little or as much as you like. Only those portions
that you use will be incorporated into (i.e. will take up ROM and
RAM in) your final executable. By choosing configuration options
you can control how much functionality Salvo delivers to your application.

What file(s) do I include in my main.c?
In terms of Salvo services, all you need to include is salvo.h. For
some target processors, including salvo.h is enough to automatically include the necessary processor-specific header files. If not,
you'll also need to include target-specific header files in all of your
source files – see your compiler's documentation for more information.

What is the purpose of OSENABLE_SEMAPHORES and
similar configuration options?
Salvo Pro users who compile their applications by linking multiple
Salvo source files may find this type of configuration option useful. That's because entire modules can be disabled simply setting
the configuration option to FALSE in salvocfg.h instead of changing the setup to your compiler / project / IDE.

Can I collect run-time statistics with Salvo?
By enabling OSGATHER_STATISTICS Salvo will track and report the
number of context switches, warnings, errors, timeouts and calls to
the idle function (if enabled).

How can I clear my processor's watchdog timer with
Salvo?
Good coding practice dictates that watchdog timers only be cleared
from a single place within an application. An excellent place to do
so is from within Salvo's scheduler, and by default, this is what

Salvo User Manual

Chapter 6 • Frequently Asked Questions (FAQ)

207

Salvo does. Therefore, if a task fails to release control back to the
scheduler, the watchdog will time out, indicating a fault.
Salvo Pro users can clear the processor's watchdog timer from another location by redefining OSCLEAR_WATCHDOG_TIMER() in salvocfg.h to do nothing, and clearing the watchdog timer elsewhere
in their code.

I enabled timeouts and my RAM and ROM grew
substantially– why?
Salvo makes the most efficient use of RAM and ROM based on the
configuration options you've chosen. Adding support for timeouts
requires an additional amount of RAM for each task, and extra
code in ROM, in order to support a task's ability to wait on an
event with a timeout. RAM- and ROM-wise, this is probably the
most "expensive" Salvo configuration option.

Timer and Timing
Do I have to install the timer?
If you want to make any use of Salvo's time-based functions (task
delays, timeouts when waiting for a resource, elapsed time, etc.)
you must install the timer. Simple multitasking and support for
events do not require the timer, but delays and timeouts do.
Salvo Pro users can configure OSBYTES_OF_DELAYS to a non-zero
value appropriate for the application in order to use Salvo's delay
and timeout features in a source-code build. Similarly, configuring
OSBYTES_OF_TICKS to a non-zero value in a source-code build enables the use of Salvo's elapsed time features.

How do I install the timer?
In your application you must call OSTimer() at the tick rate you
feel is appropriate for your application. Usually this is done by creating a periodic interrupt at the desired tick rate, and having the
associated ISR call OSTimer(). OSTimer() must be called in only
one place in your application.

208

Chapter 6 • Frequently Asked Questions (FAQ)

Salvo User Manual

I added the timer to my ISR and now my ISR is huge and
slow. What should I do?
See "Why did my interrupt service routine grow and become
slower when I added a call to OSTimer()" in this FAQ.

How do I pick a tick rate for Salvo?
The ideal Salvo "tick" rate is dependent on the application, and
hence is configurable. Rates on the order of 10-100Hz are commonly used. The tick rate defines the timer resolution in Salvo, but
does not directly affect the latency of a task made ready-to-run.
The context-switching rate is independent of the tick rate. A faster
tick rate requires more processor, but it gives better timer resolution, and may require additional memory for the delay fields in the
task blocks.
Once you've chosen a tick rate, you must configure your system to
call OSTimer() each time the tick occurs. This is usually done via
a periodic interrupt.

How do I use the timer prescalar?
A linear prescalar for the Salvo timer is provided to create a slower
Salvo "tick" rate independent of the timer to which the Salvo timer
is chained. For example, on a 4MHz system with a hardware timer
that generates interrupts at a 500 Hz rate (i.e. every 2 ms), by defining OSTIMER_PRESCALAR to 5 the desired Salvo tick rate will be
100Hz (i.e. every 10ms). The maximum value for the prescalar is
(2^32)-1, and to disable it altogether simply set it to 0 (the default).

I enabled the prescalar and set it to 1 but it didn't make
any difference. Why?
The Salvo timer prescalar is enabled if OSTIMER_PRESCALAR is set
to a number greater than or equal to 1, resulting in prescalar rates
of 1:1, 1:2, 1:3, ... 1:(2^32)-1. A prescalar value of 1 will add a few
instructions to OSTimer() and will require a byte of RAM storage
for OStimerPS, but it will not change the rate at which OSTimer()
is called, since the prescalar rate is 1:1. In order to change the rate
at which OSTimer() is called in your application, choose a value
for the timer prescalar that is 2 or greater.

Salvo User Manual

Chapter 6 • Frequently Asked Questions (FAQ)

209

What is the accuracy of the system timer?
As long as the system tick rate is slow enough to give Salvo's system timer OSTimer() enough time to do its job, the system timer
will have no more than 1 timer tick of inaccuracy.

What is Salvo's interrupt latency?
Salvo must disable interrupts while certain internal operations are
being performed. Every effort has been made to minimize Salvo's
interrupt latency. However, because of Salvo's configurability it's
difficult to provide a general answer to this question. Your best bet
is to create your own test programs with Salvo Lite to test Salvo's
interrupt latency.

What if I need to specify delays larger than 8 bits of ticks?
You have three options. You can call OS_Delay() multiple times
(sequentially, or in a loop) to create longer delays.
With Salvo Pro, you can change the configuration parameter
OSBYTES_OF_DELAYS to use 16- or 32-bit delays instead of 8-bit
delays. This will consume an additional 1 or 3 bytes of RAM per
task, respectively.
Or you can make use of the OSTIMER_PRESCALAR configuration
parameter with Salvo Pro. However, this approach will reduce the
resolution of the system timer.

How can I achieve very long delays via Salvo? Can I do
that and still keep task memory to a minimum?
The maximum delay and timeout length is user-configurable as
(2^(n x 8))-1, where n is the size in bytes for the task's delay field.
For example, if 16-bit delays are selected, delays and timeouts of
up to 65535 clock ticks are possible. Since all tasks have the samesize delay field, the total amount of RAM memory dedicated to
holding the delays is
sizeof(delay field) x #(tasks).
If your application uses delays and timeouts sparingly, but requires
a very long timeout, you can use a small value for OSBYTES_OF_DELAYS (e.g. 1, for 1 byte / 8 bits / maximum count of 255) and

210

Chapter 6 • Frequently Asked Questions (FAQ)

Salvo User Manual

nest the call within a local loop to achieve a multiple of the maximum timeout supported by Salvo. For example, using
for (i = 0; i <= TIMEOUT_MULTIPLE; i++) {
OS_WaitSem(SEM_NAME_P, MAX_TIMEOUT);
if ( !OSTimedOut() )
break;
}
if (OSTimedOut()) {
/* loop is over, are we here because of a
/* timeout or did we wait the semaphore
/* successfully?
}

*/
*/
*/

within a task (where the loop counter i is static) will result in a
maximum timeout of TIMEOUT_MULTIPLE x MAX_TIMEOUT. With a
looping construct like this a timeout or delay can be made arbitrarily long at the cost of only a single static variable local to the task
of interest.
Note that many target processors do math efficiently only for their
native data size. Therefore Salvo's timer code will grow substantially on an 8-bit PICmicro if you use 32-bit delays.
An alternative method is to use Salvo's timer prescalar. This
method will affect all Salvo delays and timeouts, system-wide. In
order to use Salvo's delays and timeouts OSBYTES_OF_DELAYS must
be non-zero. In order to use the timer prescalar,
OSTIMER_PRESCALAR must be set to a non-zero value.

Can I specify a timeout when waiting for an event?
Yes. When waiting for an event you can specify an optional timeout in system ticks. OSENABLE_TIMEOUTS must be TRUE in order to
wait with timeouts.

Does Salvo provide functions to obtain elapsed time?
Yes. Salvo provides two elapsed time functions, OSGetTicks()
and OSSetTicks(). These functions get and set, respectively, the
current number of timer ticks since the free-running timer ticks
counter rolled over. To use these elapsed time functions, the configuration parameter OSBYTES_OF_TICKS must be non-zero.

Salvo User Manual

Chapter 6 • Frequently Asked Questions (FAQ)

211

In this example, a task waits for a message, and once obtained, calculates the amount of elapsed time in timer ticks (OSBYTES_OF_TICKS is defined to be 4 in salvocfg.h):
...
static OStypeMsgP msgP;
static OStypeTick elapsedTicks;
...
while (1) {
...
OSSetTicks(0);
OS_WaitMsg(MSG_ID, &msgP, OSNO_TIMEOUT);
elapsedTicks = OSGetTicks();
printf("%lu ticks have passed\n", elapsedTicks);
...
}

How do I choose the right value for
OSBYTES_OF_TICKS?
Salvo uses a free-running counter to monitor system ticks. This
counter is incremented by 1 each time the system timer OSTimer()
is called by your application.60 The size of this counter, and hence
the rollover period, is controlled by the configuration parameter
OSBYTES_OF_TICKS.
Since system ticks are used only for obtaining elapsed time and
statistics, your choice for the value of OSBYTES_OF_TICKS is entirely dependent on the longest elapsed time you wish to be able to
measure accurately.
For example, let's assume that you have written your application to
have an effective tick rate of 100Hz by enabling Salvo's system
timer, choosing an appropriate value for OSTIMER_PRESCALAR, and
calling OSTimer() from inside a timer-interrupt ISR. If
OSBYTES_OF_TICKS were defined to be 2, the longest time interval
you could measure would be (65535/100) seconds, or just under 11
minutes. If more than 11 minutes elapse before calling OSGetTicks(), the reported elapsed time will be the actual elapsed time
modulo 11 minutes, an erroneous result.

60

212

For every OSTIMER_PRESCALAR calls to OSTimer() if OSTIMER_PRESCALAR
is nonzero.

Chapter 6 • Frequently Asked Questions (FAQ)

Salvo User Manual

My processor has no interrupts. Can I still use Salvo's
timer services?
Yes. As long as you have some form of a timer, you can use
OSTimer(). For example, you can monitor a free-running counter
for overflow, and each time this occurs, you can call OSTimer().
This results in a system tick period equal to the timer overflow period. You can lengthen this period by using Salvo's timer prescalar.
As long as you check often enough not to miss an overflow, you'll
have an accurate system timer.
See How can I avoid re-initializing Salvo's variables when I wake
up from sleep on a PIC12C509 PICmicro MCU?, above, for an example of how to do this.

Context Switching
How do I know when I'm context switching in Salvo?
All Salvo with an "OS_" prefix (e.g. OS_Yield())cause a context
switch. Context switches do not occur anywhere else in Salvo.

Why can't I context switch from something other than the
task level?
Because Salvo is designed to run on processors with minimal
amounts of RAM memory and no general-purpose stack, it does
not presume that a stack is available to store context-switching information. Without it, there's no way to store the return addresses
for the function calls nested within the task. If you were to contextswitch from a function nested within a task, upon returning from
that function the processor's program counter would be undefined.

Why does Salvo use macros to do context switching?
Context switching in Salvo is an inherently in-line action, and is
not generally conducive to the use of functions or subroutines. The
context-switching macros use function calls wherever possible to
keep code size to a minimum.

Salvo User Manual

Chapter 6 • Frequently Asked Questions (FAQ)

213

Can I context switch in more than one place per task?
There is no limit on how many context switches you write into a
given task.
For example, you could add several unconditional context switches
(OS_Yield()) to the main loop of a low-priority yet long (in terms
of lines of code) task. This way, if a higher-priority task needs to
run, it will have several opportunities to run for each full path
taken through the low-priority task's loop. For example,
void TaskLong( void )
{
while (1) {
...
/* give other tasks a chance to run.
OS_Yield(TaskLong1);
...
/* let's take a break to let higher/* tasks run.
OS_Yield(TaskLong2);
...
/* we're about to hog the processor for a
/* while, so let's yield in case another
/* more important task is ready to run.
OS_Yield(TaskLong3);
....
}
}

*/

*/
*/

*/
*/
*/

When must I use context-switching labels?
Prior to Salvo v4, Salvo required context-switching labels.
Unless otherwise specified for a particular target and compiler,
Salvo no longer requires context-switching labels.
Use of context-switching labels where they are not required will
generate an error message.

Tasks & Events
What are taskIDs?
TaskIDs are just integers used to refer to a task. They are numbered from 1 to OSTASKS. There's a one-to-one mapping between a
task's taskID and the task control block (tcb) assigned to it. You'll

214

Chapter 6 • Frequently Asked Questions (FAQ)

Salvo User Manual

rarely use taskIDs when writing your Salvo application. Instead,
Salvo uses pointers as handles to tasks. For example, the pointer to
the task with taskID 3 is OSTCBP(3).

Does it matter which taskID I assign to a particular task?
No. The only rule to follow is that each task needs its own, unique
taskID, and hence its own, unique tcb. A task's priority is independent of its taskID.

Is there an idle task in Salvo?
Salvo has a built-in facility for automatically calling a user-defined
function when the system is idling. OSIdlingHook() is enabled via
the configuration option OSENABLE_IDLING_HOOK.
If you prefer, you can create your own idle task with the lowest
possible priority (OSLOWEST_PRIO). Be sure that no other tasks
have this priority. Then, your idle task will run whenever none of
the other tasks are eligible.
You can context-switch inside an idle task of your own making,
but you cannot context-switch inside the built-in idling hook function. This is an important distinction. Which one you use will depend on what sort of functionality you want to occur when the
system is idling. The scheduler must perform a context switch each
time the idle task runs. Overall performance is better when using
the idling hook function, since no real context switch is performed
when calling OSIdlingHook().

How can I monitor the tasks in my application?
Salvo provides a task monitor function that you can link to your
application. The monitor is intended to work with a simple ASCII
terminal program. The monitor can display the status of all tasks
and events, and can control tasks. See OSRpt() for more information.

What exactly happens in the scheduler?
Salvo's scheduler OSSched() performs three major functions each
time it is called. First, it processes the event queue, if events are in
use. This means that for every event that had a waiting task when it

Salvo User Manual

Chapter 6 • Frequently Asked Questions (FAQ)

215

was signaled, the scheduler makes that task eligible to run. Next, it
processes the delay queue. Any tasks that timed out while being
delayed or waiting with a timeout will be made eligible to run. Finally, the scheduler runs the most eligible task. Interrupts are enabled and disabled at various times in the scheduler.

What about reentrant code and Salvo?
An RTOS requires a call…return stack, but Salvo works without a
general-purpose stack. Therefore none of its functions are reentrant. In order to avoid problems with reentrancy, 1) do not directly
call a task from anywhere within your program – let the scheduler
handle it, and 2) carefully observe the restrictions on calling Salvo
services from ISRs. By explicitly controlling interrupts and/or setting certain configuration parameters, you can call certain Salvo
services from mainline, task and interrupt levels all in a single application.

What are "implicit" and "explicit" OS task functions?
The explicit OS functions require that you specify a task number as
a parameter. A good example is OSCreateTask(), which creates
and starts a specified task. Explicit OS task function names contain
the word "Task". Implicit OS functions like OS_Delay() operate
only on the current task, i.e. the task that is running. Once a task is
running, most or all of the OS functions called are likely to be implicit ones, i.e. they operate on the current task.

How do I setup an infinite loop in a task?
A simple way in C is to use the following syntax:
void Task ( void )
{
/* initialization code. */
...
while (1) {
/* body of task. */
...
}
}

Note that somewhere in the for loop the task needs to return to the
scheduler (e.g. via OS_Yield()) to make the highest-priority eligible task run.

216

Chapter 6 • Frequently Asked Questions (FAQ)

Salvo User Manual

Why must tasks use static local variables?
Static variables are assigned their own unique address in RAM,
and may not be visible to other tasks. By declaring a task's variables as static you are guaranteeing that they will remain unchanged while the task is not running. This is the only way to
preserve the variable from one context switch to the next. If the
variable were not static (i.e. if it were an auto variable) it's likely
that it would be changed by other tasks, functions or ISRs, and
unpredictably.
It is safe to use auto variables in tasks61 as long as the task does not
require that the value of the variable be maintained in the task from
one context switch to the next. For example, if a simple for() loop
is used to repeatedly call a function, and then the task context
switches, as long as the loop index is initialized each time, it
should not pose a problem.
int i;
while (1) {
for (i = 0; i < 5; i++) {
WriteControlReg(0x55);
WriteControlReg(0xAA);
}
...
OS_Yield(here);
}

Doesn't using static local variables take more memory
than with other RTOSes?
No, it doesn't. The RAM required for saving persistent local variables in a Salvo application is the same as the RAM required to
save auto local variables in conventional RTOSes.62 In each situation, RAM must be permanently63 allocated to the variable.

Can tasks share the same priority?
When Salvo is configured to use queues, there's no reason why
more than one task cannot share the same priority. Tasks of equal
61
62

63

Salvo User Manual

Some implementations (e.g. Salvo on x86-based machines with the Mix
Software Power C compiler) do not permit the use of auto variables.
In a conventional RTOS, local auto variables are by their very nature stored
on the stack, or in the task's context save area (if the local auto variable was in
a register to begin with).
I.e. as long as the task is active.

Chapter 6 • Frequently Asked Questions (FAQ)

217

priority will round-robin (execute one after the another in a circular
queue) whenever they are the highest-priority eligible tasks. However, in many applications it is more efficient to give each task a
unique priority.
When Salvo is configured to use arrays, each task must have a
unique priority.
If an idle task is used in your Salvo application, it should be the
only task with the lowest priority (OSLOWEST_PRIO). Other tasks
OSHIGHEST_PRIO
should
use
priorities
between
and
OSLOWEST_PRIO-1.

Can I have multiple instances of the same task?
Yes. A Salvo task is essentially an address in your program at
which your application will resume execution when the scheduler
sends it there. You can configure two or more Salvo tasks to point
to the same place in your program. For example,
void TaskDelayFiveTicks( void )
{
while (1) {
OS_Delay(5, here);
}
}
...
OSCreateTask(TaskDelayFiveTicks, OSTCBP(5), 8);
OSCreateTask(TaskDelayFiveTicks, OSTCBP(6), 9);
...
while (1) {
OSSched();
}

will create two Salvo tasks with different priorities, each of which
delays itself for 5 system ticks over and over. Note that without
reentrancy, the utility of multiple instances of the same task is limited. Note also that all static variables in the task function will be
"shared" by each instance of the Salvo task.

Does the order in which I start tasks matter?
No. To start a task, it must have been created first. Creating a task
initializes the fields in its task control block, but leaves it ineligible
to run. Starting a task makes it eligible and places it in the eligible
queue. Tasks are positioned within the eligible queue based on

218

Chapter 6 • Frequently Asked Questions (FAQ)

Salvo User Manual

their priority. A task will first execute based on its priority, not on
when it was started.
If you start several tasks of equal priority together, they will begin
executing in the order they were started. If they remain at these
same priorities, they will continue to round robin.
By using OSSetPrio() or OS_Prio() to change the current task's
priority you can control the order in which tasks execute.

How can I reduce code size when starting tasks?
You may face this question of you are explicitly starting tasks
separately from when they are created (by using
OSDONT_START_TASK with OSCreateTask()). Each task is referred
to by its tcb pointer, which is specified in the call to OSCreateTask(). You can reduce the number of calls to OSStartTask()
by placing it in a loop in order to start multiple tasks at once, e.g.
char i;
...
for (i = 1; i <= OSTASKS; i++) {
OSStartTask(OSTCBP(i));
}

will start all of your tasks with just a single call to OSStartTask(),
thereby reducing the size of your application.

What is the difference between a delayed task and a
waiting task?
A task that is delayed is simply inactive for a specified number of
system ticks. It will then rejoin the eligible tasks when the delay
timer has expired. A task that is waiting will wait until an event
occurs. If the event never occurs, then the task is never made eligible again, unless a timeout was specified when the task was made
to wait. If the timeout timer expires before the event occurs, the
task is made eligible and carries with it a flag that indicates that a
timeout occurred. Your application program can handle this flag at
the task level.
In order to delay tasks, OSTimer() must be called at the system
tick rate from your application. This run-time overhead is independent of the number of tasks still delayed. Waiting tasks, on the

Salvo User Manual

Chapter 6 • Frequently Asked Questions (FAQ)

219

other hand, do not require the existence of OSTimer(),64 and require no processing power whatsoever while they are waiting.

Can I create a task to immediately wait an event?
Not with a single service call. A task can only wait an event by
calling OS_WaitXyz() while running. One way to start your application with a bunch of tasks waiting for event(s) is to create them
with the highest priority (guaranteeing that they will run before all
others) and create the events with initial values of 0. When each
task runs, have it change its priority to the desired run-time priority
with OSSetPrio() (not OS_Prio()!), and have it wait the event.
When the events are signaled, the waiting tasks will run.

I started a task but it never ran. Why?
You may have incorrectly specified one or more parameters when
calling the relevant Salvo services – check the function return
codes to see if any errors were reported. A common error when
using the freeware libraries is to create a task with a tcb pointer
that exceeds OSTCBP(OSTASKS).
If Salvo was initialized via OSInit(), the task was successfully
created and started via OSCreateTask(), the scheduler OSSched()
is active, and no other task has destroyed or stopped the task in
question, then it probably had a lower priority than the other tasks
running, and hence never ran. Try elevating the task's priority. Use
the Salvo monitor OSRpt() to view the current status of all the
tasks.

What happens if I forget to loop in my task?
You'll get some rather odd results. If your application doesn't crash
immediately, the original task may leave its own function and continue through your code until it reaches a context switch, and will
thereafter resume execution after that context switch, which will be
part of another task! So you may have inadvertently created a second instance of another task by failing to keep execution within the
intended task.

64

220

Unless they were made to wait with a timeout.

Chapter 6 • Frequently Asked Questions (FAQ)

Salvo User Manual

Why did my low-priority run-time tasks start running
before my high-priority startup task completed?
It's common to use delays in a startup task (responsible for configuring peripherals like LCDs, for instance). The other tasks ran because the high-priority startup task was delayed. Regardless of its
priority, whenever a task is delayed or waiting for an event, other
lower-priority tasks are free to run.
If your application needs a startup task that uses delays, and if it's
imperative that no other tasks run before the startup task is complete, then one elegant method is to initially create all the tasks but
only start the startup task, and then start the other tasks at the end
of the startup task. You can even "reuse" the startup task's tcb by
destroying the startup task and creating a new task with the same
tcb.

When I signaled a waiting task, it took much longer than
the context switching time to run. Why?
A task that is made eligible will only run when it becomes the
highest-priority eligible task. Other eligible tasks with higher priorities will run first, and will continue to run if they remain eligible. Also, interrupt service routines (ISRs) have the highest
priorities of all.

Can I destroy a task and (re-) create a new one in its
place?
Yes. As long as a task is destroyed, a new one can be created in its
place. A Salvo task is really just a means of executing a function in
ROM. Creating and starting a task allows that function to execute
along with the other tasks in a priority-based scheme.
Before destroying any task you must ensure that:
•
•
•

• it is not waiting for any event,
• is it in the delayed queue and
• has not acquired any resources that other tasks
might need.

It is up to you to ensure that the above conditions are met. If you
are to use OSDestroy() in a particular task that accesses resources,
you must release all resources before destroying the task. Failing to
do so would block any other tasks waiting for the resource previSalvo User Manual

Chapter 6 • Frequently Asked Questions (FAQ)

221

ously owned by the now-destroyed task. Only if those tasks were
waiting with a timeout would they ever run again.

Can more than one task wait on an event?
Yes. Up to all of the defined tasks can wait on a single event
simultaneously.

Does Salvo preserve the order in which events occur?
Yes.

Can a task wait on more than one event at a time?
Yes, but not simultaneously. At any time a task can only be waiting on a single event. It can wait on more than one event sequentially (e.g. first on one, then on the other), but not simultaneously.
In this example, a task first waits for an error message (a string),
then waits for a resource (an LCD display) to become available.
Once it receives the error message and obtains exclusive access to
the display, it writes the message to the display, waits one second,
releases the display for others to use, and then returns to waiting
for another message.
void TaskShowErrMsg( void )
{
static OStypeMsgP msgP;
static OStypeMsgP msgP2;
while (1) {
OS_WaitMsg(MSG_ERROR_STRING_P, &msgP,
OSNO_TIMEOUT);
OS_WaitMsg(MSG_LCD_DISPLAY_P, &msgP2,
OSNO_TIMEOUT);
DispStringOnLCD((char *) msgP);
OS_Delay(ONE_SECOND);
OSSignalMsg(MSG_LCD_DISPLAY_P, (OStypeMsgP)
1);
}
}

By first acquiring the display resource and later releasing it,65 the
user is guaranteed to see the error message for at least one second.
The error message will remain on the LCD display until this or an65

222

In this example, MSG_LCD_DISPLAY is being used as a binary semaphore.

Chapter 6 • Frequently Asked Questions (FAQ)

Salvo User Manual

other task obtains the LCD display resource via OS_WaitMsg(MSG_LCD_DISPLAY, ...) and writes a new string to it via
DispStringOnLCD().

How can I implement event flags?
Event flags are used to synchronize tasks to the occurrence of multiple events. Two types of synchronization are possible – conjunctive synchronization, where the task can only proceed once all of
the events it's waiting on have occurred (i.e. logical AND), and
disjunctive synchronization, where the task can proceed as soon as
any of the events it's waiting on has occurred (i.e. logical OR).
You can use Salvo's built-in event flag (eFlag) services(this is the
preferred method), or you can implement simple flags using binary
semaphores. See the Reference chapter in the Salvo User Manual
for more info on Salvo's event flag services.
To implement conjunctive synchronization (i.e. the logical AND of
multiple events) using binary semaphores, the task must wait on
multiple events in sequential order. In the example below, the task
waits for the occurrence of all three events (signified by binary
semaphores) before proceeding.
…
OS_WaitBinSem(BINSEM1_P, OSNO_TIMEOUT,
WaitForSync1);
OS_WaitBinSem(BINSEM2_P, OSNO_TIMEOUT,
WaitForSync2);
OS_WaitBinSem(BINSEM3_P, OSNO_TIMEOUT,
WaitForSync3);
…

The order in which the events occur (i.e. when each event is signaled) is unimportant. As long as the task is the highest-priority
task waiting on each event, once all of the events have been signaled the task will proceed.
To implement disjunctive synchronization (i.e. the logical OR of
multiple events) using binary semaphores, the task must wait on a
single event that can be signaled from multiple locations in your
application.
…
OS_WaitBinSem(BINSEM4_P, OSNO_TIMEOUT,
WaitForSync4);
…

Salvo User Manual

Chapter 6 • Frequently Asked Questions (FAQ)

223

In this case the task can proceed as soon as any part of your application has signaled the event. Subsequent event signaling will not
affect the task's execution until the next time it waits on the event.

What happens when a task times out waiting for an
event?
If the task does not acquire the resource within the timeout period,
it will be removed from the event queue (and the waiting queue)
and made eligible to run again. When it runs, a timeout flag will be
available at the task level to indicate that a timeout occurred. The
Salvo user service OSTimedOut() returns TRUE when this flag is
set, FALSE otherwise. The timeout flag is cleared when the task returns to the scheduler.
If a task times out waiting for an event, even if the event subsequently occurs before the task runs again, the timeout flag will remain until the task runs and returns to the scheduler. The event will
also remain until a task waits on it.

Why is my high-priority task stuck waiting, while other
low-priority tasks are running?
The unavailability of an event always takes precedence over a
task's priority. Therefore, regardless of its priority, a task that waits
on an event that is not available will become a waiting task, and it
will remain a waiting task until either a) the event happens and the
task is the highest-priority task waiting for the event, or b) a timeout (if specified) occurs.
This situation may simply be due to the fact that the event never
occurred, or it may be due to priority inversion.

When an event occurs and there are tasks waiting for it,
which task(s) become eligible?
The highest-priority waiting task becomes eligible. Only a single
task will become eligible, regardless of how many tasks of equal
priority are waiting for the event. All of Salvo's queues are priority
queues. Additionally, tasks of equal priorities are inserted into the
priority queues (i.e. they are enqueued) on a FIFO basis. For example, if a task of the highest priority is enqueued into a priority
queue that already contains a task of highest priority, the task being
enqueued will be enqueued after the existing task. In other words,

224

Chapter 6 • Frequently Asked Questions (FAQ)

Salvo User Manual

the first task to be enqueued with a particular priority will be the
first task to be dequeued when tasks of that particular priority
reach the head of the queue.
There is one exception to this behavior – namely, event flags.
When an event flag is signaled, all the tasks waiting on said event
flag will be made eligible.

How can I tell if a task timed out waiting for an event?
The macro OSTimedOut() is provided to detect timeouts. It returns
TRUE if the current task has timed out waiting for an event, and
FALSE otherwise. OSTimedOut() is only valid while the current
task is running.

Can I create an event from inside a task?
Yes. You can create an event or a task anywhere in your code, as
long as you have previously allocated the required memory at
compile time. Keep in mind that operating on an event that is not
yet defined can cause unpredictable behavior. For example, suppose you have two tasks, one to create and signal a resource, and
one that waits for it:
void Task1( void )
{
OSCreateSem(SEM1_P, 0);

/* init to 0 */

while (1) {
...
OSSignalSem(SEM1_P);
...
}
}
void Task2( void )
{
while (1) {
...
OS_WaitSem(SEM1_P, OSNO_TIMEOUT);
...
}
}

If your main() looks like this:
int main( void )
{
OSInit();

Salvo User Manual

Chapter 6 • Frequently Asked Questions (FAQ)

225

OSCreateTask(Task1, TASK1_P, 3);
OSCreateTask(Task2, TASK2_P, 1);
while (1) {
OSSched();
}
}

you will have unpredictable results because Task2() will attempt
to wait the semaphore SEM1 before Task1() can create it. That's
because Task2() has a higher priority than Task1(), and will
therefore run first when the OSSched() starts dispatching tasks.
To avoid this, you can either ensure that the task that creates the
resource has a higher priority than any task that uses it, or you can
create the resource before beginning multitasking via OSSched().
If you plan on creating events or tasks from within an ISR, you
must configure salvocfg.h appropriately to avoid interruptrelated issues.

What kind of information can I pass to a task via a
message?
Messages are application-specific – that is, a message contains
whatever you want it to contain. Examples include characters,
numbers, strings, structures and pointers. Messages are passed via
pointer, and the default type for a Salvo message pointer is OStypeMsgP, which is usually a void pointer. Since a void pointer can
point to anything, in order to obtain the information in the
message, you'll need to typecast the pointer's contents to the message's inherent type.
The only restriction on Salvo messages is that all the messages in a
particular message queue should point to the same type of information.

My application uses messages and binary semaphores. Is
there any way to make the Salvo code smaller?
(OStypeMsgP) 0 and
Yes, use messages with values of
(OStypeMsgP) 1 instead of binary semaphores with values of 0
and 1, respectively. This way you can use OSCreateMsg(), OSSignalMsg() and OS_WaitMsg() exclusively.

226

Chapter 6 • Frequently Asked Questions (FAQ)

Salvo User Manual

Why did RAM requirements increase substantially when I
enabled message queues?
Each message queue requires both an ecb and a message queue
control block (mqcb) of fixed size. The number of ecbs and mqcbs
are determined by OSEVENTS and OSMESSAGE_QUEUES, respectively.
Additionally, each message queue also requires RAM for the actual queue. Message queues are the only events that require this
extra memory.

Can I signal an event from outside a task?
Yes. Events can be signaled and created from mainline code (e.g.
from within tasks, functions or inside main()), and from within
interrupts. The default Salvo configuration expects events to be
created and signaled from mainline code. In order to create or signal tasks from interrupts and/or interrupts and mainline code, the
configuration parameters appropriate to the event's user service
(e.g. OSSignalMsg()) must be defined.

When I signal a message that has more than one task
waiting for it, why does only one task become eligible?
A task waits for a message when the corresponding mailbox is
empty. Signaling a message will fill the mailbox. The mailbox remains full (i.e. contains a single message) until the task that was
waiting on the message runs, i.e. until the task becomes the highest-priority task and is dispatched by the scheduler. Put another
way, signaling a message fills the mailbox, and running the task
that's waiting on the message empties it. If the task never becomes
eligible to run, the mailbox will remain full, and signaling it with a
message will result in an error.

I'm using a message event to pass a character variable to
a waiting task, but I don't get the right data when I
dereference the pointer. What's going on?
Let's say you're trying to pass a character to a task via a message.
To send the message you might write:
char tempVar;
...
tempVar = '!';
OSSignalMsg(MSG_CHAR_TO_TASK_P,
(OStypeMsgP) &tempVar);

Salvo User Manual

Chapter 6 • Frequently Asked Questions (FAQ)

227

...

to send a '!' to the task that's waiting for the message
MSG_CHAR_TO_TASK, which might look like this:
static OStypeMsgP msgP;
static char msgReceived;
while (1) {
OS_WaitMsg(&msgP, MSG_CHAR_TO_TASK_P,
OSNO_TIMEOUT);
msgReceived = *(char *) msgP;
switch (msgReceived) {
case '!':
printf("Received '!'\n");
break;
default:
printf("Received anything but '!'\n");
}
}

Because tasks obtain messages via pointers, the element referenced
by the message pointer must remain unchanged until
OS_WaitMsg() succeeds. In the example above, if the global or
auto variable tempVar is assigned another value before the waiting
task has a chance to obtain the message, the waiting task will receive a message quite different from what was intended. A safer
solution would be to signal the message with a pointer to a character constant:
const char BANG = '!';
...
OSSignalMsg(MSG_CHAR_TO_TASK_P,
(OStypeMsgP) &BANG);
...

This way, no matter how long it takes for the receiving task to run
and obtain the message, it is guaranteed to be the '!' character.

What happens when there are no tasks in the eligible
queue?
The scheduler loops in a very tight loop, with interrupts enabled,
when there are no tasks eligible to run. As soon as a task is made
eligible, either through the actions of OSTimer() or an interrupt
signaling an event, the scheduler will cause it to run.

228

Chapter 6 • Frequently Asked Questions (FAQ)

Salvo User Manual

In what order do messages leave a message queue?
Each message queue operates on a FIFO (first-in, first-out) basis.

What happens if an event is signaled before any task
starts to wait it? Will the event get lost or it will be
processed after task starts to wait it?
The event will not be lost, and the highest-priority task to wait the
event will get it, i.e. will remain eligible after OS_WaitXyz() instead of going to the waiting state.

What happens if an event is signaled several times before
waiting task gets a chance to run and process that
event? Will the last one signal be processed and
previous lost? Or the first will be processed and the
following signals lost?
That depends on the event – if it's a binary semaphore or a message, all further signaling results in OSSignalXyz() returning an
error code, because the event is "full". The first event to be signaled will be processed, and subsequent ones will be lost. In the
case of a counting semaphore, the value is simply incremented. In
the case of a message queue, additional messages are enqueued
until the queue is full. With these events, once the event is "full",
subsequent signals will be lost.

What is more important to create first, an event or the
task that waits it? Does the order of creation matter?
The order of creation doesn't matter. But when a task waits an
event, the event must exist before the task runs.

What if I don't need one event anymore and want to use
its slot for another event? Can I destroy event?
Absolutely! For example, you can destroy a binary semaphore and
create a counting semaphore in its place by calling OSCreateSem()
with the ecb you previously used for the binary semaphore. You
should only do this if you know that there aren't any tasks waiting
the binary semaphore.

Salvo User Manual

Chapter 6 • Frequently Asked Questions (FAQ)

229

Can I use messages or message queues to pass raw data
between tasks?
Yes, with some restrictions. With messages, a null message pointer
is treated as an empty message, and a task will wait an empty message forever. Therefore only non-zero raw data can be passed via
messages. Message queues are different in that a task will wait a
message queue indefinitely if there are no messages in it. Therefore null message pointers are allowed in message queues, and raw
data of any value can be passed from one task to another using a
message queue. In this case, the message queue acts like a FIFO
buffer.
If you want to pass null-pointer messages to a task, use a message
queue of size 1.

How can I test if there's room for additional messages in a
message queue without signaling the message queue?
Use OSMsgQEmpty(). If the message queue is full – i.e. there is no
room for an additional message in the message queue –
OSMsgQEmpty() returns 0 (FALSE). If there is room,
OSMsgQEmpty() returns the number of available slots in the message queue.

Interrupts
Why does Salvo disable all interrupts during a critical
section of code?
It is common practice in an RTOS to disable interrupts during a
critical section of code. To maintain system performance, interrupts should be disabled for the shortest times possible. However,
it's imperative that while an RTOS performs certain critical functions, it must not be interrupted for fear of certain things in the
RTOS being corrupted.
The major sources of corruption due to interference from an interrupt are access to a shared resource, and the operation of nonreentrant functions. Salvo must guarantee that while performing
certain operations on its data structures (e.g. changing an event
control block), no access (read or write) from any other part of the
application is allowed. Salvo functions that access the data struc-

230

Chapter 6 • Frequently Asked Questions (FAQ)

Salvo User Manual

tures include OSTimer(), which is normally called from within a
periodic interrupt, and OSSignalMsg(), which might be called
from an entirely different interrupt.
Since Salvo services work without a general-purpose stack, certain
steps must be taken to prevent data corruption from interrupts. Use
the OSCALL_Xyz() configuration parameters if you want to be able
to call a particular Salvo service (e.g. OSSignalSem()) from both
main-line code and an ISR.

I'm concerned about interrupt latency. Can I modify Salvo
to disable only certain interrupts during critical sections
of code?
Yes, and it will require Salvo Pro. The approach to take is to redefine Salvo's OSEi() and OSDi() to only disable those interrupts
that are associated with calls to Salvo services, and leave other interrupts alone. The implementation will differ from one target to
another based on the target's interrupt control scheme, its interrupt
vectors, its interrupt priorities, and whether Salvo controls interrupts via functions, macros, or through compiler extensions.
As an example, a Salvo customer on the PIC18 needed essentially
zero jitter so that his interrupt-driven DSP algorithm ran at exactly
1280Hz. So, the Salvo solution for that particular chip (which has
two interrupt priority levels) was to put the DSP stuff on the highpriority interrupt, and the rest on the low-priority interrupt, and
configure Salvo to only disable low-priority interrupts in its critical
sections. This, it turns out, was very easy for that particular target
and compiler – just a small header file to build a custom library
with the desired behavior. 5 minutes' work.

How big are the Salvo functions I might call from within
an interrupt?
and OSSignalXyz() are the Salvo services you might
call from an interrupt. They are all quite small and fast, and have
no nested subroutines. While it varies among different target processors, these services will in many cases be faster than the actual
interrupt save and restore.

OSTimer()

Salvo User Manual

Chapter 6 • Frequently Asked Questions (FAQ)

231

Why did my interrupt service routine grow and become
slower when I added a call to OSTimer()?
Some compilers assume the worst case with regard to register
saves and restores when an external function is called from within
an interrupt routine. As a result, the compiler may add a large
amount of code to save and restore registers or temporary registers
to preserve the program's context during an interrupt. Since it's always a good idea to have as fast an interrupt routine as possible,
one solution is to include the necessary Salvo files66 in your interrupt routine's source code instead of linking to the OSTimer() and
related services as external functions (e.g. through the Salvo library). By including those Salvo files which completely define the
necessary call chains for OSTimer() your compiler can "see" exactly which registers and temporary registers must be saved, instead of assuming the worst case and saving all of them.
Another option is to in-line OSTimer(). For more information, see
the OSUSE_INLINE_OSTIMER configuration option.

My application can't afford the overhead of signaling from
an ISR. How can I get around this problem?
Ideally you should signal from an ISR if the event that causes the
signaling is an interrupt. If this is not possible, in your ISR you can
set a simple flag (i.e. a bit) in a global variable, and then test-andclear it67 in your main loop. If the flag is set, you then call the appropriate signaling service prior to calling OSSched(), like this:
while (1) {
disable_interrupts();
localFlag = flag;
flag = 0;
enable_interrupts();
if (localFlag) {
OSSignalBinSem(binSemP);
}
OSSched();
}

This disadvantage of this approach is that it does not preserve the
order in which events occur, whereas signaling from an ISR will
preserve that order. This may affect the behavior of complex systems.
66
67

232

timer.c.
Interrupts should be disabled while you test and clear the flag.

Chapter 6 • Frequently Asked Questions (FAQ)

Salvo User Manual

Building Projects
What warning level should I use when building Salvo
projects?
Use the compiler's default warning level. More pedantic warning
levels may generate warnings that in some cases cannot be
avoided, and thus cause unnecessary confusion.

What optimization level should I use when building Salvo
projects?
Use the maximum optimization unless suggested otherwise.

Miscellaneous
Can Salvo run on a 12-bit PICmicro with only a 2-level
call…return stack?
Yes. Certain compilers (e.g. HI-TECH PICC) circumvent this limitation by converting all function calls into long jumps through table lookup. Therefore function calls require some additional
overhead and ROM, but call graphs of arbitrary depth are possible.

Will Salvo change my approach to embedded
programming?
Maybe. Stranger things have happened … ☺

Salvo User Manual

Chapter 6 • Frequently Asked Questions (FAQ)

233

234

Chapter 6 • Frequently Asked Questions (FAQ)

Salvo User Manual

Chapter 7 • Reference

Run-Time Architecture
In order to run properly, every Salvo application must follow three
basic rules. Failure to follow these rules may result in an application that compiles successfully, but does not run as expected.
These rules are explained below.

Rule #1: Every Task Needs a Context Switch
Each Salvo task must have at least one context switch.

Tip In Salvo, context switches are denoted by a "OS_" prefix.
Functions with just an "OS" prefix (e.g. OSSignalBinSem()) are
not context switches and may usually be called from anywhere in
the Salvo application.
void HappyTask ( void )
{
while (1) {
…
OS_Delay(10); // Return here in 10 ticks.
…
}
}
Listing 32: Task with a Proper Context Switch

In Listing 32 above, HappyTask() uses a single context switch (via
OS_Delay()) to yield to the scheduler during its delay of 10 system ticks. During the delay period, the task is in the delayed state,
and the application is free to run other, eligible tasks. Whenever
the scheduler dispatches HappyTask(), HappyTask() will run the
code inside its infinite loop, returning to the scheduler via
OS_Delay().

Note The requirement of having at least one context-switch per
task is a general one for cooperative RTOSes and is not specific to
Salvo.

Salvo User Manual

235

Note The number of context switches a Salvo task can have is
limited only by available program memory.
void ForlornTask ( void )
{
putchar('!'); // Bad – untimely exit from task.
}
void StuckTask ( void )
{
while (1) {
MyFn(); // Bad – never returns to scheduler.
}
}
Listing 33: Tasks that Fail to Context Switch

In Listing 33 above, ForlornTask() has no context switch. As a
result, when the scheduler dispatches that task, it will call
putchar() once and then the application will continue with whatever code lies in program memory after ForlornTask().68 ForlornTask() will not yield to the scheduler immediately after
MyFn() is executed. Therefore the application's behavior is unpredictable.
Also in Listing 33 above, once the scheduler dispatches StuckTask(), it will call MyFn() indefinitely, and will never yield back
to the scheduler. While this behavior is predictable, it is not desirable, as all multitasking will stop.

Rule #2: Context Switches May Only Occur in Tasks
The only valid location for a Salvo context switch is within a task
(see In Listing 32, above).

68

236

It is likely to continue "into" StuckTask() if and only if the linker has placed
StuckTask() immediately after ForlornTask() in memory.

Chapter 7 • Reference

Salvo User Manual

void StuckTask ( void )
{
while (1) {
MyFn(); // Bad – where's the context switch?
}
}
void MyFn ( void )
{
DoThings();
OS_Yield(); // Bad – not allowed inside a
// called function.
}
Listing 34: Incorrectly Context-Switching Outside of a
Task

In Listing 34 above, the scheduler will dispatch StuckTask() and
the task will, in turn, call MyFn(). After MyFn() calls DoThings(),
it will attempt to yield to the scheduler via OS_Yield(). This will
fail, as Salvo's context-switcher is not designed for yielding back
to the scheduler at any call…return level other than the task's. The
run-time behavior when violating this rule is unpredictable.
In C, the ability to context-switch outside of a task, at arbitrary
call…return stack levels, requires considerable RAM for saving
call…return addresses, function parameters and local (auto) variables. Salvo is designed expressly to minimize RAM requirements,
and therefore does not support context-switching outside of tasks.

Note Context switches may not occur in mainline (background)
code outside of tasks, nor in interrupt service routines (ISRs).

Rule #3: Persistent Local Variables Must be Declared as
Static
Every local variable used in a Salvo task in a manner that requires
persistence across context switches must be declared as static.
void TaskLowPrio ( void )
{
static int i;

while (1) {
i = 20000;
do {
LED_PORT &= ~LED_PORT_MASK;
LED_PORT |= ((i >> 8) & LED_PORT_MASK);
OS_Delay(1);

Salvo User Manual

Chapter 7 • Reference

237

} while (--i);
}
}
Listing 35: Task Using Persistent Local Variable

In Listing 35 above, TaskLowPrio()outputs the upper 8 bits of the
loop counter i to eight LEDs every system tick while decrementing i. If i were not declared as static, i 's value would be unpredictable and so would be the output to the LED port.
Declaring local variables that require persistence as static is necessary because Salvo's context switcher performs a minimal context save that does not include local variables. Other tasks,
functions and ISRs may use the memory allocated to the local
variable for their own purposes when the task is not running,
changing it in unpredictable ways.
With care, local variables can be used as auto variables in Salvo
tasks. Whenever a local variable is initialized and fully used before
the next context switch, it can be declared as a simple local (auto)
variable instead of a static one.

238

Chapter 7 • Reference

Salvo User Manual

void TaskCountElements( void )
{
char i;
element * p;

while (1) {
OS_WaitBinSem(BINSEM_COUNT_LIST);
i = 0;
p = headP;
while (1) {
if (p!=0) {
i++;
p = p->nextP;
}
else {
break;
}
}
LCDWrite("The list has %d elements.\n", i);
…
OS_Delay(delay);
…
}
}
Listing 36: Task Using Auto Local Variables

In Listing 36 above, i and p are used as local (auto) variables to
traverse a linked list and count the number of objects therein. Afterwards the result is displayed on an LCD, and the task continues.

Note When in doubt, declare local variables as static.

Salvo User Manual

Chapter 7 • Reference

239

User Services
This section describes the Salvo user services that you will use to
build your multitasking application. Each user service description
includes information on:
•
•
•
•
•
•
•
•
•
•
•
•
•

• the service type (function or macro),
• the service prototype (for a function) or
declaration (for a macro),
• where the service is callable from (the
foreground, the background or within a task),
• which Salvo C source or include files contain
the source code for the service,
• which configuration options (if any) enable the
service,
• which configuration options (if any) affect the
service (i.e. alter its execution speed or code
size),
• a description of what the service does,
• the parameter(s) (if any) expected by the
service call,
• the service's return value(s) (if any),
• the service's stack usage (if any), in terms of
levels of call…return stack used,69
• notes particular to the service,
• related services and
• an example using the service.

Salvo functions comprise the majority of the user services you will
call from C in your application. Salvo user services that do not result in a context switch are implemented as functions and are prefixed by just "OS".
Salvo uses macros wherever a context-switch is implicit in the action being performed (e.g. delaying for a number of ticks, via
OS_Delay()). All of Salvo's services that result in a context-switch
are implemented via macros and are prefixed by "OS_".

69

240

For call…return stack depth calculations, OSUSE_INSELIG_MACRO is assumed
to be the default value, TRUE. If FALSE, those services that cause a task to be
placed in the eligible, delay and/or event queue(s) will consume an additional
call…return stack level. Stack usage does not take into account any library
functions invoked by the compiler.

Chapter 7 • Reference

Salvo User Manual

Note Salvo context-switching services are implemented as macros and do not have return values.
It is important not to confuse a Salvo macro with its underlying
function. For instance, the OS_Delay() macro will cause the current task to delay for the specified number of system ticks. On the
other hand, using the OSDelay() function directly will have unpredictable results, and your application may crash as a result. These
underlying functions are intended for use only within a Salvo
macro, and are therefore not documented in this section. For the
curious, they can be viewed in the Salvo source code.

Note Some services (e.g.

and OSSignalXyz())
can be either a macro that invokes a function, or a standalone function, depending on OSCOMBINE_EVENT_SERVICES. In all cases the
argument list and return value and type are identical.
OSCreateXyz()

When compiling and linking Salvo into your application, the size
and speed of many user services is dependent on the chosen configuration. By referring to the detailed descriptions of each user
service below and inspecting the output of your compiler, you may
be able to correlate changes in the size (in instructions) and/or
speed (in cycles) of the Salvo services in your application against
changes you've made to your compile-time configuration. Remember that each time you change the configuration options, you must
recompile all of Salvo before linking it into your application.

Note The foreground is the interrupt level of your application.
The background is the non-interrupt level, and includes main(),
Salvo tasks and all other functions not called via interrupts.

Salvo User Manual

Chapter 7 • Reference

241

This page is intentionally left blank.

242

Chapter 7 • Reference

Salvo User Manual

OS_Delay(): Delay the Current Task and Context-switch
Type:
Declaration:

Macro (invokes OSDelay())
OS_Delay (
OStypeDelay delay);

Callable from:
Contained in:
Enabled by:
Affected by:
Description:

Task only

Parameters:
Returns:
Stack Usage:
Notes

salvompt.h
OSBYTES_OF_DELAY
OSENABLE_STACK_CHECKING, OSLOGGING
Delay the current task by the amount
specified. Return to scheduler.
delay: an integer (>=0) specifying the
desired delay in system ticks.
–
2

A delay of 0 will stop the current task. A non-zero delay will delay70 the current task by the number of ticks specified relative to
the current value of the system timer.
Do not call OS_Delay() from within an ISR!
In order to use delays, Salvo's timer must be installed.
Long delays can be accomplished in a variety of ways – See
"Timer and Timing" in Chapter 6 • Frequently Asked Questions
(FAQ).
In the example below (system tick rate = 40Hz, t = 25ms, Hitachi
44780 LCD controller), OS_Delay() is used to delay the LCD task
TaskDisp() during startup while the LCD is being configured. By
using OS_Delay() instead of an in-line delay, the other tasks may
run while TaskDisp() is delayed and the LCD is initialized.

See Also

OS_DelayTS(), OS_Stop(), OSTimer()

70

Salvo User Manual

When delaying a task repetitively, remember that there is an additional,
unpredictable delay between when the task's delay expires and when it
actually runs. This may happen if there are other, higher-priority tasks eligible
to run when the delayed task's delay expires. This can affect a task's "loop
delay."

Chapter 7 • Reference

243

Example

#define
#define
#define
#define
#define

LCD_CMD_REG
LCD_DATA_REG
LCD_CMD_CLS
LCD_CMD_MODE
LCD_CMD_ON_OFF

#define
#define
#define
#define

LCD_CMD_FN_SET
LCD_BITMASK_RS
LCD_BITMASK_RW
LCD_BITMASK_E

0
1
0x01
0x06
0x0C

/* for commands
*/
/* for data
*/
/* clear display
*/
/* auto-inc address*/
/* on, no cursor, */
/* no blink
*/
0x3F
0x01 /* reg select
*/
0x02 /* read/-write
*/
0x04 /* E (strobe)
*/

void TaskDisp ( void )
{
static OStypeMsgP msgP;
/* initialize the LCD Display
*/
char i;
/* doesn't need to be static */
TRISD = 0x00; /* all LCD ports are outputs
TRISE = 0x00; /* "
PORTE = 0x00; /* RS=0, -WRITE, E=0

*/
*/
*/

/* we want to talk to the command register,
/* and we'll wait 50ms to ensure it's
/* listening.
LCDSelReg(LCD_CMD_REG);
OS_Delay(2);

*/
*/
*/

/* Hitachi recommends 4 consecutive writes
/* to this register ...
for (i = 4; i--; )
LCDWrData(LCD_CMD_FN_SET);

*/
*/

/* configure LCD the "standard" way.
LCDWrData(LCD_CMD_ON_OFF);
LCDWrData(LCD_CMD_MODE);
LCDWrData(LCD_CMD_CLS);

*/

/* wait another 50ms.
OS_Delay(2);

*/

/* now we're done initializing LCD display.
…

*/

while (1) {
OS_WaitMsg(MSG_UPDATE_DISP_P, &msgP,
OSNO_TIMEOUT);
…
}
}

244

Chapter 7 • Reference

Salvo User Manual

OS_DelayTS(): Delay the Current Task Relative to its
Timestamp and Context-switch
Type:
Declaration:

Macro (invokes OSDelay())
OS_DelayTS (
OStypeDelay delay);

Callable from:
Contained in:
Enabled by:
Affected by:
Description:

Task only

Parameters:
Returns:
Stack Usage:
Notes

salvompt.h
OSBYTES_OF_DELAY, OSBYTES_OF_TICKS
OSENABLE_STACK_CHECKING, OSLOGGING

Delay the current task by the amount
specified, relative to the task's timestamp.
Return to scheduler.
delay: an integer (>=0) specifying the
desired delay in system ticks.
–
2

A delay of 0 will stop the current task. A non-zero delay will delay
the current task by the number of ticks specified relative to the
task's timestamp. The timestamp is automatically recorded by OSInit() and whenever a task's delay times out. In order to use delays with timestamps, Salvo's timer must be installed and the
counting of system ticks must be enabled via OSBYTES_OF_TICKS.
If more than delay and less than 2 x delay system ticks occur between the task's delay expiring and the task running,71 the task will
attempt to resynchronize itself for the following delay period. The
behavior for more than 2 x delay ticks is undefined.72
Do not call OS_Delay() from within an ISR!
In the example below, TaskA() will always run every fourth system tick because it is synchronized to the system timer. As long as
the delay between the task's delay expiring and the task actually
running73 never exceeds 2 delay periods, the task will always run
at t0 + (number of iterations x delay).

71
72
73

Salvo User Manual

I.e. the task is " very late".
In this situation you may need to chose a longer system tick period.
This might happen if, for instance, TaskA()'s priority is low, and there are
other tasks eligible to run.

Chapter 7 • Reference

245

See Also
Example

OS_Delay(), OSGetTS(), OSSetTS(), OS_Stop(), OSSyncTS(),
OSTimer()
void TaskA ( void )
{
while (1) {
OS_DelayTS(4);
…
}
}
int main ( void )
{
…
OSInit();
OSCreateTask(TaskA, OSTCBP(1), 4);
…
enable_interrupts();
while (1) {
OSSched();
}
}

246

Chapter 7 • Reference

Salvo User Manual

OS_Destroy(): Destroy the Current Task and Contextswitch
Type:
Declaration:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

Macro (invokes OSDestroy())
OS_Destroy ();
Task only
salvompt.h

–
OSENABLE_STACK_CHECKING

Destroy the current task. Return to scheduler.
–
–
1

Once a task is destroyed, it cannot be restarted. However, a new
task can be created in its place by using the same tcb.
Do not call OS_Destroy() from within an ISR!
In the example below, TaskStartup() creates and starts most of
the other tasks in the application. TaskDisp() (see example for
OS_Delay()) will run immediately after TaskStartup() begins its
two-second delay. When the delay expires, TaskStartup() will
resume, creating and starting TaskMsg(), TaskRdKey(), TaskStatus(), TaskTx() and TaskRx(). However, none of these tasks
will run until TaskStartup() destroys itself and returns to the
scheduler. Once TaskRx() runs it will create TaskRcvRsp() in
place of TaskStatus(), thereby reusing the tcb for another task.
TaskStartup() is not structured as an infinite loop – rather, it's
simply a one-time sequence of events, which ends when TaskStartup() destroys itself and returns to the scheduler.

See Also

Salvo User Manual

OSCreateTask(), OSStop()

Chapter 7 • Reference

247

Example

void TaskStartup ( void )
{
/* create all the tasks we need early on.
/* Some of these tasks create other tasks
/* and resources! Start them up, too.

*/
*/
*/

/* TaskDisp() handles display updates. It
*/
/* also creates MSG_DISP & SEM_UPDATE_DISP. */
OSCreateTask(TaskDisp, TASK_DISP_P,
TASK_DISP_PRIO);
/* Leave startup screen showing for 2s.
OS_Delay(TWO_SEC, TaskStartup1);

*/

/* TaskMsg() flashes messages. It also
/* creates MSG_FLASH_STRING.
OSCreateTask(TaskMsg, TASK_MSG_P,
TASK_MSG_PRIO);

*/
*/

/* TaskRdKey() reads the keypad. It also
/* creates MSG_KEY_PRESSED and creates and
/* starts TaskRcvKeys().
OSCreateTask(TaskRdKey, TASK_RD_KEY_P,
TASK_RD_KEY_PRIO);

*/
*/
*/

/* TaskStatus() monitors the PSR on Driver.
/* It also creates MSG_WAKE_STATUS and
/* MSG_LONG_OP_DONE.
OSCreateTask(TaskStatus, TASK_STATUS_P,
TASK_STATUS_PRIO);

*/
*/
*/

/* TaskTx() send cmds out to the Driver. It */
/* also creates MSG_WAKE_TX, MSG_RSP_RCVD
*/
/* and MSG_TX_BUFF_EMPTY.
*/
OSCreateTask(TaskTx, TASK_TX_P, TASK_TX_PRIO);
/* TaskRx() receives responses back from the */
/* Driver. It also creates SEM_RX_RBUFF and */
/* creates and starts TaskRcvRsp().
*/
OSCreateTask(TaskRx, TASK_RX_P, TASK_RX_PRIO);
/* we're finished starting up, so kill this
/* task permanently. TaskRcvKeys() will
/* "take over" its tcb – see
/* TaskRdKeys().
OS_Destroy();

*/
*/
*/
*/

}

248

Chapter 7 • Reference

Salvo User Manual

OS_Replace(): Replace the Current Task and Contextswitch
Type:
Declaration:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:

Returns:
Stack Usage:
Notes

Macro (invokes OSCreateTask())
OS_Replace (tFP, prio);
Task only
salvompt.h

–
–
Replace the current task with the one
specified. Return to scheduler.
tFP: a pointer to the task's start address.
This is also the task's function prototype
name.
prio: the desired priority for the task. If
OR'd with OSDONT_START_TASK, the task
will not be started.
–
3

The task that replaces the current task will use the same tcb. Once
a task is replaced, it can be restarted only with a call to OSCreateTask().
Do not call OS_Replace() from within an ISR!
is useful in various situations. For instance, you
could have a system initialization task that replaces itself with one
of your run-time tasks when all initialization is complete. Or you
could replace a large task containing a state machine with independent tasks for each state. OS_Replace() can be used wherever
multiple tasks need never run at the same time, thus conserving tcb
RAM.
OS_Replace()

In the example below, TaskCountUp() runs first. After 250 iterations, it replaces itself with TaskCountDown(). TaskCountDown()
also runs for 250 iterations, but at a faster rate, and replaces itself
with TaskCountUp() when done. The task priorities can be varied,
as shown. This continues indefinitely. Only a single tcb is used.

See Also

Salvo User Manual

OSCreateTask(), OSDestroyTask(), OSStop()

Chapter 7 • Reference

249

Example

void TaskCountUp ( void );
void TaskCountDown ( void );
void TaskCountUp ( void )
{
static char i;
for (i = 0; i <= 250; i++) {
PORTB = i;
OS_Delay(25);
}
OS_Replace(TaskCountDown, 5);
}

void TaskCountDown ( void )
{
static char i;
for (i = 250; i >= 0; i--) {
PORTB = i;
OS_Delay(5);
}
OS_Replace(TaskCountUp, 3);
}

int main ( void )
{
…
OSInit();
OSCreateTask(TaskCountUp, OSTCBP(1), 4);
…
while (1) {
OSSched();
}
}

250

Chapter 7 • Reference

Salvo User Manual

OS_SetPrio(): Change the Current Task's Priority and
Context-switch
Type:
Declaration:

Macro (invokes OSSetPrio())
OS_SetPrio (
OStypePrio prio);

Callable from:
Contained in:
Enabled by:
Affected by:
Description:

Task only

Parameters:
Returns:
Stack Usage:
Notes

salvompt.h

–
OSENABLE_STACK_CHECKING

Change the current task's priority. Return
to scheduler.
prio: the desired (new) priority for the
current task.
–
1

0 (OSHIGHEST_PRIO) is the highest priority, 15 (OSLOWEST_PRIO)
is the lowest.
Do not call OS_SetPrio() from within an ISR!
Tasks can share priorities. Eligible tasks with the same priority will
round-robin schedule as long as they are the highest-priority eligible tasks.
The change in priority takes effect when the current task returns to
the scheduler.
In the example below, TaskStartupEtc() is initially created with
a high priority. The first time it runs, it will run at that priority.
While running for the first time, it redefines its priority to be a
lower one. Each subsequent time it runs, it will run at the lower
priority. The task context-switches once at OS_SetPrio(), and
subsequently at OS_Yield().

See Also

Salvo User Manual

OSCreateTask(), OSGetPrio(), OSSetPrio(),
OSDISABLE_TASK_PRIORITIES

Chapter 7 • Reference

251

Example

#define MOST_IMPORTANT 0
#define LESS_IMPORTANT 5
int main ( void )
{
…
/* startup task gets highest priority.
OSCreateTask(TaskStartupEtc,
OSTCBP(1), MOST_IMPORTANT);
…
}
/* while starting up this task runs at
/* the highest priority, then it changes
/* its priority to a lower one.
void TaskStartupEtc ( void )
{
/* do initialization and other
/* startup code.
…
/* MonitorSystem() will always be
/* called from this task while
/* running at a lower priority.
OS_SetPrio(LESS_IMPORTANT);

*/

*/
*/
*/

*/
*/

*/
*/
*/

while (1) {
MonitorSystem();
OS_Yield();
}
}

252

Chapter 7 • Reference

Salvo User Manual

OS_Stop(): Stop the Current Task and Context-switch
Type:

Macro (invokes OS_Delay() or
OS_Stop())

Declaration:
Callable from:
Contained in:
Enabled by:
Affected by:

OS_Stop ();

Description:
Parameters:
Returns:
Stack Usage:
Notes

Task only
salvompt.h

–
OSBYTES_OF_DELAY,
OSENABLE_STACK_CHECKING, OSLOGGING

Stop the current task. Return to scheduler.
–
–
1

A stopped task can only be restarted via OSStartTask().
Do not call OS_Stop() from within an ISR!
If delays are enabled via OSBYTES_OF_DELAYS, OS_Stop() stops
the current task via a call to OSDelay(0). Otherwise it calls OSStop(). This is done to reduce the code size of your Salvo application.
In the example below, TaskRunOnce() is created and started, and
will run as soon as it becomes the highest-priority eligible task. It
will run only once. In order to make it run again, a call to
OSStartTask(TASK_RUN_ONCE) is required. Note that TaskRunOnce() would also work without the infinite loop, but subsequent calls to OSStartTask(TASK_RUN_ONCE) would result in
unpredictable behavior because task execution would resume outside of TaskRunOnce().

See Also

Salvo User Manual

OSStartTask(), OSStopTask()

Chapter 7 • Reference

253

Example

int main ( void )
{
…
OSCreateTask(TaskRunOnce, TASK_RUN_ONCE_P, 6);
…
}
void TaskRunOnce ( void )
{
while (1) {
/* do one-time things ... */
…
OS_Stop();
}
}

254

Chapter 7 • Reference

Salvo User Manual

OS_WaitBinSem(): Context-switch and Wait the Current
Task on a Binary Semaphore
Type:
Declaration:

Macro (invokes OSWaitEvent())
OS_WaitBinSem (
OStypeEcbP ecbP,
OStypeDelay timeout);

Callable from:
Contained in:
Enabled by:

Task only

Affected by:
Description:

Parameters:

Returns:
Stack Usage:
Notes

salvompt.h
OSENABLE_BINARY_SEMAPHORES, OSEVENTS
OSENABLE_STACK_CHECKING,
OSENABLE_TIMEOUTS, OSLOGGING

Wait the current task on a binary semaphore, with a timeout. If the semaphore is
0, return to the scheduler and continue
waiting. If the semaphore is 1, reset it to 0
and continue. If the timeout expires before the semaphore becomes 1, continue
execution of the task, with the timeout
flag set.
ecbP: a pointer the binary semaphore's
ecb.
timeout: an integer (>=0) specifying the
desired timeout in system ticks.
–
2

Specify a timeout of OSNO_TIMEOUT if the task is to wait the binary
semaphore indefinitely.
Do not call OS_WaitBinSem() from within an ISR!
After a timeout occurs the binary semaphore is undefined.
In the example below for a rocket launching system, a rocket is
launched via a binary semaphore BINSEM_LAUNCH_ROCKET used as
a flag. The semaphore is initialized to zero so that the rocket does
not launch on system power-up.74 Once the rocket is ready and the
order has been given to launch (via OSSignalBinSem() elsewhere
in the code), TaskLaunchRocket() starts the rocket on its journey.
Since the rocket cannot be recalled, there is no need to continue
running TaskLaunchRocket(), and it simply stops itself. There74

Salvo User Manual

That would be undesirable.

Chapter 7 • Reference

255

fore in order to launch a second rocket, the system must be restarted.

See Also

Example

OSCreateBinSem(), OSReadBinSem(), OSSignalBinSem(),
OSTryBinSem()

#define BINSEM_LAUNCH_ROCKET_P OSECBP(2)
…
/* startup code: no clearance given to launch
/* rocket.
OSCreateBinSem(BINSEM_LAUNCH_ROCKET_P, 0);

*/
*/

…
void TaskLaunchRocket ( void )
{
/* wait here forever until the order is
/* given to launch the rocket.
OS_WaitBinSem(BINSEM_LAUNCH_ROCKET_P,
OSNO_TIMEOUT);

*/
*/

/* launch rocket.
IgniteRocketEngines();
…

*/

/* rocket is on its way, therefore task is
/* no longer needed.
OS_Stop();

*/
*/

}

256

Chapter 7 • Reference

Salvo User Manual

OS_WaitEFlag(): Context-switch and Wait the Current
Task on an Event Flag
Type:
Declaration:

Macro (invokes OSWaitEvent())
OS_WaitEFlag (
OStypeEcbP
ecbP,
OStypeEFlag mask,
OStypeOption options,
OStypeDelay timeout);

Callable from:
Contained in:
Enabled by:
Affected by:

Task only

Description:

Wait the current task on an event flag, with
a timeout. The bits in the event flag specified by the mask parameter are tested according to the condition specified by the
options parameter. If the condition is not
satisfied, return to the scheduler and continue waiting. If the condition is satisfied, continue without changing the event
flag. If the timeout expires before the
condition is satisfied, continue execution
of the task, with the timeout flag set.
ecbP: a pointer the event flag's ecb.
mask: a bitmask to apply to the event flag.
options: OSANY_BITS, OSALL_BITS or
OSEXACT_BITS.
timeout: an integer (>=0) specifying the
desired timeout in system ticks.
–
2

Parameters:

Returns:
Stack Usage:
Notes

salvompt.h
OSENABLE_EVENT_FLAGS, OSEVENTS
OSBYTES_OF_EVENT_FLAGS,
OSENABLE_STACK_CHECKING,
OSENABLE_TIMEOUTS, OSLOGGING

Specify a timeout of OSNO_TIMEOUT if the task is to wait the event
flag indefinitely.
Do not call OS_WaitEFlag() from within an ISR!
After a timeout occurs the event flag is undefined.
Salvo's event flag bits are "active high", i.e. an event is said to have
occurred when its corresponding bit in the event flag is set to 1.
The event has not occurred if the bit is cleared to 0.

Salvo User Manual

Chapter 7 • Reference

257

When specifying OSANY_BITS, OS_WaitEFlag() checks if any of
the corresponding mask parameter's bits in the event flag are set to
1, and if so, the task continues. With OSALL_BITS, all of the corresponding mask parameter's bits must be set to 1 for the task to continue. With OSEXACT_BITS, the event flag must match the mask
parameter exactly for the task to continue.
In contrast to Salvo's other event services, successfully waiting an
event flag does not automatically reset the bits in the event flag
that resulted in the condition being satisfied. You must explicitly
clear event flag bits via OSClrEFlag(). Failing to clear the
appropriate event flag bits will cause unpredictable results –
generally the task will fail to yield back to the scheduler.
In the example below for a secure access system with a powerassisted door, three separate interlocks must be deactivated before
the door can be opened by TaskOpenDoor(). The three least significant bits of an eight-bit event flag are used to signify that the
bottom, side and top interlocks have been deactivated by TaskReleaseBottomLock(), etc. Bits three and four in the event flag signify whether the door is fully open or fully closed and are
maintained by TaskCheckDoor(). When the door is fully open, it's
safe to re-activate (release) the door locks so that when it closes it's
automatically locked shut.
The remaining three bits in the eight-bit event flag can be used for
other purposes entirely independent of the interlock mechanism.

See Also

Example

OSCreateEFlag(), OSClrEFlag(), OSReadEFlag(), OSSetEFlag()

#define
#define
#define
#define
#define
#define

DOOR_EFLAG_P
BOTTOM
SIDE
TOP
OPEN
CLOSED

OSECBP(1)
0x01
0x02
0x04
0x08
0x10

void TaskReleaseBottomLock ( void )
{
while (1) {
/* wait for request to release bottom lock.*/
…
/* release bottom door lock.
*/
ReleaseBottomLock();

258

Chapter 7 • Reference

Salvo User Manual

/* tell TaskOpenDoor() about it.
OSSetEFlag(DOOR_EFLAG_P, BOTTOM);

*/

/* verify that door is fully opened by
*/
/* by waiting for the signal.
*/
OS_WaitEFlag(DOOR_EFLAG_P, OPEN, OSANY_BITS,
OSNO_TIMEOUT);
/* re-engage bottom door lock. When door
/* closes it will remain locked.
OSClrEFlag(DOOR_EFLAG_P, BOTTOM);
EngageBottomLock();

*/
*/

/* remain inactive until the door closes. */
OS_WaitEFlag(DOOR_EFLAG_P, CLOSED, OSANY_BITS,
OSNO_TIMEOUT);
}
}
void TaskReleaseSideLock ( void )
{
while (1) {
…
ReleaseSideLock();
OSSetEFlag(DOOR_EFLAG_P, SIDE);
OS_WaitEFlag(DOOR_EFLAG_P, OPEN, OSANY_BITS,
OSNO_TIMEOUT);
OSClrEFlag(DOOR_EFLAG_P, SIDE);
EngageSideLock();
OS_WaitEFlag(DOOR_EFLAG_P, CLOSED, OSANY_BITS,
OSNO_TIMEOUT);
}
}
void TaskReleaseTopLock ( void )
{
while (1) {
…
ReleaseTopLock();
OSSetEFlag(DOOR_EFLAG_P, TOP);
OS_WaitEFlag(DOOR_EFLAG_P, OPEN, OSANY_BITS,
OSNO_TIMEOUT);
OSClrEFlag(DOOR_EFLAG_P, TOP);
EngageTopLock();
OS_WaitEFlag(DOOR_EFLAG_P, CLOSED, OSANY_BITS,
OSNO_TIMEOUT);
}
}
void TaskOpenTheDoor ( void )
{
/* door is initially closed.
OSCreateEFlag(DOOR_EFLAG_P, CLOSED );
while (1) {
/* wait forever for all interlocks to be
/* released.

Salvo User Manual

Chapter 7 • Reference

*/

*/
*/

259

OS_WaitEFlag(DOOR_EFLAG_P,
TOP | BOTTOM | SIDE, OSALL_BITS,
OSNO_TIMEOUT);
/* all locks are released – open door.
OpenDoor();

*/

/* wait for the door to close again before */
/* repeating the cycle.
*/
OS_WaitEFlag(DOOR_EFLAG_P, CLOSED, OSANY_BITS,
OSNO_TIMEOUT);
}
}
void TaskCheckDoor ( void )
{
while (1) {
/* check sensors every 1s.
OS_Delay(100);
/* if open door has closed
/* sensor, then door must
if (DoorFullyOpen()) {
OSSetEFlag(DOOR_EFLAG_P,
}
else {
OSClrEFlag(DOOR_EFLAG_P,
}

*/

contact on its
be open!

*/
*/

OPEN);

OPEN);

/* similarly, if closed door has closed
*/
/* contact on its sensor, then it must be */
/* closed!
*/
if (DoorFullyClosed()) {
OSSetEFlag(DOOR_EFLAG_P, CLOSED);
}
else {
OSClrEFlag(DOOR_EFLAG_P, CLOSED);
}
}
}

260

Chapter 7 • Reference

Salvo User Manual

OS_WaitMsg(): Context-switch and Wait the Current Task
on a Message
Type:
Declaration:

Macro (invokes OSWaitEvent())
OS_WaitMsg (
OStypeEcbP ecbP,
OStypeMsg
msgP,
OStypeDelay timeout);

Callable from:
Contained in:
Enabled by:
Affected by:

Task only

Description:

Wait the current task on a message, with a
timeout. If the message is available, make
msgP point to it, and continue. If it's not
available, return to the scheduler and continue waiting. If the timeout expires before the message becomes available,
continue execution of the task, with the
timeout flag set.
ecbP: a to the message's ecb.
msgP: a pointer to a message
timeout: an integer (>=0) specifying the
desired timeout in system ticks.
–
2

Parameters:

Returns:
Stack Usage:
Notes

salvompt.h
OSENABLE_MESSAGES, OSEVENTS
OSENABLE_STACK_CHECKING,
OSENABLE_TIMEOUTS, OSLOGGING

Specify a timeout of OSNO_TIMEOUT if the task is to wait the message indefinitely.
Do not call OS_WaitMsg() from within an ISR!
Should a timeout occur while waiting the message queue, the message pointer is invalid. A task may only extract the message's contents via the message pointer if it has successfully waited the
message queue event without a timeout.
In the example below, TaskRcvKeys() waits forever for the message MSG_KEY_PRESSED. No processing power is allocated to
TaskRcvKeys() while it is waiting. Once the message arrives, its
contents (the key pressed) are copied to a local variable and appropriate action is taken. Note that correct casting and dereferencing
of the pointer msgP are required in order to extract the contents of

Salvo User Manual

Chapter 7 • Reference

261

the message correctly. After TaskRcvKeys() acts on the key
pressed, it resumes waiting for the message.

See Also
Example

OSCreateMsg(), OSReadMsg(), OSSignalMsg(), OSTryMsg()

void TaskRcvKeys ( void )
{
static char key;
static OStypeMsgP msgP;
while (1) {
/* Wait forever for a new key.
OS_WaitMsg(MSG_KEY_PRESSED_P,
&msgP, OSNO_TIMEOUT);

*/

/* User pressed a key! – get it.
key = *(char *) msgP;

*/

/* Act on key pressed.
switch (tolower(key)) {
case KEY_MEM:
…
}

*/

}
}

262

Chapter 7 • Reference

Salvo User Manual

OS_WaitMsgQ(): Context-switch and Wait the Current
Task on a Message Queue
Type:
Declaration:

Macro (invokes OSWaitEvent())
OS_WaitMsgQ (
OStypeEcbP ecbP,
OStypeMsg
msgP,
OStypeDelay timeout);

Callable from:
Contained in:
Enabled by:
Affected by:
Description:

Task only

Parameters:

Returns:
Stack Usage:
Notes

salvompt.h
OSENABLE_MESSAGE_QUEUES, OSEVENTS
OSLOGGING, OSENABLE_STACK_CHECKING

Wait the current task on a message queue,
with a timeout. If the message queue contains a message, make msgP point to it,
and continue. If it's empty, return to the
scheduler and continue waiting. If the
timeout expires before a message is added
to the message queue, continue execution
of the task, with the timeout flag set.
ecbP: a pointer to the message queue's ecb.
msgP: a pointer to a message.
timeout: an integer (>=0) specifying the
desired timeout in system ticks.
–
2

Specify a timeout of OSNO_TIMEOUT if the task is to wait the message queue indefinitely.
Do not call OS_WaitMsgQ() from within an ISR!
Should a timeout occur while waiting the message queue, the message pointer is invalid. A task may only extract the message's contents via the message pointer if it has successfully waited the
message queue event without a timeout.
In the first example below, TaskRcvInt() forever waits a message
queue containing messages to objects of type int. When a message arrives, the TaskRcvInt() extracts the message from the
message queue and prints a message. The task continues printing
messages until the message queue is empty, whereupon the task a
context switch occurs.

Salvo User Manual

Chapter 7 • Reference

263

Message queues can also be used to pass raw data. In the second
example below, TaskRcvRawData() extracts unsigned-char-sized
raw data instead of message pointers from the message queue.
sizeof(raw
sizeof(OStypeMsgP).

data
type)
must
not
exceed
E.g. on a target with 16-bit void pointers,
raw data of up to 16 bits in size can be passed in each message.

Note

See Also

Example #1

OSCreateMsgQ(), OSReadMsgQ(), OSSignalMsgQ(),
OSTryMsgQ()

void TaskRcvInt ( void )
{
static int myNum;
static OStypeMsgP msgP;
while (1) {
/* Wait forever for a message.
OS_WaitMsgQ(MSGQ1, &msgP, OSNO_TIMEOUT);
/* A message has arrived – get it.
myNum = *(int *) msgP;

*/

*/

printf("The number was %d. \n", myNum);
}
}

Example #2

/* send raw data in this message.
OSSignalMsgQ(MSGQ1_P, (OStypeMsgP) 'r');
…
void TaskRcvRawData( void )
{
OStypeMsgP msgP;
unsigned char rcvdChar;
while (1) {
/* wait forever for a message.
OS_WaitMsgQ(MSGQ1_P, &msgP);

*/

*/

/* cast (don't dereference) message
/* pointer since raw data was passed.
localUC = (unsigned char) msgP;
printf("received %c \n", rcvdChar);

*/
*/

}
}

264

Chapter 7 • Reference

Salvo User Manual

OS_WaitSem(): Context-switch and Wait the Current Task
on a Semaphore
Type:
Declaration:

Macro (invokes OSWaitEvent())
OS_WaitSem (
OStypeEcbP ecbP,
OStypeDelay timeout);

Callable from:
Contained in:
Enabled by:
Affected by:

Task only

Description:

Wait the current task on a semaphore, with
a timeout. If the semaphore is 0, return to
the scheduler and continue waiting. If the
semaphore is non-zero, decrement the
semaphore and continue. If the timeout
expires before the semaphore becomes
non-zero, continue execution of the task,
with the timeout flag set.
ecbP: a pointer to the semaphore's ecb.
timeout: an integer (>=0) specifying the
desired timeout in system ticks.
–
2

Parameters:
Returns:
Stack Usage:
Notes

salvompt.h
OSENABLE_SEMAPHORES, OSEVENTS
OSENABLE_STACK_CHECKING,
OSENABLE_TIMEOUTS, OSLOGGING

Specify a timeout of OSNO_TIMEOUT if the task is to wait the semaphore indefinitely.
Do not call OS_WaitSem() from within an ISR!
After a timeout occurs the semaphore is undefined.
In the example below, TaskRcvRsp() removes incoming characters from a receive buffer one at a time and processes them.
SEM_RX_BUFF always indicates how many characters are present in
rxBuff[], and is signaled by another task which puts the characters into rxBuff[] one-by-one. TaskRcvRsp() runs as long as
there are characters present in rxBuff[] – when is empty,
TaskRcvRsp() waits. By using a semaphore for inter-task communications there's no need to poll for the existence of characters
in the buffer, and hence overall performance is improved.

See Also
Salvo User Manual

OSCreateSem(), OSReadSem(), OSSignalSem(), OSTrySem()

Chapter 7 • Reference

265

Example

void TaskRcvRsp ( void )
{
static char rcChar;
while (1) {
/* wait until there are response chars
/* waiting ... (TaskRx() signals us when
/* there are).
OS_WaitSem(SEM_RX_RBUFF_P, OSNO_TIMEOUT);

*/
*/
*/

/* then deal with them.
/* get the next char from the buffer
rcChar = rxBuff[rxHead];
rxHead++;
if (rxHead >= SIZEOF_RX_BUFF) {
rxHead = 0;
}
rxCount--;

*/
*/

/* alphanumeric characters are the _only_
/* chars (other than reserved ones) we
/* expect to see in the incoming rcChar.
if (isalnum(rcChar) || ( rcChar == '-' ))
{
…
}
else
{
…
}

*/
*/
*/

}
}

266

Chapter 7 • Reference

Salvo User Manual

OS_Yield(): Context-switch
Type:
Declaration:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

OS_Yield()

Macro
OS_Yield ();

Task only
salvompt.h

–
–
Return to scheduler.
–
–
1 or 2, depending on compiler and target.

causes an immediate, unconditional return to the

scheduler.
Do not call OS_Yield() from within an ISR!
In the example below, TaskUnimportant() is assigned a low priority and runs only when no other higher-priority tasks are eligible
to run. Each time it runs, it increments a counter by 1.

Salvo User Manual

Chapter 7 • Reference

267

Example

unsigned long int unimportantCounter = 0;
int main ( void )
{
OSCreateTask(TaskUnimportant,
TASK_UNIMPORTANT_P, 14);
…
}
void TaskUnimportant ( void )
{
while (1) {
unimportantCounter++;
OS_Yield();
}
}

268

Chapter 7 • Reference

Salvo User Manual

OSClrEFlag(): Clear Event Flag Bit(s)
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:

Description:
Parameters:
Returns:

Stack Usage:
Notes

Function
OStypeErr OSClrEFlag (
OStypeEcbP ecbP,
OStypeEFlag mask );

Anywhere
salvoeflag.c, salvoevent.c
OSENABLE_EVENT_FLAGS, OSEVENTS
OSCALL_OSSIGNALEVENT,
OSENABLE_STACK_CHECKING,
OSCOMBINE_EVENT_SERVICES,
OSLOGGING, OSUSE_EVENT_TYPES

Clear bits in an event flag. No task will be
made eligible by this operation.
ecbP: a pointer to the event flag's ecb.
mask: mask of bits to be cleared.
OSERR_BAD_P if event flag pointer is incorrectly specified.
OSERR_EVENT_BAD_TYPE if specified event
is not an event flag.
OSERR_EVENT_CB_UNINIT if event flag's
control block is uninitialized.
OSERR_EVENT_FULL if event flag doesn't
change.
OSNOERR if event flag bits are successfully
cleared.
1

No tasks are made eligible by clearing bits in an event flag.
This service is typically used immediately after successfully waiting an event flag, since the bits in question are not automatically
cleared by OS_WaitEFlag().
In the example below, a task is configured to run only when two
particular bits in an event flag are set. It then clears one of them
and returns to the waiting state. It will run again when and only
when both bits are set.

See Also

Salvo User Manual

OS_WaitEFlag(), OSCreateEFlag(), OSReadEFlag(), OSSetEFlag()

Chapter 7 • Reference

269

Example

#define EFLAG1_P OSECBP(2)
…
void TaskC ( void )
{
while (1) {
/* wait forever for both bits to be set
OS_WaitEFlag(EFLAG1_P, 0x0C, OSALL_BITS,
OSNO_TIMEOUT);
/* clear the upper bit, leave the lower
/* one alone.
OSClrEFlag(EFLAG1_P, 0x08);

*/

*/
*/

…
}
}

270

Chapter 7 • Reference

Salvo User Manual

OSCreateBinSem(): Create a Binary Semaphore
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:

Description:
Parameters:

Returns:
Stack Usage:
Notes

Function
OStypeErr OSCreateBinSem (
OStypeEcbP
ecbP,
OStypeBinSem binSem );

Anywhere
salvobinsem.c
OSENABLE_BINARY_SEMAPHORES,
OSEVENTS
OSCALL_OSCREATEEVENT,
OSENABLE_STACK_CHECKING,
OSCOMBINE_EVENT_SERVICES,
OSLOGGING, OSUSE_EVENT_TYPES

Create a binary semaphore with the initial
value specified.
ecbP: a pointer to the binary semaphore's
ecb.
binSem: the binary semaphore's initial
value (0 or 1) .
OSNOERR

1

Creating a binary semaphore assigns an event control block (ecb)
to the semaphore.
A newly-created binary semaphore has no tasks waiting for it.
Signaling or waiting a binary semaphore before it has been created
will result in an error if OSUSE_EVENT_TYPES is TRUE.
You can also implement binary semaphores via messages – see
OSCreateMsg().

In the example below, a binary semaphore is used to control access
to a shared resource, an I/O port. The port is initially available for
use, so the semaphore is initialized to 1.

See Also
Example

Salvo User Manual

OS_WaitBinSem(), OSReadBinSem(), OSSignalBinSem(),
OSTryBinSem()
/* PORTB is a general-purpose I/O port.
#define BINSEM_PORTB_P OSECBP(6)
…
/* PORTB is initially available to task that

Chapter 7 • Reference

*/

*/

271

/* wants to use it.
OSCreateBinSem(BINSEM_PORTB_P, 1);
…

272

Chapter 7 • Reference

*/

Salvo User Manual

OSCreateCycTmr(): Create a Cyclic Timer
Type:
Prototype:

Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:

Returns:

Stack Usage:
Notes

Function
OStypeErr OSCreateCycTmr (
OStypeTFP
tFP,
OStypeTcbP
tcbP,
OStypeDelay delay,
OStypeDelay period,
OStypeCTMode mode );

Background only
salvoyclic.c
OSENABLE_CYLIC_TIMERS
–

Create a cyclic timer with the initial delay
and period specified.
tFP: a pointer to the cyclic timer's start
address. This is also the cyclic timer's
function prototype name.
tcbP: a pointer to the cyclic timer's tcb.
delay: the initial delay (> 0), in ticks before the cyclic timer is first called.
period: the time, in ticks (> 0), between
successive calls of the cyclic timer
mode: OSCT_ONE_SHOT (the cyclic timer
will run only once) or OSCT_CONTINUOUS
(the cyclic timer will run indefinitely).
OSNOERR if task is successfully created.
OSERR_BAD_P if the specified tcb pointer is
invalid (i.e. out-of-range).
OSERR_BAD_CT_MODE if mode is unrecognized.
OSERR_BAD_CT_DELAY if delay or period
are 0.
3

Cyclic timers are structured like common functions (with a clear
entry and exit), not like tasks. Cyclic timers take no arguments and
return no values.
Creating a cyclic timer assigns a task control block (tcb) to the cyclic timer.
If you prefer to create the task now and explicitly start it later, OR
mode
parameter
with

OSCreateCycTmr()'s

Salvo User Manual

Chapter 7 • Reference

273

OSDONT_START_CYCTMR.

Then use OSStartCycTmr() to start the
cyclic timer at a later time.
Cyclic timers

require

OSENABLE_CYLIC_TIMERS

that timeouts be enabled. Setting
to TRUE will automatically enable time-

outs.
In the example below, cyclic timer CycTmr1() toggles bit 1 of an
I/O port. CycTmr1() will begin running 23 system ticks after the
scheduler is called, and will repeatedly toggle the port pin every
177 system ticks. CycTmr2() will set bit 2 of an I/O port 12 systems ticks after the scheduler is called, and will then stop.

See Also

OSCycTmrRunning(), OSDestroyCycTmr(), OSResetCycTmr(),
OSSetCycTmrPeriod(), OSStartCycTmr(), OSStopCycTmr()

Example

/* Cyclic timer toggles I/O pin indefinitely.
void CycTmr1 ( void )
{
PORT ^= 0x02;
}

*/

/* Cyclic timer sets I/O pin once.
void CycTmr2 ( void )
{
PORT |= 0x04;
}

*/

…
/* Create the cyclic timers.
OSCreateCycTmr(CycTmr1, OSTCBP(1), 23, 177,
OSCT_CONTINUOUS);
OSCreateCycTmr(CycTmr2, OSTCBP(5), 12, 7,
OSCT_ONE_SHOT);

274

Chapter 7 • Reference

*/

Salvo User Manual

OSCreateEFlag(): Create an Event Flag
Type:
Prototype:

Function
OStypeErr OSCreateEFlag (
OStypeEcbP ecbP,
OStypeEfcbP efcbP,
OStypeEFlag eFlag );

Callable from:
Contained in:
Enabled by:
Affected by:

Anywhere
salvoeflag.c
OSENABLE_EVENT_FLAGS, OSEVENTS
OSCALL_OSCREATEEVENT,
OSENABLE_STACK_CHECKING,
OSCOMBINE_EVENT_SERVICES,
OSLOGGING, OSUSE_EVENT_TYPES

Description:

Create an event flag with the initial value
specified.
ecbP: a pointer to the event flag's ecb.
efcbP: a pointer to the event flag's efcb.
eFlag: the event flag's initial value.

Parameters:
Returns:
Stack Usage:
Notes

OSNOERR

1

Creating an event flag assigns an event control block (ecb) and an
event flag control block (efcb) to the event flag.
A newly-created event flag has no tasks waiting for it.
Signaling or waiting an event flag before it has been created will
result in an error if OSUSE_EVENT_TYPES is TRUE.
Event flags

can

be 8, 16 or 32 bits, depending on
OSBYTES_OF_EVENT_FLAGS. OSCreateEFlag() stores the value of
the event flag in the event flag's pre-existing event flag control
block (efcb) of type OSgltypeEfcb. The number of efcb's in your
application is set by OSEVENT_FLAGS. The first efcb is accessed via
OSEFCBP(1), the second by OSEFCBP(2), etc.
In the example below, an 8-bit event flag is used to signify the occurrence of keypresses from an 8-key machine control keypad.
Each bit maps to a single key. The event flag is initialized to all 0's
to
indicate
that
no
keypresses
have
occurred.
OSBYTES_OF_EVENT_FLAGS is set to 1 in this example's salvocfg.h.

Salvo User Manual

Chapter 7 • Reference

275

See Also
Example

276

OS_WaitEFlag(), OSReadEFlag(), OSSignalEFlag(), OSTryEFlag()
/* event flag is event #3, uses event flag
/* control block #1.
#define EFLAG_KEYS_P
OSECBP(3)
#define EFLAG_KEYS_CB_P OSEFCBP(1)
…
/* Initially no keys have been pressed.
OSCreateEFlag(EFLAG_KEYS_P, EFLAG_KEYS_CB_P,
0x00);
…

Chapter 7 • Reference

*/
*/

*/

Salvo User Manual

OSCreateMsg(): Create a Message
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:

Description:
Parameters:
Returns:
Stack Usage:
Notes

Function
OStypeErr OSCreateMsg (
OStypeEcbP ecbP,
OStypeMsgP msgP );

Anywhere
salvosg.c
OSENABLE_MESSAGE, OSEVENTS
OSCALL_OSCREATEEVENT,
OSENABLE_STACK_CHECKING,
OSCOMBINE_EVENT_SERVICES,
OSLOGGING, OSUSE_EVENT_TYPES

Create a message with the initial value
specified.
ecbP: a pointer to the message's ecb.
msgP: a pointer to a message.
OSNOERR

1

Creating a message assigns an event control block (ecb) to the
message. A newly-created message has no tasks waiting for it.
Messages are passed via pointer so that a message can point to
anything.
Signaling or waiting a message before it has been created will result in an error if OSUSE_EVENT_TYPES is TRUE.
Binary semaphores and resource locking can be implemented via
messages using the values (OStypeMsgP) 0 and (OStypeMsgP) 1
for the messages.
In the example below, a message is created to pass the key pressed
(which is detected by the task TaskReadKey()) to the task TaskHandleKey(), which acts on the keypress. The message is initialized to zero because no keypress is initially detected. If, due to task
priorities and timing, TaskReadKey() signals a new message before TaskHandleKey() reads the existing message, the new key
will be lost.

See Also

Salvo User Manual

OS_WaitMsg(), OSReadMsg(), OSSignalMsg(), OSTryMsg()

Chapter 7 • Reference

277

Example

/* pass key via a message. */
#define MSG_KEY_PRESSED_P OSECBP(4)
…
/* this task reads key presses from a keypad
/* and sends them to TaskHandleKey via a
/* message.
void TaskReadKey ( void )
{
static char key;
/* holds key pressed

*/
*/
*/

*/

/* initially no key has been pressed.
*/
OSCreateMsg(MSG_KEY_PRESSED_P, (OStypeMsgP) 0);
while (1) {
if (kbhit()) {
key = getch();
/* do debouncing, key-repeat, etc.

*/

/* send new key via message.
OSSignalMsg(MSG_KEY_PRESSED_P,
(OStypeMsgP) &key);

*/

}
/* wait 10msec, then test for keypress
/* again.
OS_Delay(TEN_MSEC);

*/
*/

}
}
/* this task acts upon keypresses.
void TaskHandleKey ( void )
{
static char key;
/* holds new key
static OStypeMsgP msgP; /* get msg via ptr
while (1) {
/* do nothing until a key is pressed.
OS_WaitMsg(MSG_KEY_PRESSED_P, &msgP,
OSNO_TIMEOUT);
/* then get the new key and act on it.
key = *(char *)msgP;
switch (tolower(key)) {
case KEY_UP:
MoveUp();
break;
…
}

*/

*/
*/

*/

*/

}
}

278

Chapter 7 • Reference

Salvo User Manual

OSCreateMsgQ(): Create a Message Queue
Type:
Prototype:

Callable from:
Contained in:
Enabled by:
Affected by:

Notes

Function
OStypeErr OSCreateMsgQ (
OStypeEcbP
ecbP,
OStypeMqcbP
mqcbP,
OStypeMsgQPP
msgPP,
OStypeMsgQSize size );

Anywhere
salvomsgq.c
OSENABLE_MESSAGE_QUEUES, OSEVENTS
OSCALL_OSCREATEEVENT,
OSENABLE_STACK_CHECKING,
OSCOMBINE_EVENT_SERVICES,
OSLOGGING, OSUSE_EVENT_TYPES

Description:
Parameters:

Create an empty message queue.
ecbP: a pointer to the message queue's ecb.
mqcbP: a pointer to the message queue's
message queue control block.
msgPP: a pointer to the buffer that will hold
the message queue's message pointers.
size: the number of messages (0 < size <
256) that the message queue can hold.

Returns:
Stack Usage:

OSNOERR

1

Creating a message queue assigns an event control block (ecb) to
the message.
Each message queue has a message queue control block (mqcb)
associated with it. Salvo message queue services use mqcbs to
manage the insertion and removal of messages into and out of each
message queue. You must allocate memory for mqcbs using the
OSMESSAGE_QUEUES configuration option. You must associate a
unique mqcb with each message queue using a message queue control block pointer. These range from OSMQCBP(1) to
OSMQCBP(OSMESSAGE_QUEUES). A newly-created message queue
contains no messages.
A message queue75 holds its message pointers76 within a circular
buffer. You must declare this buffer in your source code as a simple array, and give OSCreateMsgQ() a handle to it via the msgPP
75
76

Salvo User Manual

Of type OSgltypeMsgQP.
Of type OStypeMsgP.

Chapter 7 • Reference

279

parameter. The buffer must hold size message pointers. OSCreateMsgQ() does not have any effect on the contents of the buffer.
In the example below, a 7-element and a 16-element message
queue are created with the buffers MsgQBuff1[] and
MsgQBuff2[], respectively. The message queue control block IDs
are 1 and 2, since memory was allocated for two message queues
via OSMESSAGE_QUEUES in salvocfg.h.
For this example salvocfg.h contains:
#define OSEVENTS
5
#define OSMESSAGE_QUEUES 2

In this example, all of the OSLOC_XYZ configuration options are at
their default values. By using OSLOC_MSGQ and OSLOC_MQCB you
can relocate the buffers and the mqcbs, respectively, into RAM
banks other than the default banks.

See Also

Example

280

OS_WaitMsgQ(), OSReadMsgQ(), OSSignalMsgQ(), OSTryMsgQ(),
OSLOC_MSGQ, OSLOC_MQCB

/* use #defines for legibility
#define SEM1_P
OSECBP(1)
#define SEM2_P
OSECBP(2)
#define BINSEM1_P
OSECBP(3)
#define MSGQ1_P
OSECBP(4)
#define MSGQ2_P
OSECBP(5)
#define MQCB1_P
OSMQCBP(1)
#define MQCB2_P
OSMQCBP(2)
#define SIZEOF_MSGQ1 7
#define SIZEOF_MSGQ2 16

*/

/* allocate memory for buffers
OSgltypeMsgQP MsgQBuff1[SIZEOF_MSGQ1];
OSgltypeMsgQP MsgQBuff2[SIZEOF_MSGQ2];

*/

/* create message queues from existing
/* buffers and mqcbs.
OSCreateMsgQ(MSGQ1_P, MQCBP1_P, MsgQBuff1,
SIZEOF_MSGQ1);
OSCreateMsgQ(MSGQ2_P, MQCBP2_P, MsgQBuff2,
SIZEOF_MSGQ2);

*/
*/

Chapter 7 • Reference

Salvo User Manual

OSCreateSem(): Create a Semaphore
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:

Description:
Parameters:
Returns:
Stack Usage:
Notes

Function
OStypeErr OSCreateSem (
OStypeEcbP ecbP,
OStypeSem sem );

Anywhere
salvosem.c
OSENABLE_SEMAPHORES, OSEVENTS
OSBIG_SEMAPHORES,
OSCALL_OSCREATEEVENT,
OSENABLE_STACK_CHECKING,
OSCOMBINE_EVENT_SERVICES,
OSLOGGING, OSUSE_EVENT_TYPES

Create a counting semaphore with the initial value specified.
ecbP: a pointer to the semaphore's ecb.
sem: the semaphore's initial value.
OSNOERR

1

Creating a semaphore assigns an event control block (ecb) to the
semaphore.
A newly-created semaphore has no tasks waiting for it.
Signaling or waiting a semaphore before it has been created will
result in an error if OSUSE_EVENT_TYPES is TRUE.
In the example below, a counting semaphore is created to mark
how much space is available in a transmit ring buffer. The buffer is
initially empty, so the semaphore is initialized to the size of the
buffer.

See Also

Salvo User Manual

OS_WaitSem(), OSReadSem(), OSSignalSem(), OSTrySem()

Chapter 7 • Reference

281

Example

/* Ring buffer is used to receive characters.
#define SEM_TX_RBUFF_P OSECBP(3)

*/

…
/* initialize semaphore (ring buffer is
/* empty).
OSCreateSem(SEM_TX_RBUFF_P, 16);

*/
*/

…

282

Chapter 7 • Reference

Salvo User Manual

OSCreateTask(): Create and Start a Task
Type:
Prototype:

Callable from:
Contained in:
Enabled by:
Affected by:
Description:

Parameters:

Returns:
Stack Usage:
Notes

Function
OStypeErr OSCreateTask (
OStypeTFP tFP,
OStypeTcbP tcbP,
OStypePrio prio );

Background only
salvoinit2.c

–
OSLOGGING, OSENABLE_STACK_CHECKING

Create a task with the specified start address, tcb pointer and priority. Starts the
task unless overridden by the user in the
prio parameter.
tFP: a pointer to the task's start address.
This is also the task's function prototype
name.
tcbP: a pointer to the task's tcb.
prio: the desired priority for the task. If
OR'd with OSDONT_START_TASK, the task
will not be started.
OSNOERR if task is successfully created.
OSERR_BAD_P if the specified tcb pointer
is invalid (i.e. out-of-range).
3

Creating a task assigns a task control block (tcb) to the task.
0 (OSHIGHEST_PRIO) is the highest priority, 15 (OSLOWEST_PRIO)
is the lowest. If the specified task priority is out-of-range, the task
will still be created, but with the lowest possible priority.
Tasks created via OSCreateTask() are automatically started, i.e.
they are in the eligible state.
If you prefer to create the task now and explicitly start it later, OR
OSCreateTask()'s prio parameter with OSDONT_START_TASK.
Then use OSStartTask() to start the task at a later time.
If task priorities are disabled via OSDISABLE_TASK_PRIORITIES,
OSCreateTask()'s third argument (prio) is used only with
OSDONT_START_TASK, and the priority value is disregarded.

Salvo User Manual

Chapter 7 • Reference

283

OSCreateTask() overwrites the task control block
specified via the tcbP parameter, i.e. it overwrites the tcb. When
calling OSCreateTask() after task scheduling has started via OSSched(), extreme caution must be used to avoid overwriting an ex-

Caution

isting eligible, running, delayed, waiting or stopped task.
In the example below, a single task is created from the function
TaskDoNothing() by assigning it a tcb pointer of TASK1_P, and a
priority of 7.

See Also
Example

OSStartTask(), OSStopTask()
#define TASK1_P OSTCBP(1)/* taskIDs start at 0 */
/* this task does nothing but run, context/* switch, run, context-switch, etc.
void TaskDoNothing ( void )
{
while (1) {
OS_Yield();
}
}

*/
*/

/* create a single task and run it (over and
/* over).
int main ( void )
{
…
/* initialize Salvo. */
OSInit();

*/
*/

/* create a task to do nothing but context/* switch. Tcb pointer is 0, priority is 7
/* (middle). A call to OSSTartTask() is not
/* required …
OSCreateTask(TaskDoNothing, TASK1_P, 7);
…
/* start multitasking.
while (1) {
OSSched();
}

*/
*/
*/
*/

*/

}

284

Chapter 7 • Reference

Salvo User Manual

OSDestroyCycTmr(): Destroy a Cyclic Timer
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

OSDestroyCycTmr()

Function
OStypeErr OSDestroyCycTmr (
OStypeTcbP tcbP );

Background only
salvoyclic4.c
OSENABLE_CYCLIC_TIMERS

–
Destroy the specified cyclic timer.
tcbP: a pointer to the cyclic timer's tcb.
OSNOERR if cyclic timer is destroyed.
OSERR_BAD_CT if the tcb in question does
not belong to a cyclic timer.
3
destroys both running and stopped cyclic

timers.
In the example below, CycTmr3() is created and then destroyed
from within a task after being allowed to run for 200 system ticks.
The task then continues, creating another task – Task4()– which
uses the same tcb.

See Also

Salvo User Manual

OSCreateCycTmr(), OSCycTmrRunning(), OSResetCycTmr(),
OSSetCycTmrPeriod(), OSStartCycTmr(), OSStopCycTmr()

Chapter 7 • Reference

285

Example

286

…
OSCreateCycTmr(CycTmr3, OSTCBP(7), 1, 2,
OSCT_CONTINUOUS);
OS_Delay(200);
OSDestroyCycTmr(OSTCBP(7));
OSCreateTask(Task4, OSTCBP(7), 12);

Chapter 7 • Reference

Salvo User Manual

OSDestroyTask(): Destroy a Task
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:

Stack Usage:
Notes

Function
OStypeErr OSDestroyTask (
OStypeTcbP tcbP,
OStypeID
events );

Task or Background
salvotask3.c

–
OSENABLE_STACK_CHECKING

Destroy the specified task.
tcbP: a pointer to the task's tcb.
events: OSEVENTS.
OSNOERR if specified task was successfully
destroyed.
OSERR if unable to destroy the specified
task.
3

can destroy any task that is not already destroyed or waiting an event.
OSDestroyTask()

The destroyed task's tcb is re-initialized.
The second parameter of OSEVENTS is required for all configurations where events are enabled. If events are not enabled, then OSDestroyTask() takes only a single parameter.
In the example below, TaskMain() has a relatively high priority of
3. When it runs, it creates another, lower-priorty task, TaskWarmUp(). During the next thirty seconds, TaskWarmUp() runs whenever it is the highest-priority eligible task. Then TaskMain()
destroys TaskWarmUp(). Thereafter, OSCreateTask() can be used
to create another task in TaskWarmUp()'s place, using the same tcb
pointer.

See Also

Salvo User Manual

OSCreateTask(), OS_Destroy()

Chapter 7 • Reference

287

Example

OSCreateTask(TaskMain, TASKMAIN, 3);
…
void TaskMain ( void )
{
OSCreateTask(TaskWarmUp, TASKWARMUP_P, 7);
while (1) {
OS_Delay(THIRTY_SEC);
OSDestroyTask(TASKWARMUP_P, OSEVENTS);
…
}
}

288

Chapter 7 • Reference

Salvo User Manual

OSGetPrio(): Return the Current Task's Priority
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

Macro (invokes OSGetPrioTask())
OStypePrio OSGetPrio ( );
Task only
salvoprio2.c

–
OSENABLE_STACK_CHECKING

Return the priority of the current (running)
task.
–
–
1

0 (OSHIGHEST_PRIO) is the highest priority, 15 (OSLOWEST_PRIO)
is the lowest.
In the example below, TaskB() lowers its priority each time it
runs, until it reaches the lowest allowed priority and remains there.

See Also

Salvo User Manual

OS_SetPrio(), OSGetPrioTask(), OSSetPrio(), OSSetPrioTask(), OSDISABLE_TASK_PRIORITIES

Chapter 7 • Reference

289

Example

void TaskB ( void )
{
OStypePrio prio;

while (1) {
…
prio-- = OSGetPrio();
OS_SetPrio(prio);
}
}

290

Chapter 7 • Reference

Salvo User Manual

OSGetPrioTask(): Return the Specified Task's Priority
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

Function
OStypePrio OSGetPrioTask (
OStypeTcbP tcbP );

Task or Background
salvoprio2.c

–
OSENABLE_STACK_CHECKING

Return the priority of the specified task.
tcbP: a pointer to the task's tcb.
–
1

0 (OSHIGHEST_PRIO) is the highest priority, 15 (OSLOWEST_PRIO)
is the lowest.
In the example below, DispTaskPrio() displays the priority of the
specified task.

See Also

Salvo User Manual

OS_SetPrio(), OSGetPrio(), OSSetPrio(), OSSetPrioTask(),
OSDISABLE_TASK_PRIORITIES

Chapter 7 • Reference

291

Example

292

#define TASKE_P OSTCBP(5)
…
void DispTaskPrio ( OStypeTcbP tcbP )
{
printf("Task %d has priority %d.\n",
OStID(tcbP, OSTASKS), OSGetPrioTask(tcbP));
}

Chapter 7 • Reference

Salvo User Manual

OSGetState(): Return the Current Task's State
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

Macro (invokes OSGetStateTask())
OStypeState OSGetState ( );
Task only
salvompt.h

–
OSENABLE_STACK_CHECKING

Return the state of the current (running)
task.
–
Task state.
1

The current task's state is always OSTCB_TASK_RUNNING. This service is included for completeness.
In the example below, TaskG() verifies that it is in fact running.

See Also

Salvo User Manual

OSGetStateTask()

Chapter 7 • Reference

293

Example

294

void TaskC ( void )
{
while (1) {
if (OSGetState() != OSTCB_TASK_RUNNING)
printf("Houston, we have a problem.\n");
}
}

Chapter 7 • Reference

Salvo User Manual

OSGetStateTask(): Return the Specified Task's State
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

Function
OStypeState OSGetState (
OStypeTcbP tcbP );

Task or Background
salvotask5.c

–
OSENABLE_STACK_CHECKING

Return the state of the specified task.
–
Task state.
1

A task may be in one of the following states:
OSTCB_DESTROYED
OSTCB_TASK_STOPPED
OSTCB_TASK_DELAYED
OSTCB_TASK_WAITING
OSTCB_TASK_WAITING_TO

OSTCB_TASK_ELIGIBLE
OSTCB_TASK_SIGNALED

OSTCB_TASK_RUNNING

destroyed / uninitialized
stopped
delayed
waiting on an event
waiting on an event,
with a timeout if in an event
queue. Waited for an event
and timed out if in the
eligible queue
eligible to run
in the eligible queue,
having waited an event that
was signaled
running

In the example below, mainline code verifies that a particular task
has indeed been stopped.

See Also

Salvo User Manual

OSGetState()

Chapter 7 • Reference

295

Example

296

#define TASKC_P OSTCBP(3)
…
if (OSGetStateTask(TASKC_P) != OSTCB_TASK_STOPPED)
/* something's wrong with TaskC().
*/
…

Chapter 7 • Reference

Salvo User Manual

OSGetTicks(): Return the System Timer
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

Function
OStypeTick OSGetTicks ( void );

Anywhere
salvotick.c
OSBYTES_OF_TICKS
OSENABLE_STACK_CHECKING

Obtain the current value of the system
timer (in ticks).
–
Current system timer in ticks.
1

The system timer is initialized to 0 via OSInit().
In the example below, the current value of the system timer is
stored in a variable.

See Also

Salvo User Manual

OSSetTicks()

Chapter 7 • Reference

297

…

Example

OStypeTick ticksNow;
…
/* obtain current value of system ticks.
ticksNow = OSGetTicks();
…

*/

On certain targets it may be advantageous to read the current system ticks (OStimerTicks) directly instead of through OSGetTicks(). Possible scenarios include substantial function call
overhead and/or no need to manage interrupts.77 In the example
below, the current value of the system timer is stored in a variable
by accessing OStimerTicks directly.
…
OStypeTick ticksNow;
…
/* obtain current value of system ticks.
disable_interrupts();
ticksNow = OStimerTicks;
enable_interrupts();
…

77

298

*/

Both of these conditions occur on the baseline PICmicro devices, e.g.
PIC12C509.

Chapter 7 • Reference

Salvo User Manual

OSGetTS(): Return the Current Task's Timestamp
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

Macro (invokes OSGetTSTask())
OStypeTS OSGetTS (void);
Task only
salvodelay3.c
OSBYTES_OF_TICKS
OSENABLE_STACK_CHECKING

Obtain the value of the current task's timestamp (in ticks).
–
Current task's timestamp in ticks.
1

When a task is created, its timestamp is initialized to an OStypeTSsized version of the system timer ticks, i.e. (OStypeTS) OStimerTicks.
In the example below, the current task's timestamp is displayed
whenever it times out.
See OS_DelayTS() for more information on timestamps.

See Also

Salvo User Manual

OS_DelayTS(), OSSetTS(), OSSyncTS()

Chapter 7 • Reference

299

void Task ( void )
{
while (1) {
OS_Delay(7);78

Example

printf("Task %d timed out at %d\n",
OStID(OScTcbP, OSTASKS), OSGetTS());
…
}
}

78

300

The timestamp is redefined whenever a delay expires, whether through
OS_Delay() or OS_DelayTS().

Chapter 7 • Reference

Salvo User Manual

OSInit(): Prepare for Multitasking
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

OSInit()

Function
void OSInit ( void );

Background only
salvoinit.c

–
OSBYTES_OF_DELAYS, OSCLEAR_GLOBALS,
OSENABLE_STACK_CHECKING, OSEVENTS,
OSLOGGING, OSTASKS

Initialize Salvo's pointers, counters, etc.
–
–
2

must be called first, before any other Salvo functions.

The executable code size of OSInit() can be minimized by setting
OSCLEAR_GLOBALS to FALSE. Do this only if you are certain that
your compiler initializes all global variables to 0 at runtime, and
you do not call OSInit() more than once in your application.
does not initialize tcbs or ecbs – this is done on a per-tcb
and per-ecb basis when tasks and events are created, respectively.
OSInit()

In the example below, OSInit() is called before any other Salvo
calls.

Salvo User Manual

Chapter 7 • Reference

301

Example

302

int main ( void )
{
…
/* initialize Salvo.
OSInit();
…
/* start multitasking.
while (1) {
OSSched();
}
}

Chapter 7 • Reference

*/

*/

Salvo User Manual

OSMsgQCount(): Return Number of Messages in Message
Queue
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

Function
OStypeMsgQSize OSMsgQCount (
OStypeTcbP ecbP );

Anywhere
salvomsgq4.c
OSENABLE_MESSAGE_QUEUES
OSCALL_OSMSGQCOUNT

Check whether the specified message
queue has room for additional message(s).
ecbP: a pointer to the message queue's ecb.
Number of messages in message queue,
i.e. returns 0 if message queue is empty.
1

can be used to obtain the current status of the
message queue. OSMsgQCount() returns the count record in the
message queue's message queue control block (mqcb) – therefore
it's very fast.
OSMsgQCount()

No error checking is performed on the ecbP parameter. Calling
OSMsgQCount() with an invalid ecbP, or an ecbP belonging to an
event other than a message queue, will return an erroneous result.
In the example below, OSMsgQCount() is used to obtain the number of messages in a message queue, and the space available for
new messages. When using OSMsgQCount() to calculate available
space in a message queue, it must be subtracted from the size parameter originally used to create the message queue.
See Also

Salvo User Manual

OS_WaitMsgQ(), OSCreateMsgQ(), OSMsgQEmpty(), OSReadMsgQ(), OSSignalMsgQ(), OSTryMsgQ()

Chapter 7 • Reference

303

Example

#define MSGQ1_P OSECBP(1)
printf("msgQ contains %d messages\n",
OSMsgQCount(MSGQ1_P));
printf("msgQ has room for %d messages\n",
SIZEOF_MSGQ1 - OSMsgQCount(MSGQ1_P));

304

Chapter 7 • Reference

Salvo User Manual

OSMsgQEmpty(): Check for Available Space in Message
Queue
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

Function
OStypeMsgQSize OSMsgQEmpty (
OStypeTcbP ecbP );

Anywhere
salvomsgq3.c
OSENABLE_MESSAGE_QUEUES
OSCALL_OSMSGQEMPTY

Check whether the specified message
queue has room for additional message(s).
ecbP: a pointer to the message queue's ecb.
Number of available (empty) spots in message queue, i.e. returns 0 (FALSE) if
message queue is full.
1

Each message queue can contain up to a maximum number of
messages. If messages are added to the message queue (via OSSignalMsgQ()) faster than they are removed (via OS_WaitMsgQ()),
the queue will eventually fill up. OSMsgQEmpty() can be used to
obtain the current status of the message queue without signaling
the message queue.
No error checking is performed on the ecbP parameter. Calling
OSMsgQEmpty() with an invalid ecbP, or an ecbP belonging to an
event other than a message queue, will return an erroneous result.

Note OSMsgQEmpty() performs pointer subtraction when computing the available room in the specified message queue. On some79
targets, this may result in very slow execution. Since interrupts are
disabled during OSMsgQEmpty(), this is not desirable.
OSMsgQCount() always executes very quickly, and is preferred in
these cases.
In the first example below, mainline code signals a message queue
with a message from the user's msg array only if space is available.
If not, an error counter is incremented. This example will give erroneous results if messages are also signaled to the same message
queue from within an interrupt handler. That's because interrupts
79

Salvo User Manual

For example, on an 8-bit target where data pointers are 16 bits.

Chapter 7 • Reference

305

are enabled between the call to OSMsgQEmpty() and the call to OSSignalMsgQ(). In that case, OSSignalMsgQ()'s return code of
OSERR_EVENT_FULL can be used to detect the inability to enqueue a
message into a message queue.
In the second example below, the message queue is filled to capacity with new message pointers of ascending value, starting at 0.

See Also

Example #1

OS_WaitMsgQ(), OSCreateMsgQ(), OSMsgQCount(), OSReadMsgQ(), OSSignalMsgQ(), OSTryMsgQ()
#define MSGQ3_P OSECBP(4)
unsigned int counter;
if (OSMsgQEmpty(MSGQ3_P)) {
OSSignalMsgQ(MSGQ3_P, (OStypeMsgP) &msg[i]);
}
else {
counter++;
}

Example #2

OStypeMsgQSize roomLeft;
roomLeft = OSMsgQEmpty(MSGQ1_P);
for (i = 0; i < roomLeft; i++) {
OSSignalMsgQ(MSGQ1_P, (OStypeMsgP) i);
}

306

Chapter 7 • Reference

Salvo User Manual

OSReadBinSem(): Obtain a Binary Semaphore
Unconditionally
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

Function
OStypeBinSem OSReadBinSem (
OStypeEcbP ecbP );

Anywhere
salvobinsem.c
OSENABLE_BINARY_SEMAPHORES,
OSENABLE_EVENT_READING, OSEVENTS
OSCALL_OSRETURNEVENT

Returns the binary semaphore specified by
ecbP.
ecbP: a pointer to the binary semaphore's
ecb.
Binary semaphore (0 or 1).
1

has no effect on the specified binary semaphore.
Therefore it can be used to obtain the binary semaphore's value
without affecting the state(s) of any task(s).
OSReadBinSem()

No error checking is performed on the ecbP parameter. Calling
OSReadBinSem() with an invalid ecbP, or an ecbP belonging to an
event other than a binary semaphore, will return an erroneous result.
In the example below, a binary semaphore employed as a resource
is tested before making a decision to delay a task.

See Also

Salvo User Manual

OS_WaitBinSem(), OSCreateBinSem(), OSTryBinSem(), OSSignalBinSem()

Chapter 7 • Reference

307

Example

…
/* initially, resource #2 is available.
OSCreateBinSem(BINSEM_RSRC2_P, 1);

*/

void TaskD ( void )
{
while (1) {
…
if (OSReadBinSem(BINSEM_RSRC2_P)) {
MyFn();
}
else {
OS_Delay(100);
}
}
}

308

Chapter 7 • Reference

Salvo User Manual

OSReadEFlag(): Obtain an Event Flag Unconditionally
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

Function
OStypeEFlag OSReadEFlag (
OStypeEcbP ecbP );

Anywhere
salvoeflag2.c
OSENABLE_EVENT_FLAGS,
OSENABLE_EVENT_READING, OSEVENTS
OSCALL_OSRETURNEVENT

Returns the event flag specified by ecbP.
ecbP: a pointer to the event flag's ecb.
Event flag.
1

has no effect on the specified event flag. Therefore it can be used to obtain the event flag's value without affecting
the state(s) of any task(s).
OSReadEFlag()

No error checking is performed on the ecbP parameter. Calling
OSReadEFlag() with an invalid ecbP, or an ecbP belonging to an
event other than an event flag, will return an erroneous result.
In the example below, TaskF() waits on one of two bits to be set
in an event flag pointed to by EFLAG_P. OSReadEFlag() is then
used to determine which of the two bits was set.

See Also

Salvo User Manual

OS_WaitEFlag(),OSClrEFlag(), OSCreateEFlag(), OSSetEFlag()

Chapter 7 • Reference

309

Example

void TaskF ( void )
{
OStypeEFlag eFlag;

while (1) {
OS_WaitEFlag(EFLAG_P, 0xC0, OSANY_BITS,
OSNO_TIMEOUT);
eFlag = OSReadEFlag(EFLAG_P);
if (eFlag & 0x80) {
/* topmost bit was set …
…
}
else {
/* other bit was set …
…
}

*/

*/

}
}

310

Chapter 7 • Reference

Salvo User Manual

OSReadMsg():Obtain a Message's Message Pointer
Unconditionally
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

Function
OStypeMsgP OSReadMsg (
OStypeEcbP ecbP );

Anywhere
salvomsg.c
OSENABLE_MESSAGES,
OSENABLE_EVENT_READING, OSEVENTS
OSCALL_OSRETURNEVENT

Returns a pointer to the message specified
in ecbP.
ecbP: a pointer to the message's ecb.
Message pointer.
1

has no effect on the specified message. Therefore it
can be used to obtain the message's message pointer without affecting the state(s) of any task(s).
OSReadMsg()

No error checking is performed on the ecbP parameter. Calling
OSReadMsg() with an invalid ecbP, or an ecbP belonging to an
event other than a message, will return an erroneous result.
In the example below, a task checks to see if a message is nonempty before signaling the message.80 Thus it avoids losing the
message.

See Also

OS_WaitMsg(), OSCreateMsg(), OSSignalMsg(), OSTryMsg()

80

Salvo User Manual

If the application allowed signaling the message from an interrupt, additional
interrupt control would be required in TaskC() in order to guarantee that the
message is empty before signaling it.

Chapter 7 • Reference

311

Example

/* send this when there's a problem.
*/
const char strImpMsg[] = "Important Message!\n";
void TaskC ( void )
{
while (1) {
…
/* delay one system tick as long as MSG
/* has a message in it.
while (OSReadMsg(MSG_P)) {
OS_Delay(1);
}

*/
*/

/* now that MSG is empty, we can send our */
/* important message.
*/
OSSignalMsg (MSG_P, (OStypeMsgP) &strImpMsg);
}
}

312

Chapter 7 • Reference

Salvo User Manual

OSReadMsgQ(): Obtain a Message Queue's Message
Pointer Unconditionally
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

Function
OStypeMsgP OSReadMsgQ (
OStypeEcbP ecbP );

Anywhere
salvomsgq.c
OSENABLE_EVENT_READING,
OSENABLE_MESSAGE_QUEUES, OSEVENTS
OSCALL_OSRETURNEVENT

Returns a pointer to the next message in
the message queue specified in ecbP.
ecbP: a pointer to the message's ecb.
Message pointer.
1

has no effect on the specified message queue.
Therefore it can be used to obtain the message queue's message
pointer without affecting the state(s) of any task(s).
OSReadMsgQ()

No error checking is performed on the ecbP parameter. Calling
OSReadMsgQ() with an invalid ecbP, or an ecbP belonging to an
event other than a message queue, will return an erroneous result.
In the example below, message queue #2 is slowly filled with a
new character message every few seconds. TaskB() monitors the
message queue every second. Whenever there are one or more
valid messages in the message queue, TaskB() displays the first
message's contents.81 As the waiting task (not shown) waits the
message queue and obtains the messages, TaskB()'s output will
change as well.

See Also

OS_WaitMsgQ(), OSCreateMsgQ(), OSSignalMsgQ(),
OSTryMsgQ()

/* message queue #2 contains single chars.
#define MSGQ2_P OSECBP(6)

Example

81

Salvo User Manual

*/

Note that TaskB(), as written, cannot distinguish between successive,
identical messages. Therefore it will report on a stream of messages
'h','e','l','l','o' as 'h','e','l','o'. However, the waiting task will receive all five
characters in the string.

Chapter 7 • Reference

313

void TaskB ( void )
{
static char oldchar;
char newchar;
OStypeMsgP msgP;
while (1) {
OS_Delay(ONE_SEC);
…
/* test message queue #2
msgP = OSReadMsgQ(MSGQ2_P);

*/

/* get the message if there is one.
if (msgP) {
newchar = *(char *) msgP;
if ( newchar != oldchar ) {
oldchar = newchar;
printf("The new message is: %c\n.",
newchar);
}
}
…

*/

}
}

314

Chapter 7 • Reference

Salvo User Manual

OSReadSem(): Obtain a Semaphore Unconditionally
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

Function
OStypeSem OSReadSem (
OStypeEcbP ecbP );

Anywhere
salvosem.c
OSENABLE_EVENT_READING,
OSENABLE_SEMAPHORES, OSEVENTS
OSCALL_OSRETURNEVENT

Returns the current value of the semaphore
specified in ecbP.
ecbP: a pointer to the semaphore's ecb.
Semaphore.
1

has no effect on the specified semaphore. Therefore
it can be used to obtain the semaphore's value without affecting the
state(s) of any task(s).
OSReadSem()

No error checking is performed on the ecbP parameter. Calling
OSReadSem() with an invalid ecbP, or an ecbP belonging to an
event other than a semaphore, will return an erroneous result.
In the example below, a binary semaphore is used to manage a 15character ring buffer. In case of an error, the program displays a
descriptive message82 before re-initializing the buffer.

See Also

OS_WaitSem(), OSCreateSem(), OSSignalSem(), OSTrySem()

82

Salvo User Manual

printf() does not use the system's Tx facilities.

Chapter 7 • Reference

315

Example

/* initially, Tx buffer has room for 15 chars. */
#define SIZEOF_TXBUFF 15
…
/* manage the Tx buffer as a resource.
*/
OSCreateSem(SEM_TXBUFF_P, SIZEOF_TXBUFF);
…
/* if there's a Tx error, flush and recreate
*/
/* the buffer after displaying a message.
*/
if (TxErr)
{
DisableTxInts();
printf("Error: %d chars stuck in Tx buffer.\n",
SIZEOF_TXBUFF - OSReadSem(SEM_TXBUFF_P));
FlushTxBuff();
OSCreateSem(SEM_TXBUFF_P, SIZEOF_TXBUFF);
EnableTxInts();
}
…

316

Chapter 7 • Reference

Salvo User Manual

OSResetCycTmr(): Reset a Cyclic Timer
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:

Stack Usage:
Notes

Function
OStypeErr OSResetCycTmr (
OStypeTcbP tcbP );

Background only
salvocyclic6.c
OSENABLE_CYCLIC_TIMERS

–
(Re-)set the specified cyclic timer.
tcbP: a pointer to the cyclic timer's tcb.
OSNOERR if cyclic timer is successfully reset.
OSERR_BAD_CT if the tcb in question does
not belong to a cyclic timer.
3

restarts the cyclic timer with its period regardless of whether the cyclic timer is running or not.
OSResetCycTmr()

A cyclic timer can be re-synchronized with OSResetCycTmr().
In the example below, a task waits for a signal to restart a cyclic
timer. When that signal is received, the cyclic timer is stopped and
restarted. Regardless of how close it was previously to timing out,
it will now time out in its normal period.

See Also

Salvo User Manual

OSCreateCycTmr(), OSCycTmrPeriod(), OSCycTmrRunning(),
OSDestroyCycTmr(), OSStartCycTmr(), OSStopCycTmr()

Chapter 7 • Reference

317

Example

318

…
OS_WaitBinSem(BINSEM_RESTART_CYCTMR3,
OSNO_TIMEOUT);
OSResetCycTmr(OSTCBP(6));

Chapter 7 • Reference

Salvo User Manual

OSRpt(): Display the Status of all Tasks, Events, Queues
and Counters
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:

Task or Background

Description:

Display the current status of all Salvo
tasks, events and counters in tabular form.
tasks: OSTASKS.
events: OSEVENTS.
–
3 + printf()'s stack usage

Parameters:
Returns:
Stack Usage:
Notes

Function
void OSRpt (
OStypeID tasks,
OStypeID events );
salvorpt.c

–
OSBYTES_OF_COUNTS,
OSBYTES_OF_DELAYS,
OSENABLE_STACK_CHECKING,
OSENABLE_STATISTICS, OSMON_HIDE_INVALID_PTRS,
OSMON_SHOW_ONLY_ACTIVE, OSMON_SHOW_TOTAL_DELAY,
OSUSE_EVENT_TYPES

requires a working printf() function in the target
application.83 OSRpt() is quite large and is intended for use only in
those systems that have sufficient code space (e.g. x86-based systems) to include it in the target application.
OSRpt()

displays the current task, the members of the eligible and
delayed queues (shown in their priority order), and the fields of
each task control block (tcb) and event control block (ecb). If so
configured, it also displays error, warning and timeout counter values, the maximum call ... return depth, and the total delay of the
tasks in the delay queue.
OSRpt()

reads and displays Salvo's data structures on-the-fly, i.e.
no local copy is made. Depending on the speed at which the
printf() function is able to output characters, OSRpt() may take
quite a while to complete. This may result in a display of informaOSRpt()

83

Salvo User Manual

Some libraries (e.g. Hi-Tech PICC) contain a dummy putch() function
called by printf(). You must supply your own, working putch() for
printf() output to occur.

Chapter 7 • Reference

319

tion that appears to be contradictory (e.g. a task is shown in the
eligible queue and simultaneously waiting for an event). In order to
avoid this, your application must control or disable interrupts while
OSRpt() is executing.

See Also

Chapter 5 • Configuration

…
/* display the current status of all tasks
/* and events (and counters, if so enabled)
/* to the system's terminal screen.
OSRpt(OSTASKS, OSEVENTS);
…

Example

*/
*/
*/

A call to OSRpt() resulted in the following display on a simple
terminal program connected via RS-232 to a Salvo system84 with a
working printf():

Figure 31: OSRpt() Output to Terminal Screen

In Figure 31 we can see that when OSRpt() was called, three tasks
were eligible, five were waiting and/or delayed, and over one billion context switches had occurred over a nearly four-day-long period.85

84
85

320

This output is from the program in \salvo\demo\d1\sysa, running on a
PIC16C77 with a 4MHz crystal.
System tick rate of 100Hz.

Chapter 7 • Reference

Salvo User Manual

OSSched(): Run the Highest-Priority Eligible Task
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:

Description:
Parameters:
Returns:
Stack Usage:

Notes

Function
void OSSched ( void );
main()
salvosched.c

–
OSCLEAR_UNUSED_POINTERS,
OSCLEAR_WATCHDOG_TIMER,
OSENABLE_STACK_CHECKING, OSENABLE_STATISTICS, OSLOGGING,
OSOPTIMIZE_FOR_SPEED,

Dispatch Salvo's tasks via a cooperative
multitasking priority-based scheme.
–
–
2 if OSUSE_INLINE_OSSCHED is FALSE.
Tasks will run 2 levels below scheduler.
1 if OSUSE_INLINE_OSSCHED is TRUE.
Tasks will run 1 level below scheduler.

causes the highest-priority task currently in the eligible
queue to execute.
OSSched()

Your application must call OSInit() before calling OSSched().
Your application must repeatedly call OSSched() in order for multitasking to continue.
In the example below, OSSched() is called from within an infinite
loop.

See Also

Salvo User Manual

OSCreateTask(), OSInit(), OSStartTask()

Chapter 7 • Reference

321

Example

322

int main ( void )
{
/* OS must be initialized.
*/
OSInit();
…
/* create and start several tasks ...
*/
OSCreateTask(Task0, OSTCBP(1), TASK0_PRIORITY);
OSCreateTask(Task1, OSTCBP(2), TASK1_PRIORITY);
…
/* tasks are ready to run – begin multi*/
/* tasking.
*/
while (1) {
/* OSSched() is usually the only function */
/* called inside this never-ending loop. */
OSSched();
}
}

Chapter 7 • Reference

Salvo User Manual

OSSetCycTmrPeriod(): Set a Cyclic Timer's Period
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:

Stack Usage:
Notes

Function
OStypeErr OSSetCycTmrPeriod (
OStypeTcbP tcbP,
OstypeDelay period );

Background only
salvocyclic5.c
OSENABLE_CYCLIC_TIMERS

–
(Re-)set the specified cyclic timer's period.
tcbP: a pointer to the cyclic timer's tcb.
period: the new period.
OSNOERR if cyclic timer's period is successfully redefined.
OSERR_BAD_CT if the tcb in question does
not belong to a cyclic timer.
3

(re-)sets the cyclic timer's period regardless of whether the cyclic timer is running or not.
OSSetCycTmrPeriod()

A cyclic timer's period can be changed on-the-fly with OSSetCycTmrPeriod().
In the example below, the cyclic timer's period is changed from its
previous value to 200 system ticks. If it is already running, it will
begin running once every 200 system ticks as soon as its current
period timer times out.

See Also

Salvo User Manual

OSCreateCycTmr(), OSCycTmrRunning(), OSDestroyCycTmr(),
OSResetCycTmr(), OSStartCycTmr(), OSStopCycTmr()

Chapter 7 • Reference

323

Example

324

…
OSSetCycTmrPeriod(OSTCBP(11), 200);

Chapter 7 • Reference

Salvo User Manual

OSSetEFlag(): Set Event Flag Bit(s)
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:

Stack Usage:
Notes

Macro or Function
OStypeErr OSSetEFlag (
OStypeEcbP ecbP,
OStypeEFlag mask );

Anywhere
salvoeflag.c, salvoevent.c
OSENABLE_EVENT_FLAGS, OSEVENTS
OSLOGGING, OSENABLE_STACK_CHECKING,
OSCOMBINE_EVENT_SERVICES,
OSUSE_EVENT_TYPES

Set bits in an event flag. If any bits
change, every task waiting it is made eligible.
ecbP: a pointer to the event flag's ecb.
mask: mask of bits to be set.
OSERR_BAD_P if event flag pointer is incorrectly specified.
OSERR_EVENT_BAD_TYPE if specified event
is not an event flag.
OSERR_EVENT_CB_UNINIT if event flag's
control block is uninitialized.
OSERR_EVENT_FULL if event flag doesn't
change.
OSNOERR if event flag bits are successfully
set.
1

All tasks86 waiting an event flag are made eligible by forcing any
zeroed bits to one in the event flag via OSSetEFlag(). Upon running, each such task will either continue running or will return to
the waiting state, depending on the outcome of its call to
OS_WaitEFlag(). Thus, multiple tasks waiting a single event flag
can be made eligible simultaneously.
In the example below, two tasks are each waiting different bits of
an event flag. When those bits are set via OSSetEFlag(), both
tasks are made eligible. Each task will run when it becomes the
highest-priority eligible task.

86

Salvo User Manual

Not just the highest-priority waiting task.

Chapter 7 • Reference

325

See Also

Example

OS_WaitEFlag(), OSClrEFlag(), OSCreateEFlag(), OSReadEFlag()

#define EFLAG2_P OSECBP(4)
…
/* force TaskA() and TaskB() to wake up.
OSSetEFlag(EFLAG2_P, 0x03);
…
void TaskA ( void )
{
while (1) {
/* wait forever for bit 0 to be set
OS_WaitEFlag(EFLAG2_P, 0x01, OSALL_BITS,
OSNO_TIMEOUT);
/* clear it and continue
OSClrEFlag(EFLAG2_P, 0x01);

*/

*/

*/

…
}
}
void TaskB ( void )
{
while (1) {
OS_WaitEFlag(EFLAG2_P, 0x02, OSALL_BITS,
OSNO_TIMEOUT);
OSClrEFlag(EFLAG2_P, 0x02);
…
}
}

326

Chapter 7 • Reference

Salvo User Manual

OSSetPrio(): Change the Current Task's Priority
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

Function
void OSSetPrio (
OStypePrio prio );

Task only
salvoprio.c

–
OSENABLE_STACK_CHECKING

Change the priority of the current (running) task.
priority: the desired (new) priority for
the current task.
–
1

0 (OSHIGHEST_PRIO) is the highest priority, 15 (OSLOWEST_PRIO)
is the lowest.
Tasks can share priorities. Eligible tasks with the same priority will
round-robin schedule as long as they are the highest-priority eligible tasks.
The new priority will take effect immediately after the next context
switch.
In the example below, TaskStatusLED() is dedicated to flashing
an LED at one of two rates – 1Hz for a simple heartbeat indication,
and 25Hz for an alert indication. The system timer ticks every
10ms. When an alert is not present, it's sensible to run TaskStatusLED() at a low priority, so that other more important tasks
can run. However, when an alert condition occurs, it's imperative
that the user see the LED flash at 25Hz, so TaskStatusLED() elevates itself to a higher priority to ensure that it runs often enough
to flash the LED at 25Hz. This example assumes that all other
tasks are either delayed or waiting at any particular time. Note that
in this example TaskStatusLED() will fail to flash the LED at
25Hz if it is blocked (i.e. if there are always higher-priority tasks
running) at priority 14 when alert is TRUE.

See Also

Salvo User Manual

OS_SetPrio(), OSGetPrio(), OSGetPrioTask(), OSSetPrioTask(), OSDISABLE_TASK_PRIORITIES

Chapter 7 • Reference

327

Example

char alert = FALSE;

/* global, set & reset
/* elsewhere in code

void TaskStatusLED ( void )
{
while (1) {
/* toggle alert LED
PORT_LED ^= 0x01;

*/
*/

*/

/* if there's an alert, elevate the task's */
/* priority (to ensure that we see the LED*/
/* flash) and change the flash rate to
*/
/* 25Hz to be sure to catch the user's
*/
/* attention.
*/
if ( alert )
{
OSSetPrio(5);
OS_Delay(2);
}
/* otherwise lower the task's priority to */
/* rock-bottom and toggle the LED at 1Hz. */
else
{
OSSetPrio(OSLOWEST_PRIO);
OS_Delay(50);
}
}
}

328

Chapter 7 • Reference

Salvo User Manual

OSSetPrioTask(): Change a Task's Priority
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:

Stack Usage:
Notes

Function
OStypeErr OSSetPrioTask (
OStypeTcbP tcbP,
OStypePrio prio );

Task or Background
salvotask6.c

–
OSENABLE_STACK_CHECKING

Change the priority of the specified task.
tcbP: a pointer to the task's tcb.
prio: the desired (new) priority for the
specified task.
OSNOERR if specified task's priority was
changed successfully
OSERR if OSSetPrioTask() was unable to
change the specified task's priority.
3

OSSetPrioTask()can

change the priority of any task that is not
already destroyed or waiting an event.
0 (OSHIGHEST_PRIO) is the highest priority, 15 (OSLOWEST_PRIO)
is the lowest.
Tasks can share priorities. Eligible tasks with the same priority will
round-robin schedule as long as they are the highest-priority eligible tasks.
The new priority will take effect immediately.
In the example below, every ten minutes TaskE() elevates the priority of TaskC()for one minute, then reduces TaskC()'s priority
back to its original priority.

See Also

Salvo User Manual

OSGetPrioTask(), OSDISABLE_TASK_PRIORITIES

Chapter 7 • Reference

329

Example

/* initially, run TaskD() at priority 7.
OSCreateTask(TaskD, TASKD_P, 7);
OSCreateTask(TaskE, TASKE_P, 3);
void TaskE ( void )
{
while (1) {
/* delay ten minutes.
OS_Delay(TEN_MINUTES);

*/

*/

/* elevate TaskD()'s priority.
OSSetPrioTask(TASKD_P, 5);

*/

/* delay another minute.
OS_Delay(ONE_MINUTE);

*/

/* restore TaskD()'s priority.
OSSetPrioTask(TASKD_P, 7);

*/

}
}

330

Chapter 7 • Reference

Salvo User Manual

OSSetTicks(): Initialize the System Timer
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

Function
void OSSetTicks (
OStypeTick tick );

Anywhere
salvoticks.c
OSBYTES_OF_TICKS
OSENABLE_STACK_CHECKING

(Re-)define the current value of the system
timer (in ticks).
tick: an integer (>=0) value for the system timer.
–
1

The system timer is initialized to 0 via OSInit().
In the example below, the current value of the system timer is reset
to zero during runtime.

See Also

Salvo User Manual

OSGetTicks()

Chapter 7 • Reference

331

Example

…
/* reset system ticks to 0.
OSSetTicks(0);

*/

…

On certain targets it may be advantageous to write the current system ticks (OStimerTicks) directly instead of through OSSetTicks(). Possible scenarios include substantial function call
overhead and/or no need to manage interrupts. In the example below, the current value of the system timer is reset to zero during
runtime.
…
/* reset system ticks to 0.
disable_interrupts();
OStimerTicks = 0;
enable_interrupts();

*/

…

332

Chapter 7 • Reference

Salvo User Manual

OSSetTS(): Initialize the Current Task's Timestamp
Type:
Prototype:

Macro (invokes OSSetTSTask())
void OSSetTS (
OStypeTS timestamp );

Callable from:
Contained in:
Enabled by:
Affected by:
Description:

Task only

Parameters:
Returns:
Stack Usage:
Notes

salvodelay3.c
OSBYTES_OF_TICKS
OSENABLE_STACK_CHECKING

(Re-)define the current task's timestamp
(in ticks).
timestamp: an integer (>=0) value for the
timestamp.
–
1

When a task is created, its timestamp is initialized to an OStypeTSsized version of the system timer ticks, i.e. (OStypeTS) OStimerTicks.
In the example below, the task resets its timestamp upon starting. It
then preserves its timestamp prior to invoking OS_Delay() as part
of a hardware initialization sequence. Thereafter, it will time out
every 6 ticks relative to when it started. If OS_Delay() had been
used, it would time out every six ticks relative to when
OS_Delay() was called.
See OS_DelayTS() for more information on timestamps.

See Also

Salvo User Manual

OS_DelayTS(), OSGetTS(), OSSyncTS()

Chapter 7 • Reference

333

void Task ( void )
{
OStypeTS timestamp;

Example

/* synchronize delays with the start of this */
/* task, i.e. timestamp = now.
*/
OSSetTS((OStypeTS) OSGetTicks());
/* do various things here.
…
OS_Yield();
…

*/

/* initialize some peripheral that requires
/* a short delay. Must preserve timestamp
/* when calling OS_Delay().
…
timestamp = OSGetTS();
OS_Delay(1);
OSSetTS(timestamp);
/* continue initializing said peripheral.
…

*/
*/
*/

while (1)
{
/* as long as no more than 5 ticks have
/* passed since this task was started,
/* the task will timeout at timestamp + 6
/* ticks, and then timestamp + 12, + 18,
/* etc.
OS_DelayTS(6);
…
}

*/

*/87
*/
*/
*/
*/

}

87

334

5 ticks because of the system timer's inherent +/- 1 tick accuracy.

Chapter 7 • Reference

Salvo User Manual

OSSignalBinSem(): Signal a Binary Semaphore
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:

Description:
Parameters:
Returns:

Stack Usage:
Notes

Macro or Function
OStypeErr OSSignalBinSem (
OStypeEcbP ecbP );

Anywhere
salvobinsem.c
OSENABLE_BINARY_SEMAPHORES,
OSEVENTS
OSCALL_OSSIGNALEVENT,
OSENABLE_STACK_CHECKING,
OSCOMBINE_EVENT_SERVICES,
OSLOGGING, OSUSE_EVENT_TYPES

Signal a binary semaphore. If one or more
tasks are waiting for the semaphore, the
highest-priority task is made eligible.
ecbP: a pointer to the semaphore's ecb.
OSERR_BAD_P if binary semaphore pointer
is incorrectly specified.
OSERR_EVENT_BAD_TYPE if specified event
is not a binary semaphore.
OSERR_EVENT_FULL if binary semaphore is
already 1.
OSNOERR on success.
1

No more than one task can be made eligible by signaling a binary
semaphore.
In the example below, a binary semaphore is used to signal a waiting task. TaskWaveformGenerator() outputs an 8-bit waveform
to a DAC whenever it receives a signal to do so. The binary semaphore is initialized to 0, so TaskWaveformGenerator() remains in
the waiting state until the BINSEM_GEN_WAVEFORM is signaled elsewhere in the program, whereupon it outputs an array of 8-bit values to a port. It then resumes waiting until BINSEM_GEN_WAVEFORM
is signaled again.

See Also

Salvo User Manual

OS_WaitBinSem(), OSCreateBinSem(), OSReadBinSem(),
OSTryBinSem()

Chapter 7 • Reference

335

Example

…
#define BINSEM_GEN_WAVEFORM_P OSECBP(5)
…
OSCreateBinSem(BINSEM_GEN_WAVEFORM_P, 0);
…
/* tell waveform-generating task to create a
/* single waveform.
OSSignalBinSem(BINSEM_GEN_WAVEFORM_P);

*/
*/

…
void TaskWaveformGenerator ( void )
{
char i;
while (1) {
/* wait forever for signal to generate
/* waveform.
OS_WaitBinSem(BINSEM_GEN_WAVEFORM_P,
OSNO_TIMEOUT);
/* output waveform to DAC.
for (i = 0; i < 256; i++) {
DACPORT = WAVEFORM_TABLE[i];
}

*/
*/

*/

}
}

336

Chapter 7 • Reference

Salvo User Manual

OSSignalMsg(): Send a Message
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:

Description:

Parameters:
Returns:

Stack Usage:
Notes

Macro or Function
OStypeErr OSSignalMsg (
OStypeEcbP ecbP,
OStypeMsgP msgP );

Anywhere
salvomsg.c
OSENABLE_MESSAGES, OSEVENTS
OSCALL_OSSIGNALEVENT,
OSENABLE_STACK_CHECKING,
OSCOMBINE_EVENT_SERVICES,
OSLOGGING, OSUSE_EVENT_TYPES

Signal a message with the value specified.
If one or more tasks are waiting for the
message, the highest-priority task is made
eligible.
ecbP: a pointer to the message's ecb.
msgP: a pointer to a message.
OSERR_BAD_P if message pointer is incorrectly specified.
OSERR_EVENT_BAD_TYPE if specified event
is not a message.
OSERR_EVENT_FULL if message is already
defined.
OSNOERR on success.
1

No more than one task can be made eligible by signaling a message.
In the example below, a message is used (in place of a binary
semaphore) to control access to a shared resource, an LCD. When
either TaskDisplay() or TaskFlashWarning() needs to write to
the display, it must first acquire the display by successfully waiting
on the message MSG_LCD_RSRC. Once obtained, the task can write
to the LCD. When finished, it must release the resource by signaling the message.
displays a warning message for five seconds by writing to the display and then delaying itself for five seconds before releasing the resource. The use of a message to control
access to the LCD prevents TaskDisplay() from overwriting the
LCD while the warning message is displayed.
TaskFlashWarning()

Salvo User Manual

Chapter 7 • Reference

337

See Also
Example

OS_WaitMsg(), OSCreateMsg(), OSReadMsg(), OSTryMsg()
#define MSG_DISP_UPDATE_P OSECBP(2)
/*
#define MSG_LCD_RSRC_P
OSECBP(3)
/*
#define MSG_WARNING_P
OSECBP(4)
/*
char strLCD[LCD_LENGTH+1]; /* 1 row chars

flag
rsrc
flag
+ \0

*/
*/
*/
*/

void TaskDisplay ( voi d)
{
static OStypeMsgP msgP;
/* display is initially available to all.
*/
OSCreateMsg(MSG_LCD_RSRC_P, (OStypeMsgP) 1);
while (1) {
/* wait until display update is required
OS_WaitMsg(MSG_DISP_UPDATE_P, &msgP,
OSNO_TIMEOUT);

*/

/* wait if we can't acquire the resource.
OS_WaitMsg(MSG_LCD_RSRC_P, &msgP,
OSNO_TIMEOUT);

*/

/* write global string to display.
WriteLCD(strLCD);

*/

/* free display for others to use.
*/
OSSignalMsg(MSG_LCD_RSRC_P, (OStypeMsgP) 1);
}
}
void TaskFlashWarning ( void )
{
static OStypeMsgP msgP, msgP2;
while (1) {
/* wait for the warning ...
OS_WaitMsg(MSG_WARNING_P, &msgP,
OSNO_TIMEOUT);

*/

/* grab the LCD, locking others out.
OS_WaitMsg(MSG_LCD_RSRC_P, &msgP2,
OSNO_TIMEOUT);

*/

/* Flash warning on LCD for 5 seconds.
WriteLCD((char *)msgP);
OS_Delay(FIVE_SEC);

*/

/* refresh / restore LCD, and free it.
*/
WriteLCD(strLCD);
OSSignalMsg(MSG_LCD_RSRC_P, (OStypeMsgP) 1);
}
}

338

Chapter 7 • Reference

Salvo User Manual

OSSignalMsgQ(): Send a Message via a Message Queue
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:

Description:

Parameters:
Returns:

Stack Usage:
Notes

Macro or Function
OStypeErr OSSignalMsgQ (
OStypeEcbP ecbP,
OStypeMsgP msgP );

Anywhere
salvomsgq.c
OSENABLE_MESSAGE_QUEUES, OSEVENTS
OSCALL_OSSIGNALEVENT,
OSENABLE_STACK_CHECKING,
OSCOMBINE_EVENT_SERVICES,
OSLOGGING, OSUSE_EVENT_TYPES

Send a message to a task via the message
queue specified with ecbP. If one or more
tasks are waiting the message queue, the
highest-priority task is made eligible.
ecbP: a pointer to the message queue's ecb.
msgP: a pointer to a message.
OSERR_BAD_P if message queue pointer is
incorrectly specified.
OSERR_EVENT_BAD_TYPE if specified event
is not a message queue.
OSERR_EVENT_CB_UNINIT if the message
queue's control block is uninitialized.
OSERR_EVENT_FULL if message queue is
full.
OSNOERR on success.
1

No more than one task can be made eligible by signaling a message.
In the example below, Commands[] is a constant array of onecharacter commands. A message queue is used to send multiple
commands to a waiting task. The two successive calls to OSSignalMsg() will place the HALT ('h') and EXIT ('x') commands into
the message queue, but only if room is available. Upon arrival of
the messages, the receiving task will act accordingly.

See Also

Salvo User Manual

OS_WaitMsgQ(), OSCreateMsgQ(), OSReadMsgQ(), OSTryMsgQ()

Chapter 7 • Reference

339

Example

const char Commands[4] = { 'a', 'g', 'h', 'x' };
…
OSSignalMsgQ(MSGQ5_P, (OStypeMsgP) &Commands[2]);
OSSignalMsgQ(MSGQ5_P, (OStypeMsgP) &Commands[3]);
…

340

Chapter 7 • Reference

Salvo User Manual

OSSignalSem(): Signal a Semaphore
Type:
Prototype:

Macro or Function
OStypeErr OSSignalSem (
OStypeEcbP ecbP );

Callable from:
Contained in:
Enabled by:
Affected by:

Description:
Parameters:
Returns:

Stack Usage:
Notes

Anywhere
salvosem.c
OSENABLE_SEMAPHORES, OSEVENTS
OSBIG_SEMAPHORES,
OSCALL_OSSIGNALEVENT,
OSENABLE_STACK_CHECKING,
OSCOMBINE_EVENT_SERVICES,
OSLOGGING, OSUSE_EVENT_TYPES

Increment a counting semaphore. If one or
more tasks are waiting for the semaphore,
the highest-priority task is made eligible.
ecbP: a pointer to the semaphore's ecb.
OSERR_BAD_P if semaphore pointer is incorrectly specified.
OSERR_EVENT_BAD_TYPE if specified event
is not a semaphore.
OSERR_EVENT_FULL if semaphore is already at its maximum allowed value.
OSNOERR on success.
1

No more than one task can be made eligible by signaling a semaphore.
8-

or

16-bit

semaphores can be
OSBIG_SEMAPHORES configuration option.

selected

via

the

In the example below, a counting semaphore is used to keep track
of how many characters are waiting in the receive buffer rxBuff.
Another task that waits on SEM_RX_BUFF will remove and process
them, one at a time, from the buffer. By communicating between
the tasks with a semaphore, the tasks can run at different priorities
– TaskRx() can run at a high priority to ensure that the UART's
receive buffer is not overrun, and the processing task (which waits
on SEM_RX_BUFF) can run at a lower priority while parsing incoming command strings.

See Also

Salvo User Manual

OS_WaitSem(), OSCreateSem(), OSReadSem(), OSTrySem()

Chapter 7 • Reference

341

Example

void TaskRx ( voi d)
{
/* initially there are no Rx chars for
/* TaskRcvRsp() to process.
OSCreateSem(SEM_RX_RBUFF_P, 0);
/* The task to interpret responses is driven
/* solely by TaskRx()'s collecting incoming
/* incoming chars for it, so we'll launch
/* it from here.
OSCreateTask(TaskRcvRsp, TASK_RCV_RSP_P,
TASK_RCV_RSP_PRIO);

*/
*/

*/
*/
*/
*/

/* deal with Rx chars. */
while (1) {
/* if there are any Rx chars waiting,
*/
/* signal the command interpreter.
*/
while (SioRxQue(Port) > 0)
{
/* put new Rx char into local buffer
*/
rxBuff[rxTail] = (char) SioGetc(Port, 10);
/* massage buffer pointers
rxTail++;
rxCount++;
if (rxTail >= SIZEOF_RX_BUFF)
rxTail = 0;

*/

/* signal the command interpreter that
/* there's work to be done. In this
/* implementation we signal once for
/* every new character received.
OSSignalSem(SEM_RX_RBUFF_P);

*/
*/
*/
*/

}
/* wait a while and poll again.
OS_Delay(1);
}

*/

}

342

Chapter 7 • Reference

Salvo User Manual

OSStartCycTmr(): Start a Cyclic Timer
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:

Stack Usage:
Notes

OSStartCycTmr()

Function
OStypeErr OSStartCycTmr (
OStypeTcbP tcbP );

Background only
salvocyclic2.c
OSENABLE_CYCLIC_TIMERS

–
Start the specified cyclic timer.
tcbP: a pointer to the cyclic timer's tcb.
OSNOERR if cyclic timer is successfully
started.
OSERR_BAD_CT if the tcb in question does
not belong to a cyclic timer.
OSERR_BAD_P if the specified tcb pointer is
invalid (i.e. out-of-range).
OSERR_CT_RUNNING if the cyclic timer is
already running.
3
can only start a cyclic timer that is stopped.

If OSStartCycTmr() operates on a cyclic timer that has not yet
started (e.g. it was created with OSDONT_START_CYCTMR), then it
will begin with its delay period, followed by its normal period. If,
on the other hand, the cyclic timer was already started and then
stopped, invoking OSStartCycTmr() will cause it to restart after
its normal period.
In the example below, Task3() allows the cyclic timer to run for
400ms88 while bit 3 of the port is high, and stops the cyclic timer
from running when bit 3 is low. This is repeated indefinitely, and
requires that the cyclic timer be in continuous mode.

See Also

OSCreateCycTmr(), OSCycTmrRunning(), OSDestroyCycTmr(),
OSResetCycTmr(), OSSetCycTmrPeriod(), OSStopCycTmr()

88

Salvo User Manual

Assumes 10ms system tick period.

Chapter 7 • Reference

343

Example

void Task3( void )
{
while (1) {
OS_Delay(40);
PORT ^= 0x08;
if (PORT & 0x08) {
OSStartCycTmr(OSTCBP(1));
}
else {
OSStopCycTmr(OSTCBP(1));
}
}
}

344

Chapter 7 • Reference

Salvo User Manual

OSStartTask(): Make a Task Eligible To Run
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:

Function
OStypeErr OSStartTask (
OStypeTcbP tcbP );

Anywhere
salvotask.c

–
OSLOGGING, OSENABLE_STACK_CHECKING
Start the specified task.
tcbP: a pointer to the task's tcb.
OSNOERR if task is successfully started.
OSERR if either the specified tcb pointer is
invalid (i.e. out-of-range), or if the specified task's state is not
OSTCB_TASK_STOPPED.

Stack Usage:
Notes

3

OSStartTask() can only start
(OSTCB_TASK_STOPPED) state.

a task that is in the stopped

Starting a task simply places it into the eligible queue. It will not
run until it becomes the highest-priority eligible task.
A task that has been started is in the eligible state.
A task must be created via OSCreateTask() before it can be
started via OSStartTask().
In the example below, TaskToggleLED() is created but is only
made eligible to run via the call to OSStartTask(). Without the
call to OSStartTask(), the task would remain stopped indefinitely.

See Also

Salvo User Manual

OSCreateTask(), OSInit()

Chapter 7 • Reference

345

Example

…
/* this task toggles an LED each time it
/* runs, i.e. whenever it's the highest/* priority eligible task.
void TaskToggleLED ( void )
{
while (1) {
/* toggle LED on pin 0 of PORT B */
PORTB ^= 0x01;

*/
*/
*/

OS_Yield();
}
}
int main ( void )
{
…
/* create and start TaskToggleLED0() with
/* the lowest priority. We'll observe the
/* LED toggling when no other tasks are
/* eligible to run.
OSCreateTask(TaskToggleLED, OSTCBP(5),
OSDONT_START_TASK | OSLOWEST_PRIO);

*/
*/
*/
*/

…
OSStartTask(OSTCBP(5));
…
while (1) {
OSSched();
}
}

346

Chapter 7 • Reference

Salvo User Manual

OSStopCycTmr(): Stop a Cyclic Timer
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:

Stack Usage:
Notes

OSStopCycTmr()

Function
OStypeErr OSStopCycTmr (
OStypeTcbP tcbP );

Background only
salvocyclic3.c
OSENABLE_CYCLIC_TIMERS

–
Stop the specified cyclic timer.
tcbP: a pointer to the cyclic timer's tcb.
OSNOERR if cyclic timer is already stopped
or is successfully stopped.
OSERR_BAD_CT if the tcb in question does
not belong to a cyclic timer.
3
takes no action when the cyclic timer is already

stopped.
In the example below, the cyclic timer occupying the fifth task
control block is stopped.

See Also

Salvo User Manual

OSCreateCycTmr(), OSCycTmrRunning(), OSDestroyCycTmr(),
OSResetCycTmr(), OSSetCycTmrPeriod(), OSStartCycTmr()

Chapter 7 • Reference

347

Example

348

…
OSStopCycTmr(OSTCBP(5));

Chapter 7 • Reference

Salvo User Manual

OSStopTask(): Stop a Task
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:

Stack Usage:
Notes

Function
OStypeErr OSStopTask (
OStypeTcbP tcbP );

Task or Background
salvotask2.c

–
OSENABLE_STACK_CHECKING

Stop the specified task.
tcbP: a pointer to the task's tcb.
OSNOERR if specified task was successfully
stopped.
OSERR if OSStopTask() was unable to stop
the specified task.
3

can stop any task that is not already destroyed or
waiting an event.
OSStopTask()

A stopped task can be restarted with OSStartTask().
In the example below, TaskStopBeep() exists only to stop another
task, TaskBeep(). TaskStopBeep() waits forever for the binary
semaphore BINSEM_STOP_BEEP to be signaled. When this occurs, it
calls OSStopTask(), which stops TaskBeep(). TaskStopBeep()
then begins waiting the binary semaphore again. By setting TaskStopBeep()'s priority to be higher than TaskBeep()'s, TaskStopBeep() is able to stop TaskBeep() at the earliest opportunity.
This example also illustrates how program control can pass from
an interrupt through a task and affect another task, even if OSStopTask() is not called from an interrupt. By calling OSSignalBinSem(BINSEM_STOP_BEEP) from an ISR, TaskBeep() will be
stopped by TaskStopBeep() before its earliest opportunity to run
again.

See Also

Salvo User Manual

OSStartTask(), OS_Stop()

Chapter 7 • Reference

349

Example

350

OSCreateTask(TaskBeep,
TASK_BEEP_P,
7);
OSCreateTask(TaskStopBeep, TASK_STOPBEEP_P, 6);
OSCreateSem(BINSEM_STOP_BEEP_P, 0);
…
void TaskStopBeep ( void )
{
while (1) {
OS_WaitBinSem(BINSEM_STOP_BEEP_P,
OSNO_TIMEOUT);
OSStopTask();
}
}

Chapter 7 • Reference

Salvo User Manual

OSSyncTS(): Synchronize the Current Task's Timestamp
Type:
Prototype:

Macro (invokes OSSyncTSTask())
void OSSyncTS (
OStypeInterval interval );

Callable from:
Contained in:
Enabled by:
Affected by:
Description:

Task only

Parameters:
Returns:
Stack Usage:
Notes

salvodelay2.c

–
OSENABLE_DELAYS, OSENABLE_TICKS

Synchronize the current task's timestamp
against the current timer ticks.
interval: a signed offset relative to the
current timer ticks.
–
2

is used in conjunction with OS_DelayTS() to synchronize the current task's delays against an absolute value of the
system's timer ticks. With OSSyncTS(), you can increment or decrement the value of current task's timestamp.89
OSSyncTS()

In the example below, TaskPeriodic() begins by running every
16 system ticks. If the global variable shiftTicks is found to be
non-zero, it is copied to a local variable offset, cleared, and then
used to phase-shift TaskPeriodic() with a resolution of 1 system
tick.
See Also

OS_DelayTS(), OSGetTS(), OSSetTS()

89

Salvo User Manual

Use OSSetTS() to change the absolute value of the current task's timestamp.

Chapter 7 • Reference

351

Example

OStypeInterval shiftTicks;
…
void TaskPeriodic ( void )
{
OStypeInterval offset;

/* -15 to +15

*/

while (1) {
OS_DelayTS(16);
…
if (shift) {
disable_interrupts();
offset = shiftTicks;
shiftTicks = 0;
enable_interrupts();
OSSyncTS(offset);
}
}
}

352

Chapter 7 • Reference

Salvo User Manual

OSTimer(): Run the Timer
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:

Description:
Parameters:
Returns:
Stack Usage:

Notes

Function
void OSTimer ( void );

Foreground (preferred) or background.
salvotimer.c
OSBYTES_OF_DELAYS, OSBYTES_OF_TICKS
OSDISABLE_ERROR_CHECKING,
OSENABLE_DELAYS,
OSENABLE_STACK_CHECKING,
OSENABLE_TICKS, OSTIMER_PRESCALAR

Perform Salvo's timer-based services.
–
–
2 if OSUSE_INLINE_OSTIMER is FALSE.
1 if OSUSE_INLINE_OSTIMER is TRUE.

If delay, elapsed time and/or timeout services are desired,
OSTimer() must be called at the desired system tick rate. Context
switching and event services do not require OSTimer() to be installed.
The rate at which OSTimer() is called by your application (typically every 5-100ms) must allow sufficient time for OSTimer() to
complete its actions.
In the example below, the timer is called from within an interrupt
service routine (ISR) as a periodic event. Each time OSTimer() is
called it checks to see if any delayed or waiting tasks have timed
out, and if so, re-enters them into the eligible queue.
is very small and is easily incorporated into an ISR
without major deleterious effects.
OSTimer()

Salvo User Manual

Chapter 7 • Reference

353

Example

void interrupt ISR ( void )
{
/* OSTimer() is called on every timer0
/* interrupt.
if (TOIF) {
/* must clear timer0 interrupt flag.
TOIF = 0;
/* let Salvo handle delays, ticks
/* and timeouts.
OSTimer();

*/
*/

*/

*/
*/

}
/* handle other interrupt sources.
…

*/

}

354

Chapter 7 • Reference

Salvo User Manual

OSTryBinSem(): Obtain a Binary Semaphore if Available
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

Function
OStypeBinSem OSTryBinSem (
OStypeEcbP ecbP );

Anywhere
salvobinsem2.c
OSENABLE_BINARY_SEMAPHORES,
OSENABLE_EVENT_READING, OSEVENTS
OSCALL_OSRETURNEVENT

Returns the binary semaphore specified by
ecbP. If the semaphore is 1, reset it to 0.
ecbP: a pointer to the binary semaphore's
ecb.
Binary semaphore (0 or 1).
1

is like OS_WaitBinSem(), but it does not contextswitch the current task if the binary semaphore is not available (i.e.
has a value of 0). Therefore OSTryBinSem() can be used outside of
the current task to obtain the binary semaphore, e.g. in an ISR.
OSTryBinSem()

No error checking is performed on the ecbP parameter. Calling
OSTryBinSem() with an invalid ecbP, or an ecbP belonging to an
event other than a binary semaphore, will return an erroneous result.
In the example below, TaskC()has a higher priority than TaskD()
and obtains the binary semaphore whenever it is set to 1. Signaling
the binary semaphore does not change the state of TaskC(). As
long as TaskC() is running, TaskD() will wait forever for the binary semaphore.90

See Also

OS_WaitBinSem(), OSCreateBinSem(), OSReadBinSem(), OSSignalBinSem()

90

Salvo User Manual

This assumes that TaskD() unsuccessfully waited the binary semaphore
before TaskC() started running.

Chapter 7 • Reference

355

Example

/* priority of 3
void TaskC ( void )
{
while (1) {
if (OSTryBinSem(BINSEM2_P)) {
printf("binSem #2 was 1, now 0.\n");
}
else {
printf("binSem #2 is 0.\n");
}

*/

OS_Yield();
…
}
}
/* priority of 9 (lower)
void TaskD ( void )
{
while (1) {
OS_WaitBinSem(BINSEM2_P,
OSNO_TIMEOUT);
…
}
}

356

Chapter 7 • Reference

*/

Salvo User Manual

OSTryMsg(): Obtain a Message if Available
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

Function
OStypeMsg OSTryMsg (
OStypeEcbP ecbP );

Anywhere
salvomsg2.c
OSENABLE_MESSAGES,
OSENABLE_EVENT_READING, OSEVENTS
OSCALL_OSRETURNEVENT

Returns a pointer to the message specified
by ecbP. If the message exists, the message's own pointer is cleared.
ecbP: a pointer to the message's ecb.
Message pointer.
1

is like OS_WaitMsg(), but it does not context-switch
the current task if the message is not available (i.e. the message
pointer has a value of 0). Therefore OSTryMsg() can be used outside of the current task to obtain the message, e.g. in an ISR.
OSTryMsg()

No error checking is performed on the ecbP parameter. Calling
OSTryMsg() with an invalid ecbP, or an ecbP belonging to an
event other than a binary semaphore, will return an erroneous result.
Waiting on a message (i.e. via OS_WaitMsg()) is not permitted
within an interrupt service routine. In the example below,
OSTryMsg() is used within the ISR in order to obtain a message
without waiting. Regardless of whether or not a message was
available, the message will be empty at the end of the ISR.

See Also

Salvo User Manual

OS_WaitMsg(), OSCreateMsg(), OSReadMsg(), OSSignalMsg()

Chapter 7 • Reference

357

Example

void interrupt myISR ( void )
{
OStypeMsgP msgP;
/* get message pointer (may be 0).
msgP = OSTryMsg(MSG3_P);
while (1) {
/* do something with the message.
…
}
else
{
/* message wasn't available.
…
}
…

*/

*/

*/

}

358

Chapter 7 • Reference

Salvo User Manual

OSTryMsgQ(): Obtain a Message from a Message Queue
if Available
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:

Parameters:
Returns:
Stack Usage:
Notes

Function
OStypeMsgQ OSTryMsgQ (
OStypeEcbP ecbP );

Anywhere
salvomsgq2.c
OSENABLE_MESSAGE_QUEUES,
OSENABLE_EVENT_READING, OSEVENTS
OSCALL_OSRETURNEVENT

Returns a pointer to the first available
message in the message queue specified
by ecbP. If the message queue contains
any messages, remove the message from
the queue.
ecbP: a pointer to the message queue's ecb.
Message pointer.
1

is like OS_WaitMsgQ(), but it does not contextswitch the current task if the message queue is empty. Therefore
OSTryMsgQ() can be used outside of the current task to obtain the
message in the message queue, e.g. in an ISR.
OSTryMsgQ()

No error checking is performed on the ecbP parameter. Calling
OSTryMsgQ() with an invalid ecbP, or an ecbP belonging to an
event other than a binary semaphore, will return an erroneous result.
In the example below, after each call to the scheduler, a char message is removed from a message queue and then re-inserted. As
long as no services involving this message queue are called from
within an interrupt, this will rotate the order of the messages in the
message queue indefinitely. For example, a message queue containing the four single-character messages 's', 't', 'o' and 'p' becomes
't', 'o', 'p' and 's'.

See Also

Salvo User Manual

OS_WaitMsgQ(), OSCreateMsgQ(), OSReadMsgQ(), OSSignalMsgQ()

Chapter 7 • Reference

359

Example

OStypeMsgP msgP;
…
while (1) {
OSSched();
msgP = OSTryMsgQ(MSGQ3_P);
if (msgP) {
printf("removed message %c from msgQ.\n",
*(char *) msgP);
OSSignalMsgQ(MSGQ3_P, msgP);
printf("re-inserted message into msgQ.\n");
}
}

360

Chapter 7 • Reference

Salvo User Manual

OSTrySem(): Obtain a Semaphore if Available
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

Function
OStypeSem OSTrySem (
OStypeEcbP ecbP );

Anywhere
ssalvosem2.c
OSENABLE_SEMAPHORES,
OSENABLE_EVENT_READING, OSEVENTS
OSCALL_OSRETURNEVENT

Returns the semaphore specified by ecbP.
If the semaphore is non-zero, decrement
it.
ecbP: a pointer to the semaphore's ecb.
Semaphore.
1

is like OS_WaitSem(), but it does not context-switch
the current task if the semaphore is not available (i.e. has a value of
0). Therefore OSTrySem() can be used outside of the current task
to obtain the semaphore, e.g. in an ISR.
OSTrySem()

No error checking is performed on the ecbP parameter. Calling
OSTrySem() with an invalid ecbP, or an ecbP belonging to an
event other than a binary semaphore, will return an erroneous result.
In the example below, OSTrySem() is used by FlushBuffer()91 to
flush a buffer that is managed through a counting semaphore. Afterwards, i holds the count of the items that were in the the buffer
before it was flushed.

See Also

OS_WaitSem(), OSCreateSem(), OSReadSem(), OSSignalSem()

91

Salvo User Manual

Note that FlushBuffer() is a simple function, and not a task. The flushing
operation could also be performed in a task.

Chapter 7 • Reference

361

Example

/* buffer is initially empty.
OSCreateSem(SEM2_P, 0);
…
void FlushBuffer ( void )
{
char i;

*/

/* count and remove the buffer's contents.
i = 0;
while (OSTrySem(SEM2_P)) {
i++;
}

*/

}

362

Chapter 7 • Reference

Salvo User Manual

Additional User Services
OSAnyEligibleTasks (): Check for Eligible Tasks
Type:
Declaration:
Callable from:

Macro

Contained in:
Enabled by:
Affected by:
Description:

salvomac.h

Parameters:
Returns:
Stack Usage:
Notes

OSAnyEligibleTasks()

Outside OSSched() (background) or inside
a task or its subroutines.
–
–
Detect if any tasks are currently eligible to
run.
–
TRUE if one or more tasks are eligible,
FALSE otherwise.
0

cannot predict when waiting and/or delayed tasks will become eligible. This must be considered when
using OSAnyEligibleTasks().
OSAnyEligibleTasks()

OSAnyEligibleTasks()

returns FALSE if a task is running and no

tasks are eligible.
In the first example below, a Salvo application's main loop has
been modified to run an alternative process (e.g. some legacy code
written in assembler) in addition to the scheduler. This alternative
process must terminate within a short time in order to avoid problems scheduling tasks. By invoking the alternative process only
when no tasks are eligible, it can "steal cycles" that the scheduler
does not currently need.
In the second example, a user function (not a task) is called only
when the system is idling, i.e. when tasks are eligible to run. This
idling function must execute quickly so as not to affect task execution.
Note that in both examples, Salvo's idling hook could be used in
place of OSAnyEligibleTasks() if it were not already in use.

Salvo User Manual

Chapter 7 • Reference

363

Example #1

int main ( void )
{
…
while (1) {
OSSched();
if (!OSAnyEligibleTasks()) {
/* do alternative background process */
#asm
#include "mystuff.asm"
#endasm
}
}
}

Example #2

int main ( void )
{
…
while (1) {
OSSched();
if (!OSAnyEligibleTasks())
DoWhileIdling();
}
}

364

Chapter 7 • Reference

Salvo User Manual

OScTcbExt0|1|2|3|4|5, OStcbExt0|1|2|3|4|5(): Return a Tcb
Extension
Type:
Declaration:
Callable from:

Macro
OScTcbExt0|1|2|3|4|5, OStcbExt0|1|2|3|4|5(tcbP)
OScTcbExt0|1|2|3|4|5 should only be called
from the task level. OStcbExt0|1|2|3|4|5() can be called from any-

where.
Contained in:
Enabled by:
Affected by:
Description:

Parameters:
Returns:
Stack Usage:

salvomac.h
OSENABLE_TCBEXT0|1|2|3|4|5

–
OScTcbExt0|1|2|3|4|5

returns the specified
tcb extension of the current task. OStcbExt0|1|2|3|4|5 returns the specified tcb
extension of the specified task.
–
Tcb extension.
0.

Notes

These macros are used to obtain the desired tcb extension from the
task's tcb.

See Also

OSENABLE_TCBEXT0|1|2|3|4|5, OSTYPE_TCBEXT0|1|2|3|4|5

Salvo User Manual

Chapter 7 • Reference

365

Example
void CommTask ( void )
{
/* ascertain mode at startup */
switch (OScTcbExt3) {
case SW_HANDSHAKING:
while (1) {
/* do comms w/ XON/XOFF */
OpenSWUART();
…
OS_Yield();
}
break;
case HW_HANDSHAKING:
while (1) {
/* do comms w/ DTR & CTS */
OpenHWUART();
…
OS_Yield();
}
break;
default:
break;
}
}
int main ( void )
{
…
/* we want hardware handshaking … */
OSCreateTask(CommTask, OSTCBP(7), 5);
OStcbExt3(OSTCBP(7)) = HW_HANDSHAKING;
…
while (1) {
OSSched();
}
}

366

Chapter 7 • Reference

Salvo User Manual

OSCycTmrRunning(): Check Cyclic Timer for Running
Type:
Prototype:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:

Stack Usage:
Notes

OSCycTmrRunning()

Function
OStypeErr OSCycTmrRunning (
OStypeTcbP tcbP );

Background only
salvocyclic7.c
OSENABLE_CYCLIC_TIMERS

–
Detect if cyclic timer is running or not.
tcbP: a pointer to the cyclic timer's tcb.
FALSE if cyclic timer is stopped, or if the
tcb in question does not belong to a cyclic
timer.
TRUE if cyclic timer is running.
1
indicates whether or not a cyclic timer is run-

ning.
In the example below, a task waits for a signal to restart a cyclic
timer. When that signal is received, the cyclic timer is stopped and
restarted. Regardless of how close it was previously to timing out,
it will now time out in its normal period.

See Also

Salvo User Manual

OSCreateCycTmr(), OSCycTmrPeriod(), OSDestroyCycTmr(),
OSResetCycTmr(), OSStartCycTmr(), OSStopCycTmr()

Chapter 7 • Reference

367

Example

368

…
if (OSCycTmrRunning(OSTCBP(3)))
{
/* do something if cyclic timer is running.
}

Chapter 7 • Reference

*/

Salvo User Manual

OSProtect(), OSUnprotect(): Protect Services Against
Corruption by ISR
Type:
Declaration:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

Macro
OSProtect(), OSUnprotect()

Background
salvoportXyz.h

–
–
Disable or enable interrupts, respectively,
if such control is required on given target.
–
n/a
0, unless defined otherwise.

When compiling for a target that does not have a software stack,
certain steps must be taken to protect servicse with multiple callgraphs. By calling OSProtect() immediately before each such
service, and OSUnprotect() immediately thereafter, the service is
protected against any corruption that might occur if an interrupt
that calls the service were to occur simultaneously.
These macros are empty for all targets whose compilers pass parameters on a stack. To ensure cross-platform compatibility, all
Salvo applications should use OSProtect() and OSUnprotect()
as specified, even if these macros are empty for a particular compiler.

Warning Because a stackless compiler may overlay the local /
parameter areas of one or more services with multiple callgraphs,
OSProtect() and OSUnprotect() should be used around every
service whose OSCALL_XYZ is set to OSFROM_ANYWHERE.
In the example below, OSSignalBinSem() is called from mainline
code and from within an ISR. Therefore OSProtect() and OSUnprotect() are required in the mainline code.

See Also

OSCALL_OSXYZ, OSFROM_ANYWHERE, OSDi(), OSEi(), Salvo Com-

piler Reference Manuals

Salvo User Manual

Chapter 7 • Reference

369

Example

void TestCode ( void )
{
…
if (PutTx1Buff(data)) {
OSProtect();
OSSignalBinSem(BINSEM_TXBUFF_P);
OSUnprotect();
}
…
}
void interrupt ISR ( void )
{
…
if (txState == TXSTATE_DONE) {
txState = TXSTATE_IDLE;
OSSignalBinSem(BINSEM_TXDONE_P);
}
…
}

370

Chapter 7 • Reference

Salvo User Manual

OSTaskStopped(): Check whether Task has Stopped
Type:
Declaration:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

OSTaskStopped()

Macro
OSTaskStopped ( OStypeTcbP tcbP );

anywhere
salvomac.h

–
–
Detect if the current task is stopped.
–
TRUE if task is stopped, FALSE otherwise.
0
does not check the validity of the task handle

passed to it.
In the example below, the task pointed to by TASK_FREQ_P is (re-)
started if already stopped. Otherwise it is stopped.

See Also

Example

–

…
OS_WaitSem(SEM_CMD_CHAR_P, OSNO_TIMEOUT);
if (cmd = getchar1()) {
switch (tolower((char) cmd)) {
…
case 'f':
if (OSTaskStopped(TASK_FREQ_P)) {
OSStartTask(TASK_FREQ_P);
user_msg(STR_TASK_CMD STR_TABS "f:" \
" Started task_freq().");
}
else {
OSStopTask(TASK_FREQ_P);
user_msg(STR_TASK_CMD STR_TABS "f:" \
" Stopped task_freq().");
}
break;
…
}
}

Salvo User Manual

Chapter 7 • Reference

371

OSTimedOut(): Check for Timeout
Type:
Declaration:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

Macro
OSTimedOut()

Task only
salvomac.h
OSENABLE_TIMEOUTS

–
Detect if the current task timed out waiting
for an event.
–
TRUE if a timeout occurred, FALSE otherwise.
0

By specifying a non-zero timeout in OS_WaitBinSem(),
OS_WaitMsg(), OS_WaitMsgQ() or OS_WaitSem(), you can control program execution in the case where an event does not occur
within a specified number of system ticks. This is very useful in
handling errors that may result from expected events failing to occur.
Once a timeout occurs, the task is no longer waiting the event. The
fact that a timeout occurred only indicates that the task did not successfully wait the event in the allotted time … it does not in any
way reflect on the current status of the event, or on other tasks
waiting the event.
In the example below, a bidirectional communications channel is
used to send commands and receives a response (acknowledgments) for each command sent. A new command can be sent only
after the acknowledgment for the previous command has been received. By specifying a response timeout (RSP_TIMEOUT) that's larger than the expected time for the receiver to respond to a
command, TaskTx() can conditionally wait for the response instead of waiting indefinitely if the acknowledgment never arrives.
When a timeout occurs, a task's execution resumes where it was
originally waiting for the event, and the Salvo function OSTimedOut() returns TRUE until the task context-switches back to the
scheduler. TaskTx() checks to see if a timeout occurred after it
acquires the message.

372

Chapter 7 • Reference

Salvo User Manual

See Also

Example

OS_WaitBinSem(), OS_WaitMsg(), OS_WaitMsgQ(),
OS_WaitSem()

void TaskTx ( void )
{
static OStypeMsgP msgP;
/* No cmds have been sent yet, so no
*/
/* responses have been received.
*/
OSCreateMsg(MSG_RSP_RCVD_P, (OStypeMsgP) 0);
while (1) {
/* send command to receiver.
…
/* wait here until response has been
/* received for the command we sent.
/* if we timed out, reset the expected
/* response, STOP, clear the buffer and
/* tell the user.
OS_WaitMsg(MSG_RSP_RCVD_P, &msgP,
RSP_TIMEOUT);

*/

*/
*/
*/
*/
*/

if (OSTimedOut()) {
FlushCmdInterpreter();
setSTOP();
txBuff[0] = 0;
FlashMsg(&msgBadComms);
}
/* continue processing outgoing commands.
…

*/

Chapter 7 • Reference

373

}
}

Salvo User Manual

OSVersion(), OSVERSION: Return Version as Integer
Type:
Declaration:
Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

Macro
OSVersion(), OSVERSION

Anywhere
salvover.h

–
–
Returns the version number.
–
Returns the version number as an unsigned
integer.
0

three version number fields: OSVER_MAJOR,
and OSVER_SUBMINOR. Each field is a numeric integer constant. They are combined into a single symbol, OSVERSION,
in the following manner:

Salvo

uses

OSVER_MINOR

OSVERSION = OSVER_MAJOR
* 100
+ OSVER_MINOR
* 10
+ OSVER_SUBMINOR

Therefore in v3.0.0, OSVERSION equals 300.
OSVersion()

374

is identical to OSVERSION.

Chapter 7 • Reference

Salvo User Manual

Example

Salvo User Manual

printf("Salvo version: %d (v%c.%c.%c)\n",
'0' + OSVER_MAJOR,
'0' + OSVER_MINOR,
'0' + OSVER_SUBMINOR,
OSVersion());

Chapter 7 • Reference

375

User Macros
This section describes the Salvo user macros that you will use to
build your multitasking application.
The macros are described below.

OSECBP(), OSEFCBP(),OSMQCBP(), OSTCBP(): Return a
Control Block Pointer
Type:
Declaration:

Callable from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

Macro
OSECBP( index )
OSEFCBP( index )
OSMQCBP( index )
OSTCBP( index )

n/a
salvo.h

–
–
Shorthand for pointer to specified control
block.
index: an index from 1 to OSEVENTS, 1 to
OSEVENT_FLAGS, 1 to OSMESSAGE_QUEUES
or 1 to OSTASKS, respectively.
pointer to (i.e. address of) desired event,
message queue or task control block, respectively.
n/a

RAM memory for control blocks is allocated at compile time using
the OSEVENTS, OSEVENT_FLAGS, OSMESSAGE_QUEUES and OSTASKS
configuration options. Instead of obtaining the compile-time address of a particular event, event flag, message queue or task control block by using
&OSecbArea[i-1]
&OsefcbArea[i-1]
&OSmqcbArea[i-1]
&OStcbArea[i-1]

you can and should use these macros.

376

Chapter 7 • Reference

Salvo User Manual

Example

#define TASK1_P
#define TASK2_P
#define SEM1_P

OSTCBP(1)
OSTCBP(2)
OSECBP(1)

…
OSCreateTask(Task1, TASK1_P, 7);
…
OSCreateSem(SEM1_P, 14);
…

Salvo User Manual

Chapter 7 • Reference

377

User-Defined Services
OSDisableIntsHook(), OSEnableIntsHook(): Interruptcontrol Hooks
Type:
Declaration:
Called from:
Contained in:

Function
void OSDisableIntsHook( void )
void OSEnableIntsHook( void )
OSDi() and OSEi()
salvo.h if left undefined, otherwise in

user source code.
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

OSENABLE_INTERRUPT_HOOKS

–
User-defined.
–
–
Dependent on user definition.

You may find it useful or necessary to perform certain operations
coincident with Salvo's disabling and (re-)enabling of interrupts
during critical sections of code.
If these functions are enabled via OSENABLE_INTERRUPT_HOOKS,
OSDisableIntsHook() is called immediately after disabling
interrupts, and OSEnableIntsHook() is called immediately before
(re-)enabling interrupts. Therefore each function is called with
interrupts disabled.
By default, these functions are undefined.
In the example below, two separate counters, diCounter and eiCounter, are used to count the number of times that Salvo disables
and (re-)enables interrupts, respectively.

See Also

378

OSDi(), OSEi()

Chapter 7 • Reference

Salvo User Manual

Example

unsigned long int diCounter, eiCounter;
…
void OSDisableIntsHook( void )
{
diCounter++;
}
void OSEnableIntsHook( void )
{
eiCounter++;
}

Salvo User Manual

Chapter 7 • Reference

379

OSIdlingHook(): Idle Function Hook
Type:
Declaration:
Called from:
Contained in:
Enabled by:
Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

Function
void OSIdlingHook( void )
OSSched()

User source code, called from sched.c.
OSENABLE_IDLING_HOOK

–
User-defined.
–
–
Dependent on user definition.

Salvo's scheduler normally runs in a tight loop when no tasks are
eligible to run, i.e. when it is idling. By defining an idle function
and setting OSENABLE_IDLING_HOOK to TRUE, you can do something useful while the system is idling. Your idle function should
be short and fast, as time spent in it delays the operation of the
scheduler.
By default, OSIdlingHook() is undefined. However, Salvo libraries configured for the idling hook contain a dummy OSIdlingHook() function to avoid linker errors when the user fails to define
a OSIdlingHook().
In the example below, the least significant bit on an output port is
toggled whenever there are no eligible or running tasks.

380

Chapter 7 • Reference

Salvo User Manual

Example

Salvo User Manual

void OSIdlingHook( void )
{
PORTB ^= 0x01;
}

Chapter 7 • Reference

381

OSSchedDispatchHook(), OSSchedEntryHook(),
OSSchedReturnHook(): Scheduler Hooks
Type:
Declaration:
Called from:
Contained in:
Enabled by:

Affected by:
Description:
Parameters:
Returns:
Stack Usage:
Notes

Function
void OSSchedDispatchHook( void )
void OSSchedEntryHook( void )
void OSSchedReturnHook( void )
OSSched()

User source code, called from sched.c.
OSENABLE_OSSCHED_DISPATCH_HOOK,
OSENABLE_OSSCHED_ENTRY_HOOK, and
OSENABLE_OSSCHED_RETURN_HOOK,
respectively
–
User-defined.
–
–
Dependent on user definition.

It may be useful when debugging a Salvo application to have runtime information on the scheduler's behavior. These hooks are provided so that user-defined functions can be invoked at strategic
times within OSSched()'s execution.
OSSchedEntryHook() is called immediately
scheduler. OSSchedDispatchHook() is called

upon entry into the
immediately prior to
dispatching the current eligible task, with interrupts enabled and
OScTcbP pointing to the current task's control block. OSSchedReturnHook() is called immediately after the current task returns
(yields) to the scheduler ... the current task can be in any state, interrupts are enabled, and OScTcbP still points to the current task's
control block.
When the system is idling (i.e. there are no eligible tasks), neither
nor OSSchedReturnHook() will be
called.

OSSchedDispatchHook()

By default, OSSchedDispatchHook(), OSSchedEntryHook()
OSSchedReturnHook() are all undefined.

and

In the example below, PORTB[5] is set just prior to dispatching the
current task, and is cleared after the current task yields back to the
scheduler. The time that PORTB[5] is high represents the dispatch
overhead in OSSched(), plus the task's execution time. The time

382

Chapter 7 • Reference

Salvo User Manual

between successive rising edges of PORTB[5] represents the instantaneous context-switching speed of the application.

Example

void OSSchedDispatchHook ( void )
{
PORTB |= 0x20;
}
void OSSchedReturnHook ( void )
{
PORTB &= ~0x20;
}

Salvo User Manual

Chapter 7 • Reference

383

Return Codes
Many Salvo user services have return codes to indicate whether or
not they were called successfully. Some are listed below. See the
individual user service descriptions for more information on return
codes.
OSNOERR:
OSERR:
OSERR_TASK_BAD_P:
OSERR_EVENT_NA:
OSERR_EVENT_FULL:
OERR_EVENT_CB_UNINIT:
OSERR_TIMEOUT:

No error.
An error was encountered while executing
the user service.
An invalid pointer was passed to the user
service.
The specified event was not available
The specified event (e.g. message) is already full.
The specified control block (e.g. for message queues or event flags) has not yet
been initialized.
The current task has timed out while waiting for an event.
Table 5: Return Codes

Salvo Defined Types
The following types are defined for use with Salvo user services.
Because the types are affected by configuration options, when interfacing to Salvo user services you should always declare variables with these defined types. Failing to do so is likely to result in
unpredictable behavior.
Salvo has two classes of predefined types: those where the memory
(RAM) location of the object is not specified (normal,
OStypeXyz), and those where the location is explicitly specified
(qualified, OSgltypeXyz). The need for both types arises on those
processors with banked RAM. If your target processor has a single
linear RAM space, the two types are identical. When in doubt, use
the qualified type if one exists.
The normal types are used in the Salvo source code when declaring
auto variables, parameters and function return values. You can also
use the normal types when declaring your own local variables (e.g.

384

Chapter 7 • Reference

Salvo User Manual

message pointers of type OStypeMsgP), and when typecasting (e.g.
OSSignalMsg(MSGP, (OStypeMsgP) &array[2]));
The qualified types are used to declare Salvo's global variables,
and are also provided so that you can properly declare your own
global variables for Salvo, e.g. message queues – OSgltypeMsgQP
MsgQBuff[SIZEOF_MSGQ].

Tip Refer to the Salvo source code for examples of when to use
normal or qualified Salvo types.
The normal types are:
OStypeBinSem:
OStypeBitField:

binary semaphore: OStypeBoolean
size of bit fields in structures: int or char,
depending on
OSUSE_CHAR_SIZED_BITFIELDS

OStypeBoolean:
OStypeCount:
OStypeDelay:
OStypeDepth:
OStypeEcb:
OStypeEfcb:
OStypeEFlag:
OStypeErr:
OStypeEType:
OStypeID:
OStypeInt8u:
OStypeInt16u:
OStypeInt32u:
OStypeInterval:
OStypeMqcb:
OStypeMsg:
OStypeMsgQSize:
OStypeOption:

Salvo User Manual

Boolean: FALSE (0) or TRUE (non-zero)
counter: OStypeInt8u/16u/32u, depending on OSBYTES_OF_COUNTS
delay: OStypeInt8u/16u/32u, depending
on OSBYTES_OF_DELAYS
stack depth counter: OStypeInt8u
event control block: structure
event flag control block: structure
event flag: OStypeInt8u/16u/32u, depending on configuration
function return code or error / warning /
timeout counter: OStypeInt8u
event type: OStypeInt8u
object ID: OStypeInt8u
integer: 8-bit, unsigned
integer: 16-bit, unsigned
integer: 32-bit, unsigned
interval: OStypeInt8/16/32, depending on
OSBYTES_OF_DELAYS

message queue control block: structure
message: void or const, depending on
OSMESSAGE_TYPE

number of messages in a message queue:
OStypeInt8u

generic option: OStypeInt8u

Chapter 7 • Reference

385

OStypePrio:
OStypePS:
OStypeSem:

OStypeState:
OStypeStatus:
OStypeTcb:
OStypeTcbExt:
OStypeTick:
OStypeTS:

task priority: OStypeInt8u, values from 0
to 15 are defined
timer prescalar: OStypeInt8u/16u/32u,
depending on configuration
semaphore: OStypeInt8u or
OStypeInt16u, depending on configuration
task state: OStypeInt8u, values from 0 to 7
are defined
task status: bitfields of type OStypeInt8u
for a task's running bit, state and priority
task control block: structure
tcb extension: void *, user-(re-)definable
timer ticks: OStypeInt8u/16u/32u, depending on configuration
timestamp: OStypeInt8u/16u/32u,
depending on configuration of
OSBYTES_OF_DELAYS
Table 6: Normal Types

The normal pointer types are:
OStypeCharEcbP:
OStypeCharTcbP:
OStypeEcbP:
OStypeEfcbP:
OStypeMqcbP:
OStypeMsgP:
OStypeMsgPP:
OStypeMsgQPP:
OStypeTcbP:
OStypeTcbPP:
OStypeTFP:

pointer to banked (OSLOC_ECB) char
pointer to banked (OSLOC_TCB) char
pointer to banked (OSLOC_ECB) event control block
pointer to banked (OSLOC_EFCB) event flag
control block
pointer to banked (OSLOC_MQCB) message
queue control block
pointer to message
pointer to pointer to message
pointer to banked (OSLOC_MSGQ) pointer to
message
pointer to banked (OSLOC_TCB) task control
block
pointer to banked (OSLOC_ECB) pointer to
banked (OSLOC_TCB) task control block
pointer to (task) function
Table 7: Normal Pointer Types

The qualified types are:

386

Chapter 7 • Reference

Salvo User Manual

OSgltypeCount:
OSgltypeDepth:
OSgltypeEcb:
OSgltypeEfcb:
OSgltypeErr:
OSgltypeGlStat:
OSgltypeLogMsg:
OSgltypeMqcb:
OSgltypePS:
OSgltypeTcb:
OSgltypeTick:

qualified OStypeCount: banked
(OSLOC_COUNT) counter
qualified OStypeDepth: banked
(OSLOC_DEPTH) stack depth counter
qualified OStypeEcb: banked (OSLOC_ECB)
event control block
qualified OStypeEfcb: banked
(OSLOC_EFCB) event flag control block
qualified OStypeErr: banked (OSLOC_ERR)
error counter
qualified OStypeGlStat: banked
(OSLOC_GLSTAT) global status bits
qualified char: banked (OSLOC_LOGMSG)
log message character or string
qualified OStypeMqcb: banked
(OSLOC_MQCB) message queue control
block
qualified OStypePS: banked (OSLOC_PS)
timer prescalar
qualified OStypeTcb: banked (OSLOC_TCB)
task control block
qualified OStypeTick: banked
(OSLOC_TICK) system ticks
Table 8: Qualified Types

The qualified pointer types are:

OSgltypeCTcbP:

OSgltypeEcbP:

OSgltypeMsgQP:

OSgltypeSigQP:

OSgltypeTcbP:

qualified OStypeTcbP: banked
(OSLOC_CTCB) pointer to banked task control block
qualified OStypeEcbP: banked
(OSLOC_ECB) pointer to banked event control block
qualified OStypeMsgP: banked
(OSLOC_MSGQ) pointer to message
qualified OStypeTcbP: banked
(OSLOC_SIGQ) pointer to banked task control block
qualified OStypeTcbP: banked
(OSLOC_ECB) pointer to banked task control block
Table 9: Qualified Pointer Types

Salvo User Manual

Chapter 7 • Reference

387

Note When declaring pointers using predefined Salvo pointer
types on targets that have banked RAM, always declare each
pointer on its own, like this:
OStypeMsgP msgP1;
OStypeMsgP msgP2;

Failing to do so (i.e. declaring multiple pointers by commadelimiting them on one line) will result in an improper declaration.

Salvo Variables
Salvo's global variables (declared in salvomem.c) are listed below.
The variable, the qualified type corresponding to the variable and a
description of the variable are listed for each one. Advanced programmers may find it useful to read these variables during runtime
or while debugging. In some development environments (e.g. Microchip MPLAB), these variable names will be available for symbolic debugging.

Warning Do not modify any of these variables during runtime –
unpredictable results may occur.

OScTcbP

OSgltypeCTcbP

OSctxSws

OSgltypeCount

OSdelayQP

OSgltypeDelayQP

OSecbArea[]

OSgltypeEcb

OSefcbArea[]

OSgltypeEfcb

OSeligQP

OSgltypeEligQP

OSerrs

OSgltypeErr

OSframeP

OsgltypeFrameP

OSglStat

OSgltypeGlStat

92

388

pointer to current
task's task control
block
context switch
counter
pointer to delay
queue
event control block
storage
event flag control
block storage
pointer to eligible
queue
runtime error
counter
frame pointer92
global status bits

Used in some Salvo context switcher to assist in stack frame operations.

Chapter 7 • Reference

Salvo User Manual

OSidleCtxSws

OSgltypeCount

OSlogMsg[]

OSgltypeLogMsg

OSlostTicks

OSgltypeLostTick

OSmaxStkDepth

OSgltypeDepth

OSmqcbArea[]

OSgltypeMqcb

OSrtnAddr

OSgltypeTFP

OSsigQinP,
OSsigQoutP

OSgltypeSigQP

OSstkDepth

OSgltypeDepth

OStcbArea[]

OSgltypeTcb

OStimerTicks

OSgltypeTick

OStimerPS

OSgltypePS

OStimeouts

OSgltypeErr

OSwarns

OSgltypeErr

idle function calls
counter
log (debug) message
string
accumulated timer
ticks
maximum stack
depth achieved by
Salvo functions
message queue control block storage
task's return / resume
address
signaled event queue
insert and removal
pointers
current stack depth
of Salvo function
task control block
storage
system timer ticks
counter
runtime timer prescalar
runtime timeout
counter
runtime warning
counter

Table 10: Salvo Variables

Salvo Source Code
The Salvo source code is organized into files that handle tasks, resources, queues, data structures, utility functions, the monitor, and
the many #defines that are used to configure Salvo for a variety
of applications.
You can always review the source code if the manual is unable to
answer your question(s). Modifying the source code is not recommended, as your application may not run properly when compiled
with a later release of Salvo. Where applicable, user #defines and
hooks for user functions are provided so that you can use Salvo in

Salvo User Manual

Chapter 7 • Reference

389

conjunction with features that are not yet supported in the current
release.
Salvo's source (*.h and *.c) files are listed below.
Pumpkin\Salvo\Inc\salvo.h
Pumpkin\Salvo\Inc\salvoadc.h
Pumpkin\Salvo\Inc\salvocri.h
Pumpkin\Salvo\Inc\salvoctx.h
Pumpkin\Salvo\Inc\salvodef.h
Pumpkin\Salvo\Inc\salvofpt.h
Pumpkin\Salvo\Inc\salvolbo.h
Pumpkin\Salvo\Inc\salvolib.h
Pumpkin\Salvo\Inc\salvoloc.h
Pumpkin\Salvo\Inc\salvolvl.h
Pumpkin\Salvo\Inc\salvomac.h
Pumpkin\Salvo\Inc\salvomcg.h
Pumpkin\Salvo\Inc\salvomem.h
Pumpkin\Salvo\Inc\salvompt.h
Pumpkin\Salvo\Inc\salvoocp.h
Pumpkin\Salvo\Inc\salvoprg.h
Pumpkin\Salvo\Inc\salvopsh.h
Pumpkin\Salvo\Inc\salvoscb.h
Pumpkin\Salvo\Inc\salvoscg.h
Pumpkin\Salvo\Inc\salvostr.h
Pumpkin\Salvo\Inc\salvotyp.h
Pumpkin\Salvo\Inc\salvover.h
Pumpkin\Salvo\Inc\salvowar.h
Pumpkin\Salvo\Src\salvobinsem.c
Pumpkin\Salvo\Src\salvobinsem2.c
Pumpkin\Salvo\Src\salvochk.c
Pumpkin\Salvo\Src\salvocyclic.c
Pumpkin\Salvo\Src\salvocyclic2.c
Pumpkin\Salvo\Src\salvocyclic3.c
Pumpkin\Salvo\Src\salvocyclic4.c
Pumpkin\Salvo\Src\salvocyclic5.c
Pumpkin\Salvo\Src\salvocyclic6.c
Pumpkin\Salvo\Src\salvocyclic7.c
Pumpkin\Salvo\Src\salvodebug.c
Pumpkin\Salvo\Src\salvodelay.c
Pumpkin\Salvo\Src\salvodelay2.c
Pumpkin\Salvo\Src\salvodelay3.c
Pumpkin\Salvo\Src\salvodestroy.c
Pumpkin\Salvo\Src\salvoeflag.c
Pumpkin\Salvo\Src\salvoeflag2.c
Pumpkin\Salvo\Src\salvoeid.c
Pumpkin\Salvo\Src\salvoevent.c
Pumpkin\Salvo\Src\salvohook_idle.c
Pumpkin\Salvo\Src\salvohook_interrupt.c
Pumpkin\Salvo\Src\salvohook_wdt.c
Pumpkin\Salvo\Src\salvoidle.c
Pumpkin\Salvo\Src\salvoinit.c
Pumpkin\Salvo\Src\salvoinit2.c
Pumpkin\Salvo\Src\salvoinit3.c
Pumpkin\Salvo\Src\salvoinit4.c

390

Chapter 7 • Reference

Salvo User Manual

Pumpkin\Salvo\Src\salvointvl.c
Pumpkin\Salvo\Src\salvolicense.c
Pumpkin\Salvo\Src\salvomem.c
Pumpkin\Salvo\Src\salvomsg.c
Pumpkin\Salvo\Src\salvomsg2.c
Pumpkin\Salvo\Src\salvomsgq.c
Pumpkin\Salvo\Src\salvomsgq2.c
Pumpkin\Salvo\Src\salvomsgq3.c
Pumpkin\Salvo\Src\salvomsgq4.c
Pumpkin\Salvo\Src\salvoprio.c
Pumpkin\Salvo\Src\salvoprio2.c
Pumpkin\Salvo\Src\salvoqdel.c
Pumpkin\Salvo\Src\salvoqins.c
Pumpkin\Salvo\Src\salvorpt.c
Pumpkin\Salvo\Src\salvosched.c
Pumpkin\Salvo\Src\salvosem.c
Pumpkin\Salvo\Src\salvosem2.c
Pumpkin\Salvo\Src\salvostop.c
Pumpkin\Salvo\Src\salvotask.c
Pumpkin\Salvo\Src\salvotask2.c
Pumpkin\Salvo\Src\salvotask3.c
Pumpkin\Salvo\Src\salvotask4.c
Pumpkin\Salvo\Src\salvotask5.c
Pumpkin\Salvo\Src\salvotask6.c
Pumpkin\Salvo\Src\salvotask7.c
Pumpkin\Salvo\Src\salvotask8.c
Pumpkin\Salvo\Src\salvotick.c
Pumpkin\Salvo\Src\salvotid.c
Pumpkin\Salvo\Src\salvotimer.c
Pumpkin\Salvo\Src\salvoutil.c
Pumpkin\Salvo\Src\salvover.c
Listing 37: Source Code Files

Compiler-specific header and source files are listed in each compiler's Salvo Compiler Reference Manual.

Note Salvo source code uses tab settings of 2, i.e. tabs are
equivalent to 2 spaces.

Locations of Salvo Functions
Below is a list of each Salvo function (including user services and
certain internal functions called by user services, shown in italics)
and the source file in which it resides. This list is provided to assist
source code users in resolving compile-time link errors due to the
failure to include a particular Salvo source code file in their project.

Salvo User Manual

Chapter 7 • Reference

391

Note Under certain configurations, those functions marked with
an '*' may be macros or in-lined code instead of functions.
OSClrEFlag()*
OSCreateBinSem()*
OSCreateEFlag()*
OSCreateEvent()
OSCreateMsg()*
OSCreateMsgQ()*
OSCreateSem()*
OSCreateTask()
OSCtxSw()*
OSDelay()
OSDelDelayQ()
OSDelPrioQ()
OSDelTaskQ()
OSDestroy()
OSDestroyTask()
OSDispTcbP()
OSeID()
OSGetPrio()*
OSGetPrioTask()
OSGetTicks()
OSGetState()
OSGetStateTask()
OSGetTS()
OSInit()
OSInitEcb()
OSInitPrioTask()
OSInitTcb()
OSInsDelayQ()
OSInsElig()*
OSInsPrioQ()
OSInsTaskQ()
OSLogErr()*
OSLogMsg()*
OSLogWarn()*
OSMakeStr()
OSMsgQEmpty()
OSPrintEcb()
OSPrintEcbP()
OSPrintTcb()
OSPrintTcbP()
OSReturnBinSem()
OSReturnEFlag()
OSReturnMsg()
OSReturnMsgQ()
OSReturnSem()
OSRpt()
OSSaveRtnAddr()
OSSched()*
OSSchedEntryHook()
OSSchedDispatchHook()
OSSchedReturnHook()
OSSetEFlag()*
OSSetPrio()

392

Chapter 7 • Reference

salvoeflag.c
salvobinsem.c
salvoeflag.c
salvoevent.c
salvomsg.c
salvomsgq.c
salvosem.c
salvoinit2.c
salvoportxyz.h
salvodelay.c
salvoqdel.c
salvoqdel.c
salvotask7.c
salvodestroy.c
salvotask3.c
salvorpt.c
salvoeid.c
salvoprio2.c
salvoprio2.c
salvoticks.c
salvotask.c
salvotask5.c
salvodelay2.c
salvoinit.c
salvoinit4.c
salvoinit2.c
salvoinit3.c
salvoqins.c
salvoqins.c
salvoqins.c
salvotask8.c
salvodebug.c
salvodebug.c
salvodebug.c
salvodebug.c
salvomsgq3.c
salvorpt.c
salvorpt.c
salvorpt.c
salvorpt.c
salvobinsem2.c
salvoeflag2.c
salvomsg2.c
salvomsgq2.c
salvosem2.c
salvorpt.c
salvoutil.c
salvosched.c
salvosched.c
salvosched.c
salvosched.c
salvoeflag.c
salvoprio.c

Salvo User Manual

OSSetPrioTask()
OSSetTicks()
OSSetTS()
OSSignalBinSem()*
OSSignalEvent()
OSSignalMsg()*
OSSignalMsgQ()*
OSSignalSem()*
OSStartTask()
OSStop()
OSStopTask()
OSSyncTS()
OSTaskUsed()
OSTaskRunning()
OStID()
OSTimer()*
OSWaitEvent()

salvotask6.c
salvoticks.c
salvodelay2.c
salvobinsem.c
salvoevent.c
salvomsg.c
salvomsgq.c
salvosem.c
salvotask.c
salvostop.c
salvotask2.c
salvodelay3.c
salvotask7.c
salvotask4.c
salvotid.c
salvotimer.c
salvoevent.c

Listing 38: Location of Functions in Source Code

Abbreviations Used by Salvo
The following abbreviations are used throughout the Salvo source
code:
address
array
binary
change
check
circular
clear
create
configuration
context
current
cyclic timer
delay
delete
depth
destroy
disable
disable interrupt(s)
ecb pointer
eligible
enable
enable interrupt(s)
enter
event
event control block
event flag
event flag control block
event type
error

Salvo User Manual

Chapter 7 • Reference

addr
A
bin
change, chg
chk
circ
clr
create
config
ctx
curr, c
cycTmr
delay
del
depth
destroy
dis
di
ecbP
elig
en
ei
enter
event, e
ecb
eFlag
efcb
eType
err

393

from
global
global type
identifier
include guard
initialize
insert
length
local
location
maximum
message
message queue
message queue control block
minimum
not available
number
operating system
pointer
pointer to a pointer
prescalar
previous
priority
queue
report
reset
restore
return
save
scheduler
semaphore
set
signal
stack
status
statistics
string
switch
synchronize
task
task control block
task function pointer
tcb extension
tcb pointer
tick
timeout
timer
timestamp
toggle
utility
value
version
wait(ing) (for)
warning

fm
gl
gltype
ID
IG
init
ins
len
l
loc
max
msg
msgQ
mqcb
min
NA
num
OS
ptr, p
pp
PS
prev
prio
Q
rpt
rst
rstr
rtn
save
sched
sem
set
signal
stk
stat
stats
str
sw
sync
task, t
tcb
tFP
tcbExt
tcbP
tick
timeout
timer
TS
tgl
util
val
ver
wait, w
warn

Listing 39: List of Abbreviations

394

Chapter 7 • Reference

Salvo User Manual

Chapter 8 • Libraries
Note This chapter provides an overview of using and
(re-)building Salvo libraries. Only general issues that affect all of
Salvo's libraries are covered here.
For library particulars, please refer to your compiler's Salvo Compiler Reference Manual.

Library Types
Salvo ships with two types of precompiled libraries – standard libraries and freeware libraries. The standard libraries contain all of
Salvo's basic functionality, configured for each supported compiler
and target processor. The standard libraries are included in their
respective Salvo standard distributions. The freeware libraries are
identical to the corresponding standard libraries except for the relatively limited numbers of supported tasks and events, and are included in the Salvo Lite distributions.
Salvo Pro users can create applications using the Salvo source
files, the standard libraries, or a combination thereof. All other
Salvo users must use libraries when creating their applications. For
functionality and flexibility greater than that provided by the libraries, you'll need to purchase Salvo for full access to the Salvo
source code, and all the configuration options.

Libraries for Different Environments
The various Salvo distributions contain libraries for two different
kinds of compilers – native and non-native compilers.

Native Compilers
By native compilers we mean compilers that generate output (usually in .hex format) for a specific embedded target. You would use
a native compiler to create a Salvo application for a real product.
Native compilers are usually cross-compilers, i.e. they run on one
machine architecture (usually x86-based PCs) and generate code
for another (e.g. TI MSP430).

Salvo User Manual

395

Non-native Compilers
By non-native compilers we mean compilers that generate code for
another target altogether (usually an x86-based PC). Salvo's support for these "pure" compilers93 is intended to facilitate crossplatform development of Salvo applications for embedded targets.
Users can build C console applications and test, run, and debug
them on their main development machine (e.g. a PC) before building the same application for the intended embedded target (e.g. a
PICmicro MCU). The editing and debugging features available on
PCs are powerful tools that can aid in project management, testing
and debugging.
If you wish to develop your embedded application on the PC and
then recompile your Salvo application for your embedded target,
keep in mind that the non-native compilers generally lack any support for non-console-oriented subsystems that may exist on your
embedded target. Therefore you will need to simulate things like
serial I/O, A/D, D/A, interrupts, etc.
This "build on two, run on one" technique can be quite useful. For
example, you could write, test and debug a Salvo application that
passes floating-point data between two tasks via a message queue.
The PC's enormous94 resources (stdout buffers, memory, etc.),
coupled with a good IDE, present an ideal environment for developing this sort of application. You could debug your application
using printf() or the IDE's debugger. Once your application
works on the PC – and as long as you've used C library functions
that are also included in your target compiler's libraries – then
building a Salvo application for the embedded target should be a
snap!

Using the Libraries
In order to use a Salvo library, place the OSUSE_LIBRARY and
OSLIBRARY_XYZ configuration options particular to your compiler
into your salvocfg.h. These configuration options ensure that the
same configuration options used to generate the chosen library will
also be used in your source code.
For example, to use the full-featured standard library for HI-TECH
PICC and the PIC16F877A, your salvocfg.h file would contain
only:
93
94

396

As opposed to cross-compilers.
When compared to an embedded microcontroller.

Chapter 8 • Libraries

Salvo User Manual

#define
#define
#define
#define

OSUSE_LIBRARY
OSLIBRARY_TYPE
OSLIBRARY_CONFIG
OSLIBRARY_VARIANT

TRUE
OSL
OSA
OSB

Listing 40: Example salvocfg.h for Use with Standard
Library

and your project would link to the standard library slp42Cab.lib.
Please see Chapter 5 • Configuration for more information on these
configuration options. Figure 25: Salvo Library Build Overview illustrates the process of building a Salvo application from a Salvo
library.
OSCOMPILER and OSTARGET are not
vocfg.h file listed above. That's because in

included in the salmost cases Salvo can
automatically detect the compiler in use and then set the target
processor accordingly. This is done in the preprocessor via predefined symbols supplied by the compiler.

Note

Overriding Default RAM Settings
Each library is compiled with default values for the number of objects (tasks, events, etc.). By setting configuration parameters in
salvocfg.h it's possible to increase or decrease the RAM allocated to Salvo, and hence the number of objects in your application.
If the number of objects in your application is smaller than what
the library is compiled for, or your application doesn't use certain
objects (e.g. message queues) that have their own, dedicated control blocks, you can reduce Salvo's RAM usage. Just add the appropriate configuration options to salvocfg.h and rebuild your
project.
For example, to set the amount of RAM allocated to tasks in the
above example to just two, your salvocfg.h file would contain:
#define
#define
#define
#define
#define

OSUSE_LIBRARY
OSLIBRARY_TYPE
OSLIBRARY_CONFIG
OSLIBRARY_VARIANT
OSTASKS

TRUE
OSL
OSA
OSB
2

Listing 41: Example salvocfg.h for Use with Standard
Library and Reduced Number of Tasks

Salvo User Manual

Chapter 8 • Libraries

397

and you would link these three files:
main.obj, salvomem.obj, slp42Cab.lib

to build your application. By adding the following two lines to
your salvocfg.h:
#define OSEVENT_FLAGS
#define OSMESSAGE_QUEUES

0
0

Listing 42: Additional Lines in salvocfg.h for Reducing
Memory Usage with Salvo Libraries

you can prevent any RAM from being allocated to event flag and
message queue control blocks, respectively.

Caution This technique frees RAM for other uses in your application, and must be used with caution. If you reduce OSTASKS or
OSEVENTS from their default values, you must ensure that you do
not perform any Salvo services on tasks or events that are now "out
of range." E.g. for libraries that support three tasks, if you reduce
OSTASKS to 2 as outlined above, you must not call OSCreateTask(TaskName, OSTCBP(3), prio). If any of your own
variables are located in RAM immediately after the tcbs, they will
be overwritten with the call to OSCreateTask().
Setting the number of objects in an application above the library
defaults is only possible with the standard libraries – the preset
limits in the freeware libraries cannot be overridden.

Note Illegal or incorrect values for the number of objects in an
application that uses a library will usually be flagged by the compiler as an error.

Library Functionality
By linking your application to the appropriate library, you can use
as few or as many of Salvo's user services as you like. Each library
supports up to some number of tasks and events.

Note Because of the enormous number of possible configurations, the standard and freeware libraries support most, but not all,
of Salvo's functionality. Each library is compiled with a particular
set of configuration options. See the library-specific details (below) or Pumpkin\Salvo\Inc\salvolib.h for more information.

398

Chapter 8 • Libraries

Salvo User Manual

Warning Do not edit Pumpkin\Salvo\Inc\salvolib.h. Doing
so may cause problems when compiling and/or linking your
application to the freeware libraries.

Types
The library type is specified using the OSLIBRARY_TYPE configuration option in salvocfg.h.
The library types, shown in Table 11, are self-explanatory.
type code
f / OSF:
l / OSL:

description
Freeware library. Number of tasks, events,
etc. is restricted.95
Standard library. Number of tasks, events,
etc. is limited only by available RAM.

Table 11: Type Codes for Salvo Libraries

Note The standard libraries are slightly smaller than the corresponding freeware libraries.

Memory Models
Where applicable, Salvo libraries are compiled for different memory models. There is no configuration option for specifying the
memory model.

Options
Where applicable, Salvo libraries are compiled with different options. There is generally no configuration option for specifying the
option.

Global Variables
Salvo uses a variety of objects for internal housekeeping. Where
applicable, the OSLIBRARY_GLOBALS configuration option in salvocfg.h is used to specify the storage type for these global variables. The configuration codes vary by compiler.

95

Salvo User Manual

Most freeware libraries are compiled with OSSET_LIMITS set to TRUE.

Chapter 8 • Libraries

399

Configurations
The

library

configuration
is
specified
using
configuration option in salvocfg.h.

the

OSLIBRARY_CONFIG

The library configurations, shown in Table 12, indicate which services are included in the library specified. Use the library that includes the minimum functionality that your application requires.
For example, don't use an a-series library unless your application
requires both delay (e.g. OS_Delay()) and event (e.g. OSSignalSem()) services.
configuration code
a / OSA:

d / OSD:

e / OSE:

m / OSM:
s / OSS:
t / OST:
y / OSY:

description
Library supports multitasking with delay
and event services – all default functionality is included.
Library supports multitasking with delay
services only – event services are not supported.
Library supports multitasking with event
services only – delay services are not supported.
Library supports multitasking only – delay
and event services are not supported.
Library supports only Salvo SE features.
Library supports multitasking with delay
and event services. Tasks can wait on
events with a timeout.
Library supports only Salvo tiny features.

Table 12: Configuration Codes for Salvo Libraries

Note Using a library that's been created with support for services
you don't use will have an impact on your application's ROM and
RAM requirements.
Table 13 shows the essential differences among the library configurations.

400

Chapter 8 • Libraries

Salvo User Manual

configuration
Delay services:
Event
services:
Idling function:
Task
priorities:
Timeouts:

a

d

e

m

s

t

y

+

+

-

-

+

+

+

+

-

+

-

+96

+

+97

+

+

+

-

+

+

+

+

+

+

-

+

+

-

-

-

-

-

-

+

-

Table 13: Features Common to all Salvo Library
Configurations

+: enabled
-: disabled

Variants
The library variant is specified using the OSLIBRARY_VARIANT
configuration option in salvocfg.h.
A variety of different compilers are certified for use with Salvo.
Some compilers use the target processor's stack or registers to pass
parameters and store auto variables – this is true for all compilers
for x86 targets. There are no library variants for these conventional compilers.
Other compilers certified for use with Salvo maintain parameters
and auto variables as static objects in dedicated RAM – this is the
case for targets that do not have or use general-purpose stacks for
parameter and auto variable storage. The libraries for these compilers have variants. The remainder of this section applies to the
libraries for these compilers.
Some of Salvo's services can be called from within interrupts.
Those services include:

96
97

Salvo User Manual

Binary semaphores, semaphores and messages.
Binary semaphores and semaphores.

Chapter 8 • Libraries

401

•
•
•
•
•
•

•
•
•
•
•
•

•

• OSGetPrioTask()
• OSGetStateTask()
• OSReadBinSem()
• OSReadEFlag()
• OSReadMsg()
• OSReadMsgQ()
• OSReadSem()
• OSMsgQEmpty()
• OSSignalBinSem()
• OSSignalMsg()
• OSSignalMsgQ()
• OSSignalSem()
• OSStartTask()

Listing 43: Partial Listing of Services than can be called
from Interrupts

If the target processor does not have a general-purpose stack, the
Salvo source code must be properly configured via the appropriate
configuration parameters. The library variants, shown in Table 14,
are provided for those applications that call these services from
within interrupts.
If your application does not call any of the services above from
within interrupts, use the b variant. If you wish to these services
exclusively from within interrupts, use the f variant. If you wish to
do this from both inside and outside of interrupts, use the a variant.
In each case, you must call the services that you use from the correct place in your application, or either the linker will generate an
error or your application will fail during runtime.

402

Chapter 8 • Libraries

Salvo User Manual

variant code
a / OSA:

b / OSB:

e / OSE:

f / OSF:
- / OSNONE:

description
Applicable services can be called from
anywhere, i.e. from the foreground and
the background, simultaneously.
Applicable services may only be called
from the background (default).
Applicable services may only be called
from either the foreground or the background, but not both.
Applicable services may only be called
from the foreground.
Library has no variants.98

Table 14: Variant Codes for Salvo Libraries

See the OSCALL_OSXYZ configuration parameters for more information on calling Salvo services from interrupts.

Library Reference
Refer to your compiler's Salvo Compiler Reference Manual for
details on the associated Salvo libraries.

Rebuilding the Libraries
One common reason to rebuild the Salvo libraries occurs when the
compiler you are using has been upgraded (new versions, enhancements, bug fixes, etc.) and pre-compiled Salvo libraries built
with the new compiler have not yet been released. In a situation
like this, you must rebuild the Salvo libraries in order to build your
library-build Salvo projects.
Doing source-code builds is generally an easier way to set configuration options for a Salvo project. In multi-user environments,
however, it may be wiser to force all Salvo users working on a single application to link to a single, custom library so as to ensure
that they are all configured identically.

Note Libraries can only be rebuilt by Salvo Pro users, as the
Salvo source code is required.

98

Salvo User Manual

A library may have no variants if the target processor does not support
interrupts, or if the target processor has a conventional stack and the ability to
save and restore the state of interrupts.

Chapter 8 • Libraries

403

GNU Make and the bash Shell
The Salvo libraries are generated with GNU make in the bash
shell.99 If you have Salvo Pro you can rebuild the libraries using
the makefiles in the Pumpkin\Salvo\Src directory.

Note The Salvo library makefiles are designed to run from the
Pumpkin\Salvo\Src

directory.

In addition to the make utility, other utilities commonly used in the
bash shell are also required for a successful make, including
expr(.exe). Refer to your bash shell documentation for information on installing the various utilities.
Salvo's makefile system is relatively complex and uses make recursively. Normally, users need not edit the makefiles. However, if
you have installed your compiler(s) in places that differ from those
specified in the Salvo makefiles, you may need to edit the appropriate makefile for a successful compile.

Rebuilding Salvo Libraries

Linux/Unix Environment
To rebuild a particular library in the bash shell, simply specify it
as make's target, e.g.
•
•

$: cd /Pumpkin/Salvo/Src
$: make –f Makefile libsalvolmcc30it.a
Listing 44: Making a Single Salvo Library

The Salvo makefiles also allow for groups of libraries to be made,
e.g.
•
•

$: cd /Pumpkin/Salvo/Src
$: make –f Makefile ra430

Listing 45: Making all Salvo Libraries for a Particular
Compiler

to generate all of the Salvo libraries for the Rowley Associates
CorssWorks for MSP430 toolset (Salvo code RA430), and
99

404

Bourne-again shell, a Unix command language interpreter.

Chapter 8 • Libraries

Salvo User Manual

•
•

$: cd /Pumpkin/Salvo/Src
$: make –f Makefile msp430

Listing 46: Making all Salvo Libraries for a Particular
Target

to generate all of the Salvo libraries for MSP430 targets. Naturally,
you will need all of the compiler(s) associated with the Salvo libraries you're rebuilding.
A list of target groups can be obtained by issuing the commands:
•
•

$: cd /Pumpkin/Salvo/Src
$: make –f Makefile

Listing 47: Obtaining a List of Library Targets in the
Makefile

Multiple Compiler Versions
Some of Salvo's supported compilers are in use at different version
levels. For these compilers, the make command-line argument
CVER must also be specified, e.g.
•
•

$: cd /Pumpkin/Salvo/Src
$: make –f Makefile iar430 CVER=2

Listing 48: Making Salvo Libraries for IAR's MSP430 C
Compiler v2.x

will result in Salvo libraries being built and placed in
\Pumpkin\Salvo\Lib\IAR430-v2. CVER details are compilerdependent – see the Salvo makefiles for more information.
can be combined with CLC when building custom libraries (see below).

Note

CVER

Win32 Environment
To rebuild Salvo libraries in a Win32 environment, you will need a
bash shell along with GNU make. One free source for both is the
Cygwin bash shell. Another is the MinGW project, along with associated utilities.100
100

Salvo User Manual

A MinGW installation is reported to require only MinGW (e.g. Mingw2.0.0-3.exe) and Msys (e.g. Msys-1.0.8.exe), available on
http://www.SourceForge.net. MinGW should be installed before Msys.

Chapter 8 • Libraries

405

Currently, all libraries included in Salvo distributions are built in
the Cygwin bash shell using make recursively, as outlined
above.101 Therefore you are strongly encouraged to set up a working Cygwin bash shell from the latest Cygwin releases for generating Salvo libraries.

Customizing the Libraries
You can rebuild the Salvo libraries to a configuration that differs
from the standard build.102 This is useful in situations where you
prefer to do library builds, and the standard libraries differ somewhat from the configuration that you require.
Using custom libraries is a three-step process, involving:
•
•
•

• creating a custom library configuration file,
• building the custom library and
• using the custom library in a library build

Creating a Custom Library Configuration File
Salvo provides for 20 different user-definable custom library configuration files, salvoclc1.h through salvoclc20.h.103 When a
custom library is in use, one of these files will be included in the
salvo configuration file Pumpkin\Salvo\Inc\salvolib.h via the
C preprocessor's #include "filename" directive.

Note Because of the use of "" in the #include directive, the custom library configuration file must be located in the preprocessor's
user search path. It is up to the user to ensure that the preprocessor
can find the selected custom library configuration file. A safe location for such files is the Pumpkin\Salvo\Inc directory, or the project directory.
Each custom library configuration file includes overrides of Salvo
configuration option settings used to generate the library. For each
configuration option to be overridden, the Salvo symbol should
101
102
103

406

PCs with large (e.g. 1GB) amounts of RAM are used to avoid the recursive
make problems that have plagued Cygwin.
Note that Pumpkin cannot provide support for libraries that differ from those
provided in the Salvo distributions.
Salvo installers do not install any salvoclcN.h files. The installers will not
replace, overwrite or delete any such user files.

Chapter 8 • Libraries

Salvo User Manual

first be #undef'd, then #define'd, so as to avoid any preprocessor
warnings.

Building the Custom Library
Once your custom library configuration file is ready, you rebuild
the Salvo library or libraries using the Salvo makefiles and an
additional make command-line option, CLC=N, where N is the
number of the custom library configuration file you are using.

Note Most users of custom Salvo libraries will only need to override a few of the configuration options for the standard libraries.
The library or libraries you choose to rebuild should have a default
configuration that is as close as possible to what you are trying to
achieve with your custom library.

Using the Custom Library in a Library Build
After you have built your custom library, you must set the
OSCUSTOM_LIBRARY_CONFIG configuration option in your project's
salvocfg.h configuration file to the number of your custom library configuration file. And of course you must link to the custom
library instead of a standard library.

Example – Custom Library with 16-bit Delays and Non-Zero
Prescalar
To build a Salvo library for the Archelon / Quadravox AQ430 Development Tools that has all of the features of an "ia" library, but
also has 16-bit delays and a timer prescalar of 5, one would start
with slaq430ia.lib. Assuming this will be custom library configuration 4, create a Pumpkin\Salvo\Inc\salvoclc4.h with the
following entries:
#undef OSBYTES_OF_DELAYS
#define OSBYTES_OF_DELAYS 2
#undef OSTIMER_PRESCALAR
#define OSTIMER_PRESCALAR 5
Listing 49: Example Custom Library Configuration File
salvoclc4.h

and then build the new library:

Salvo User Manual

Chapter 8 • Libraries

407

•
•

$: cd /Pumpkin/Salvo/Src
$: make –f Makefile libsalvolmcc32l-t.a
CLC=4

Listing 50: Making a Custom Salvo Library with Custom
Library Configuration 4

Note The CLC= command-line argument to make is case-sensitive.
Making the custom library as above will result in a new library,
\Pumpkin\Salvo\Lib\MCC32\libsalvolmcc32l-t-clc4.a.
To use the new library, add OSCUSTOM_LIBRARY_CONFIG to your
project's salvocfg.h, e.g.:
#define
#define
#define
#define

OSUSE_LIBRARY
OSLIBRARY_TYPE
OSLIBRARY_CONFIG
OSCUSTOM_LIBRARY_CONFIG

TRUE
OSL
OSA
4

Listing 51: Example salvocfg.h for Library Build Using
Custom Library Configuration 4 and Archelon /
Quadravox AQ430 Development Tools

and

link

your

project

to

your

new

custom

library
\Pumpkin\Salvo\Lib\MCC32\libsalvolmcc32l-t-clc4.a.

Note In this example, we've only altered the standard library
slightly. In general, you should pick a standard library that is as
close as possible to the configuration you want in your custom library. Deviating substantially from the standard library's configuration may cause problems when building the library because of
conflicts between configuration options. Also, it may result in an
unnecessarily large library. Advanced users may want to review
Pumpkin\Salvo\Inc\salvolib.h to solve such problems using
the defined symbols contained therein.
To build a custom library for a particular library and a particular
version of the associated compiler, combine the CLC and CVER
arguments to the makefile:
•
•

$: cd /Pumpkin/Salvo/Src
$: make –f Makefile libsalvolra430-t.hza
CLC=2 CVER=1

Listing 52: Making a Custom Salvo Library with Custom
Library Configuration 4

408

Chapter 8 • Libraries

Salvo User Manual

Making the custom library as above will result in a new library,
\Pumpkin\Salvo\Lib\RA430-v1\libsalvolra430-t-clc2.hza.

Note To avoid problems associated with different compilers
and/or targets, each custom library configuration file salvoclcN.h
should only be used with a single compiler and target combination.

Preserving a User's salvoclcN.h Files
The Salvo installers will not touch or delete any existing salvoclcN.h files. Therefore custom library configuration files can be
left in place when Salvo is upgraded.

Restoring the Standard Libraries
The standard Salvo libraries can be restored by either re-installing
them from the Salvo installer, or by rebuilding the libraries without
any CLC= command-line options to make. Since the Salvo library
makefile system automatically assigns unique, descriptive names
to custom libraries, there is no good reason to alter or move the
standard libraries.

Custom Libraries for non-Salvo Pro Users
Occasionally, potential Salvo users will request a custom library
for evaluation. This will invariably be a custom Salvo Lite (freeware) library. Using a custom Salvo freeware library is no different
from using a custom Salvo standard library – just follow the steps
outlined above.

Makefile Descriptions

Pumpkin\Salvo\Src\Makefile
This makefile uses a regular expression to parse the name of the
desired library or libraries. It then calls make recursively using
Makefile2 to generate one or more libraries.

Salvo User Manual

Chapter 8 • Libraries

409

Pumpkin\Salvo\Src\Makefile2
This makefile references the compiler- and target-specific Makefile in the CODE subdirectory.

Pumpkin\Salvo\Src\CODE\Makefile
This makefile file contains drives the compiler(s) and assembler(s)
required to generate the libraries. Compiler-specific paths are located in this file.

Pumpkin\Salvo\Src\CODE\targets.mk
This include file contains the names of all valid Salvo libraries for
the selected compiler and target.

410

Chapter 8 • Libraries

Salvo User Manual

Chapter 9 • Performance

Introduction
In this chapter we'll address the runtime aspects of Salvo which
affect performance. A good understanding is essential if you wish
to extract the maximum possible performance from your target
processor.

Interrupts
Salvo controls interrupts in two distinct regions of its code – in the
context switcher, and in critical sections. These two regions of the
Salvo code are target- and sometimes compiler-specific, unlike the
main body of Salvo code, which is target-independent. These code
regions and their impact on your application are discussed below.

Context Switcher
The Salvo context switcher for each compiler and target family is
unique. In general terms, the context switcher handles:
•
•
•
•
•

Vectoring from the scheduler to the task
Generating a local stack frame for the task
Storing the task's updated resume address in the
task's task control block (tcb)
Any required register save and restores
Returning from the task to the scheduler

Note All Salvo tasks execute with interrupts enabled. Therefore
interrupts are enabled when entering and exiting the Salvo context
switcher.
For most Salvo context switchers, the operations listed above involve changes to the stack and stack pointer (SP). Wherever possible, interrupts are not disabled during the operation of the context

Salvo User Manual

411

switcher. This is possible104 in most Salvo context switchers, and
depends on the target architecture.

Note Most Salvo context switchers are implemented in assembly
language and are unaffected by any project optimizations.

Tip Each Salvo Compiler Reference Manual clearly states the
interrupt-disabling behavior of the particular context switcher.
In the rare cases where it is not possible to context switch without
disabling interrupts, every effort has been made to minimize the
number of cycles during which interrupts are disabled. Therefore,
for Salvo distributions whose context switcher have non-zero interrupt latencies, the latency represents the maximum interrupt latency due to the Salvo context switcher. Even in these cases, the
latency is usually less than 20 instruction cycles.

Note The latency of the Salvo context switcher is constant and is
independent of all other aspects of a Salvo application.

Summary
Most Salvo context switcher do not disable interrupts and therefore
introduce no interrupt latency into a Salvo application.
Those Salvo context switchers that do disable interrupts do so for
the minimum time possible.

Critical Sections
Critical sections of code are sections of code that must not be preempted. In a single-threaded application, preemption occurs
through interrupts. If a critical section of code is preempted, then
there is a real possibility of corruption of global variables. Since
the vast majority of microcontrollers do not have protected memory features, it is imperative that Salvo take steps to prevent preemption during critical sections.

Note Most callable Salvo services include critical sections.

104

412

If the Stack Pointer on the target architecture can be changed atomically, then
this usually means that interrupts need not be disabled during a Salvo context
switch.

Chapter 9 • Performance

Salvo User Manual

Salvo has two user-definable hooks (i.e. functions) that are used to
prevent preemption (and therefore corruption of Salvo's own
global variables).105 They are OSDisableHook() and OSEnableHook(). OSDisableHook() is called inside a Salvo service at the
beginning of a critical section, and OSEnableHook() is called inside a Salvo service at the end of a critical section.

Note Interrupt hooks are contained in every Salvo library. Refer
to the appropriate Salvo Compiler Reference Manual for the functionality of the hooks. All Salvo hooks can be overridden by the
user, in both source-code builds and library builds.
Inside the Salvo source code, the interrupt hooks are used like this:
… // Non-critical section of Salvo code
OSDisableHook();
… // Critical section of Salvo code
OSEnableHook();
… // Non-critical section of Salvo code
OSDisableHook();
… // Critical section of Salvo code
OSEnableHook();
… // etc.
Listing 53: Use of interrupt hooks in Salvo source code.

Note Non-dummy (i.e. non-empty) interrupt hooks are targetand sometimes even compiler-specific.106

Warning Salvo users cannot change how or when these hooks
are called. Their positions in the Salvo code have been chosen to
disable interrupts only while required for critical sections. Salvo's
critical sections have been coded to be as short as possible.

Effect on Runtime Performance
The runtime length of a Salvo service – and hence the runtime
length of a critical section107 in Salvo's code – can only be obtained

105

106

Salvo User Manual

An example of one of Salvo's global variables is the pointer to the head of the
queue of delayed tasks. If a mainline Salvo service is in the process of making
changes to the head of this queue and an interrupt occurs which calls a Salvo
service that changes the head of this queue, the result will be unpredictable
and will lead to a malfunction of the application. Therefore all interrupt-level
calls to Salvo services must be suppressed while any Salvo service is making
any changes to a Salvo global variable.
A compiler-specific hook might include the weak keyword when the compiler
supports this feature.

Chapter 9 • Performance

413

through measurement in an actual application.108 Some Salvo services have very short critical sections. Some even have no critical
sections. Yet others can potentially have very long critical sections
(e.g. when a low-priority task must be enqueued into the eligible
queue where several higher-priority tasks are already eligible). To
the Salvo user, the main area of concern here is "How long does
Salvo disable my interrupts?", as this can adversely affect onboard peripherals that are used in an interrupt-driven manner.109 As
you will see below, Salvo can be configured for zero interrupt latency for any desired interrupt source.
We will now examine various scenarios for the coding of the interrupt-disabling hooks

Controlling Interrupts Globally
The most general and safest configuration for the user interrupt
hooks is for the hooks to disable interrupts globally during a critical section. This is the default for the hooks contained in all Salvo
library builds where the target architecture has a single, consistent
method of disabling and enabling global interrupts.
void OSDisableHook ( void )
{
__disable_interrupt();
}
void OSEnableHook ( void )
{
__enable_interrupt();
}
Listing 54: Most general configuration for Salvo's
interrupt hooks.

The advantage of this approach is that it is safe for all application.
With the hooks defined as shown in Listing 54, any Salvo service
can be called from any interrupt without fear of corrupting Salvo's
global variables. That's why this is the default for all Salvo libraries.

107
108

109

414

For a given Salvo service, the runtime length of the critical section contained
therein cannot exceed the runtime length of the service itself.
This is due in no small part to the wide range of Salvo configuration options
and their effect on the runtime performance of the Salvo code. Its is also due
to the priority-queue-based priority-resolution algorithms used in Salvo.
For example, an interrupt-driven single-byte-buffer asynchronous serial
receiver operating at 115200,N,8,1 cannot tolerate its interrupts being disabled
for longer than 87µs or it risks losing incoming characters.

Chapter 9 • Performance

Salvo User Manual

The disadvantage of this approach is that all interrupt sources are
disabled while Salvo is in a critical section, even if said interrupts
do not call Salvo services. Clearly, this non-targeted approach to
controlling interrupts is not well-suited to high-performance, interrupt-driven Salvo applications, due to the substantially non-zero
interrupt latencies imposed on the application.

Controlling Interrupts Individually
For better performance from interrupt-driven peripherals, individual control of interrupts during Salvo's critical sections is recommended. With this approach, only those interrupt sources which
themselves call Salvo services need to be disabled during critical
sections. Since this approach is target-specific, it is best illustrated
by example.
void OSDisableHook ( void )
{
IE2
&= ~URXIE1;
TBCCTL6 &=

~CCIE;

}

void OSEnableHook ( void )
{
IE2
|= URXIE1;
TBCCTL6 |=

CCIE;

}

Salvo User Manual

Chapter 9 • Performance

415

#pragma vector=USART1RX_VECTOR
__interrupt void ISRRx1 (void)
{
USART_UART1_inchar();
OSSignalSem(SEM_CMD_CHAR_P);
__low_power_mode_off_on_exit();
}
#pragma vector=TIMERB1_VECTOR
__interrupt void ISRTimerB1 (void)
{
switch(__even_in_range(TBIV,14))
{
case 0x0C:
TBCCR6 += TIMER_TICKS_RELOAD;
OSTimer();
__low_power_mode_off_on_exit();
break;
default:
fatal(FATAL_ERROR_UNUSED_ISR);
break;
}
}
Listing 55: Application-specific configuration for Salvo's
interrupt hooks. Relevant ISRs also shown. Target is
TI's MSP430FG4619.

In the Salvo application associated with the interrupt hooks of
Listing 55, two ISRs call Salvo services: ISRRx1() calls OSSignalSem() when a valid incoming character has been received via
USART1 and put into a buffer, and ISRTimerB1()110 calls
OSTimer() at a period rate. Since these are the only interrupts that
calls Salvo services, these are the only interrupt sources that must
be disabled during Salvo's critical sections. Therefore we see that
OSDisableHook() disables USART1 Rx interrupt generation and
TimerB6 interrupt generation, and OSEnableHook() re-enables the
same.

Note In this example it's assumed that interrupts are globally enabled at all times, and are not controlled by Salvo.
The net effect of the hooks in this example is that other interrupt
sources operate with zero interrupt latency because Salvo does not
disable global interrupts or the individual interrupt sources, as
there is no need to. Thus, performance is maximized with these
other interrupt-driven peripherals.
110

416

On the MSP430FG4619, the TimerB1 ISR handles interrupts for Timers B1
through B6, based on the Timer B Interrupt Vector (TBIV).

Chapter 9 • Performance

Salvo User Manual

Warning Failure in the interrupt hooks to disable an interrupt
source that calls a Salvo service will inevitably lead to runtime
problems in a Salvo application due to the unavoidable corruption
of global variables. Therefore it's important to keep track of which
Salvo services are called from ISRs, and configure the interrupt
hooks accordingly.

Tip There is no limit to how many different interrupt sources can
be controlled by the interrupt hooks. Just write OSDisableHook()
and OSEnableHook() accordingly.

Avoiding Interrupt Control Altogether
Strange as it may seem, there are Salvo applications that do not
require any control of interrupts. They include:
•
•
•

Salvo applications built on microcontrollers that
do not have interrupts (e.g. Microchip
PIC12F509).
Salvo applications that do not use services that
are traditionally called from an ISR (like Salvo's
timer).
Salvo applications that cannot tolerate any
interrupt latency yet, wish to call one or more
Salvo services from an ISR.

In the first two cases above the interrupt hooks need only be redefined as shown in Listing 56.
void OSDisableHook ( void )
{
;
}
void OSEnableHook ( void )
{
;
}
Listing 56: Interrupt hooks for applications that do not
call Salvo services from any interrupts.

Here, Salvo's critical sections do not involve any change to the interrupt status of the target microcontroller. If the target's Salvo
context switcher (see Context Switcher, above) has zero interrupt

Salvo User Manual

Chapter 9 • Performance

417

latency as well, then Salvo's total contribution to overall interrupt
latency is zero for all interrupt sources.
In the case where a user wishes to call a Salvo service from an interrupt, yet cannot tolerate any interrupt latency on that interrupt
source due to Salvo, then a slightly indirect approach is required.

Tip This situation can arise for example in targets that do not have
vectored interrupts, or in targets where a single interrupt vector
services several interrupt sources.
In this situation, Salvo's interrupt hooks do not disable the source
of interrupt that would normally call the Salvo service. Instead, the
user must create a semaphore that is used to pass information up
from the ISR to the main loop of the Salvo application:
int main ( void )
{
…
while (1) {
if (HighPrioISRDataReady == 1) {
GIEH = 0;
HighPrioISRDataReady = 0;
GIEH = 1;
OSSignalBinSem(HIGH_PRIO_ISR_DATA_READY_P);
}
OSSched();
}
}
Listing 57: Passing interrupt activity up from an ISR to
call a Salvo service without a corresponding interrupt
hook. Target is Microchip PIC18F452.

In Listing 57, a Salvo application built for the Microchip
PIC18F452 passes information up from a high-priority ISR111 to
ultimately cause a Salvo binary semaphore to be signaled. It does
this simply by setting a semaphore (HighPrioISRDataReady in
this example) inside the high-priority ISR when event signaling is
required. In the application's main() loop, this semaphore is tested
prior to calling the scheduler and if set, is reset with high-priority
interrupts disabled,112 and finally OSSignalBinSem() is called.

111

112

418

The PIC18 architecture has just two interrupt vectors – the low-priority
interrupt vector and the high-priority interrupt vector. Each vector has its own
individual interrupt enable bit (GIEL and GIEH, respectively).
Note that if the semaphore can be set and reset atomically, the control of the
GIEH bit in this example is unnecessary. It is shown, however, to remind the
reader for the general requirement of protecting global variables.

Chapter 9 • Performance

Salvo User Manual

This approach has a very substantial advantage in that the application can run without Salvo's critical sections affecting interrupts.
Yet the runtime performance of signaling a Salvo event is virtually
indistinguishable from that of an application built with the interrupt disabled (and its attendant non-zero interrupt latency). This is
because Salvo's scheduler processes events all at once, and so it
makes little difference as to whether an event is signaled at an arbitrary time113 or immediately before the scheduler is called.
The disadvantages of this approach are:
•

•

•

Depending on target architecture, the interrupt
source may still need to be disabled, albeit for a
very short time (just two instruction cycles in
the example of Listing 57 above).
Event processing no longer occurs in the order
that the interrupt occurred, but rather in the
order that the event is signaled in the user code
when the semaphore is found to have been set.
This mainly affects multiple tasks waiting on a
single event.
This involves polling the semaphore prior to
every invocation of Salvo scheduler. This is
contrary to the purely event-driven (i.e. no
polling) operation of Salvo.

For most applications, these disadvantages are outweighed by the
advantage of near-zero interrupt latency while still effectively calling a Salvo service from an interrupt.

Note Depending on the target architecture, the (albeit short) disabling and re-enabling of interrupts to protect the semaphore (a
global variable) as shown in Listing 57 above can be avoided if the
semaphore is set (in the ISR) and reset (after the semaphore test in
main()) atomically. In this case, the total interrupt latency remains
0 cycles – highly desirable. Inspection of the assembly code generated by the compiler will prove whether the desired operations are
atomic.

Tip Multiple semaphores from multiple interrupt sources can be
combined in this approach. Ideally, each semaphore should be implemented as a single-bit-wide bitfield in C, inside of a structure
113

Salvo User Manual

When signaling an event from an ISR, the signaling can happen at any time
except during a critical section (because said interrupt is disabled during that
critical section).

Chapter 9 • Performance

419

consisting of a union of all the bits (e.g. in an int) and of the individual bits. Therefore all the bits can be tested once (is the int
non-zero?), and if non-zero, the individual bits can be tested and
cleared individually. This minimizes the number of instruction cycles spent polling for a change in the semaphores' status, thereby
improving runtime performance and minimizing the use of polling
(which is undesirable).

Side Effects of Interrupt Hooks
Salvo's interrupt hooks are called from all Salvo services that contain critical sections. This means that many Salvo services that can
be called from ISRs will call the interrupt hooks while in the ISR,
with attendant changes to the interrupt enable bit(s) of the target.
For the default hook for most targets (see Listing 56 above), this
means that interrupts will be enabled at the end of the Salvo service that is called in the ISR. Therefore the interrupts controlled by
the interrupt hooks will be enabled prior to the end of the ISR. This
could lead to nested interrupts where none are desired, etc.
While this is not usually a problem, it can be solved by explicitly
flagging being in an ISR and basing the interrupt hook actions on
the flag, as shown in Listing 58:

420

Chapter 9 • Performance

Salvo User Manual

static unsigned int InISR = 0;
void entering_isr ( void )
{
InISR = 1;
}
void leaving_isr ( void )
{
InISR = 0;
}
void OSDisableHook ( void )
{
if (InISR == 0) {
__disable_interrupt();
}
}
void OSEnableHook ( void )
{
if (InISR == 0) {
__enable_interrupt();
}
}
Listing 58: Interrupt hooks to avoid interrupt nesting.

With this method, any ISR that calls Salvo services begins with
entering_isr() and ends with leaving_isr(). This completely
avoids nested interrupts. This user flexibility – no need to change
any Salvo code here – is the reason for the introduction of hook
functions in Salvo 4.

Tip If the compiler or target provides an automatic means of detecting that code is executing at the ISR / foreground level, this can
be used to your advantage in the interrupt hooks.

The Fallacy of Avoiding Critical Sections at the Interrupt Level
Some inexperienced programmers might fall for the notion that
preempting a critical section can be avoided by testing for a condition inside an ISR's (i.e. in the foreground) code instead of by disabling the ISR in the critical section in mainline (i.e. in the
background) code. The idea is to forego all interrupt control in
Salvo's critical sections, in favor of setting a flag, which can then
be tested in the ISR to avoid call a Salvo service during the critical
section. While the test will in fact work, the rest will not, as there
is no way in the ISR to know when the critical section will complete. And since the critical section does not progress while in the

Salvo User Manual

Chapter 9 • Performance

421

ISR, there is in fact no way to know when the ISR can call the
Salvo service. In effect, the critical section is being blocked by the
ISR, which is the opposite of what is desired.
Therefore the prescribed methods above for configuring Salvo's
interrupt hooks for critical sections must be followed.

User Hooks
Salvo has four hook services that can be redefined by the user to
suit the chosen target and application. They are:
•
•
•

Interrupt hooks: OSDisableHook(),
OSEnableHook()
Watchdog hook: OSClrWDTHook()

Idling hook: OSIdlingHook()

Tip Since each user hook is defined in its own source code module, state information can be combined with a hook function by
declaring a local static variable in the module, and referencing the
variable from the hook function(s).

OSDisableHook(), OSEnableHook()
The use of the interrupt hooks is covered above in Interrupts.

OSClrWDTHook()
The watchdog hook provides a simple and integrated way to clear
an application's watchdog timer from within the Salvo portion of
your application. OSClrWDTHook() is called each time Salvo's
scheduler is called.
OSClrWDTHook() is not a failsafe means of properly
maintaining a software or hardware watchdog. It is provided as a
simple scheme that is useful and applicable to many applications,
especially in the early stages of their software development. If a
more sophisticated approach to watchdog management is required,
the user can either override the hook (by defining it as a dummy
function), expand the hook (by replacing the default hook with a
more sophisticated version), or augmenting the hook with other
watchdog-related application code.

Warning

422

Chapter 9 • Performance

Salvo User Manual

void OSClrWDTHook ( void )
{
WDTCTL = (WDTCTL & 0x00FF) | WDTPW | WDTCNTCL;
}
Listing 59: Example watchdog hook. Target is TI's
MSP430F1612.

In Listing 59 the watchdog hook clears the target's watchdog timer
without any other changes.

Salvo User Manual

Chapter 9 • Performance

423

424

Chapter 9 • Performance

Salvo User Manual

Chapter 10 • Porting
With its minimal RAM requirements, small code size and high performance, Salvo is an appealing RTOS for use on just about any
processor. Even if it hasn't been ported to your processor and/or
compiler, you can probably do the port in a day or two.
If you are interested in porting Salvo to a new target processor
and/or compiler, please contact Pumpkin for more details. A comprehensive Salvo Porting Manual is available.

Salvo User Manual

425

426

Chapter 10 • Porting

Salvo User Manual

Chapter 11 • Tips, Tricks and
Troubleshooting

Introduction
If you're having trouble getting your code to work properly with
Salvo, here are some suggestions on how to solve your problem.
•
•

•

•

•
•
•
•

• Read and re-read all the relevant portions of
this manual.
• Review the example programs in this manual
and in the Salvo distribution. You may find
something that is very similar to what you are
trying to do.
• Examine the postprocessed output of your
compiler, both in C and in assembly language.
Output listings contain a wealth of useful
information.
• Examine any map files generated by your
compiler. These files have information
containing the location of Salvo routines and
variables and their sizes, the calling trees, etc.
• Use the error codes returned by the user
services to verify that the desired Salvo actions
are really happening.
• If your application has the RAM and ROM to
support it, use OSRpt() to examine the status of
the system.
• If you have access to run-time debugging
tools, step through the code in question while
monitoring important variables.
• Examine the Salvo source code – it may
contain information not presented elsewhere.

Most importantly, examine your assumptions! Don't assume, for
example, that a call to OSStartTask() is working until you've
confirmed that it is in fact returning an error code of OSNOERR.

Salvo User Manual

427

Compile-Time Troubleshooting
I'm just starting, and I'm getting lots of errors.
Be sure to place
#include 

at the start of each source file that uses Salvo.

My compiler can't find salvo.h.
Make sure that your compiler's include search paths contain the
Pumpkin\Salvo\Inc directory.

My compiler can't find salvocfg.h.
Each project needs a project-specific salvocfg.h. Create one from
scratch or copy one from another project. salvocfg.h normally
resides in your current working directory – you may need to instruct your compiler to explicitly search this directory.
If you are using a Salvo freeware library, copy its salvocfg.h to
your working directory and edit it as needed.

My compiler can't find certain target-specific header files.
This problem may arise if your compiler has no generic target
processor header file that uses defined symbols to include the appropriate target-specific header file. The solution is to include the
target-specific header file in your salvocfg.h.

My compiler can't locate a particular Salvo service.
You must either include the Salvo files in your project or link to a
Salvo library. See your compiler's Salvo Compiler Reference
Manual for more information.

428

Chapter 11 • Tips, Tricks and Troubleshooting

Salvo User Manual

My compiler has issued an "undefined symbol" error for a
context-switching label that I've defined properly.
This may be happening if you have the context-switching label in
unreachable code and your compiler has removed the unreachable
code through optimization. For example, OS_Delay() below is unreachable because of an innocuous error:
if (speed = 0) { // Error – should be "=="
outPWM = 0;
}
else
{
outPWM = 1;
OS_Delay(speed);
…
}

and your compiler may be unable to find label as a result. Change
your code to make the context switch reachable114 and the error
should disappear.

My compiler is saying something about OSIdlingHook.
The configuration options in your salvocfg.h may be set to enable the user hook function, OSIdlingHook(). In a source-code
build, you must define a function with this name. For example,
void OSIdlingHook(void)
{
;
}

is a null (i.e. "do-nothing") function that satisfies this requirement.

My compiler has no command-line tools. Can I still build a
library?
You can build a library without access to a command-line librarian115 by creating a project with all of the Salvo source files, and
setting the output type of your compiler to be a library file. You
will also need a special salvocfg.h file that looks something like
this:
114
115

Salvo User Manual

Use if ( speed = = 0 ) instead of if ( speed = 0 ).
CodeWarrior v3.1 has no command-line tools, but can build a library from a
project.

Chapter 11 • Tips, Tricks and Troubleshooting

429

#define OSUSE_LIBRARY

TRUE

#define OSLIBRARY_TYPE
#define OSLIBRARY_CONFIG
#define OSLIBRARY_VARIANT

OSL
OST
OSNONE

#undef OSMAKE_LIBRARY
#define OSMAKE_LIBRARY

TRUE

This works as follows: when you set OSUSE_LIBRARY to TRUE in
your project's header file salvocfg.h, the library header file salvolib.h will be included in your project. By defining the library
type, configuration and variant symbols T, C and V, respectively,
and by setting OSMAKE_LIBRARY to TRUE, the Salvo source code is
configured for library building.
This method is inefficient for building multiple libraries. For that,
refer to Salvo's makefiles.

Run-Time Troubleshooting
Nothing's happening.
Did you remember to:
•
•
•
•
•
•
•

• Call OSInit()?
• Set OSCOMPILER, OSTARGET and OSTASKS
correctly in your salvocfg.h?
• Create at least one task with
OSCreateTask()?
• Choose valid task pointers and task priorities
that are within the allowed range?
• Call the Salvo scheduler OSSched() from
inside an infinite loop?
• Task-switch inside each task body with a call
to OS_Yield(), OS_Delay(), OS_WaitXyz() or
another context-switcher?
• Structure each task with its body in an infinite
loop?

If you've done all these things and your application still doesn't appear to work, you may have a configuration problem (e.g. parts of
your salvocfg.h do not match those used to create the freeware
library you're using) or an altogether different problem.

430

Chapter 11 • Tips, Tricks and Troubleshooting

Salvo User Manual

Also, make sure that you've done a full recompile ("re-make"),
and, if you're using some sort of integrated development environment, be sure that you've downloaded your latest compiled code
and reset the processor before running the new code.

It only works if I single-step through my program.
This is usually indicative of a problem with interrupts or the
watchdog timer. Since both are usually disabled when singlestepping with an in-circuit emulator (ICE) or in-circuit debugger
(ICD), your application may work in this mode but not in run
mode.
If your application uses interrupts, be sure that any interrupt flags
are cleared before leaving the ISR. When interrupt sources share
the same interrupt vector, failing to clear the interrupt flag will result in an endless loop of interrupt services. In general, vectored
interrupts do not have interrupt flags associated with them.
Many target processors enable the watchdog timer by default. If
you fail to reset it regularly, your application will appear to be constantly resetting itself. Depending on the watchdog timer's timeout
period, this may be a very short (e.g. < 1s) period. Either disable
the watchdog timer or use Salvo's OSCLEAR_WATCHDOG_TIMER()
configuration option.

Note All Salvo projects in the distributions are compiled with
defined to reset the watchdog timer.
This way, even if you forget to disable the watchdog timer116 in
your development environment, the application should still work.
OSCLEAR_WATCHDOG_TIMER()

It still doesn't work. How should I begin debugging?
If you have the ability to set breakpoints, a quick way to verify that
your application is multitasking is to re-load your executable (e.g.
hex) code, place breakpoints at the entry of each task, reset the
processor, and Run. If you have successfully initialized Salvo and
created tasks (check the error return codes for OSInit() and
OSCreateTask()), the first call to OSSched() should eventually
result in the processor halting at one of those breakpoints.
116

Salvo User Manual

In the Microchip development tools family, the PICMASTER and the
MPLAB-ICE disable the watchdog timer by default, but the MPLAB-ICD
enables it by default.

Chapter 11 • Tips, Tricks and Troubleshooting

431

If your application makes it this far, Salvo's internals are probably
working correctly, and your problem may have to do with improper task structure and/or use of Salvo's context-switching services. Improper control of interrupts and incorrectly-written
interrupt service routines (ISRs) are also a common problem.
If you do not have hardware debugging support, use simple methods (like turning an LED on or off from within a task) to trace a
path through your program's execution. On small, embedded systems, "printf-style debugging" may not be a viable option, or
may introduce other errors (like stack overflow) that will only frustrate your attempts to get at the root of the problem.

My program's behavior still doesn't make any sense.
You may be experiencing unintended interaction with your processor's watchdog timer. This can occur if you've compiled your application with the target processor's default (programmable)
configuration, which may enable the watchdog timer. You can
avoid this problem by using the OSCLEAR_WATCHDOG_TIMER() configuration option in your salvocfg.h configuration file. By defining this configuration option to be your target processor's
watchdog-clearing instruction, the Salvo scheduler will clear the
watchdog each time it's called, and prevent watchdog timeouts.

Compiler Issues
Where can I get a free C compiler?
Borland's C++ compilers can be had for free at:
•

http://www.borland.com/bcppbuilder/freecompil
er/

They can be used to create 16- and 32-bit PC (x86) applications.
HI-TECH software also offers free C compilers:
• http://www.htsoft.com/
•
Pacific C can be used to create PC (x86) applications, and PICC
Lite can be used on the Microchip PIC16C84 family.

432

Chapter 11 • Tips, Tricks and Troubleshooting

Salvo User Manual

Where can I get a free make utility?
You can download the GNU make utility's source code from
•

http://www.gnu.org/order/ftp.html

A precompiled DOS/Win32 version is available at
•

ftp://ftp.simtel.net/pub/simtelnet/gnu/djgpp/v2g
nu/

Look for the mak*.zip files. This is a full-featured, UNIX-like
make that works well in the Win32 environment.

Where can I get a Linux/Unix-like shell for my Windows
PC?
You can download the Cygwin bash shell from RedHat at
•

http://sources.redhat.com/cygwin/

A full installation will contain GNU make and many other utilities.
It works best on Windows NT / 2000 / XP systems. If you have the
Salvo Pro, this shell can be used to generate all of Salvo's libraries
on a Windows PC.

My compiler behaves strangely when I'm compiling from
the DOS command line, e.g. "This program has
performed an illegal operation and will be terminated."
The DOS command line is limited to a maximum of 126 characters. If you invoke your compiler with a longer command line, you
may experience very unpredictable results. The solution is to reorganize your project. Consult your compiler's user's manual for
more information.
Another possibility is that the environment size on your Windows/DOS PC is inadequate for the DOS program(s) you are running. If you run more than one DOS window under Windows and
the environment size is marginal, you may also encounter this
problem. You can fix this by adding the shell command to your
config.sys file, e.g.:
shell = c:\windows\command.com /p /e:nnnnn

Salvo User Manual

Chapter 11 • Tips, Tricks and Troubleshooting

433

where nnnnn is the size of the environment, in bytes, from 160 to
32768. The default is 256. See your DOS manual for more information on the DOS command interpreter and the shell command.

My compiler is issuing redeclaration errors when I
compile my program with Salvo's source files.
If you create your application by compiling and then linking your
files and Salvo's source files all at once, be sure that none of your
source files have the same name as any Salvo source file.

HI-TECH PICC Compiler
Salvo has been thoroughly tested with PICC and it is unlikely that
you will encounter any problems that are due directly to compiling
and linking the Salvo code to your application. However, since it is
often difficult to pinpoint the exact cause of a compile-and-link
error, you should follow the tips below if you encounter difficulties.

Running HPDPIC under Windows 2000 Pro
Some people like to run HPDPIC117 in an 80x50 "DOS window"
under Windows. Do the following:
•
•
•
•
•
•
•

• start HPDPIC
• right-click on the menu bar and select
Properties
• select Layout
• choose a Window Size of Width:80 and
Height:50
• select OK, choose "Save properties for
future windows with same title", select OK
• exit HPDPIC (alt-Q)
• restart HPDPIC

You may want to choose a different font or font size (under Properties → Font) that is better suited to a larger DOS window. If
you are having problems with your mouse, instead of changing the
window size settings in the procedure above, deselect the QuickEdit mode under Properties → Options.

117

434

The HI-TECH Integrated Development Environment (IDE) for PICC.

Chapter 11 • Tips, Tricks and Troubleshooting

Salvo User Manual

Setting PICC Error/Warning Format under Windows 2000 Pro
In Windows 2000 Pro, do either:
•
or

•
•

My Computer → Properties → Advanced →
Environment Variables ...

Start → Settings → Control Panel →
System → Advanced → Environment
Variables ...

•
then in User Variables for Userid do:
•

New → Variable, enter HTC_ERR_FORMAT
, OK,
Variable Value, enter Error[ ] %f %l : %s , OK

•

New → Variable, enter
HTC_WARN_FORMAT , OK, Variable
Value, enter Warning[ ] %f %l : %s , OK

and

Then log off and log back on for these changes to take effect. You
can see that they are in force by running the MS-DOS Prompt
(C:\WINNT\system32\command.com) and entering the SET command. Type EXIT to leave the MS-DOS command prompt.
Note that you must log off and log back on for these changes to
take effect. If you change the environment variables without logging off and back on, MPLAB may behave strangely, like do nothing when you click on the error/warning message.

Linker reports fixup errors
If the PICC linker is unable to place variables in RAM, it will report fixup errors. Interpreting these errors can be very difficult.
You must successfully place all variables in RAM before attempting to interpret any other PICC link errors. If you're having difficulty, the simplest thing is to place all of Salvo's variables in an
unused bank (e.g. Bank 3 on a PIC16C77). Then, by using PICC's
bank directives you can move your own variables around until they
all fit. A thorough understanding of the bank directives is required,
especially when banked (or unbanked) pointers to banked (or un-

Salvo User Manual

Chapter 11 • Tips, Tricks and Troubleshooting

435

banked) objects are involved. Consult the PICC manual for more
information, or the Salvo source code for examples of using the
bank directives.
See also "Placing Variables in RAM", below.

Placing variables in RAM
Because PICs have generally very little RAM, as your application
grows it's likely that you will need to explicitly manage where
variables are located in RAM. If your Salvo application has more
than a few tasks and events, it's likely that you will want to place
the Salvo data structures (e.g. tcbs and ecbs) and other variables in
a RAM memory bank other than Bank 0, the default bank for auto
variables and parameters. To do this, use the OSLOC_Xyz configuration options and recompile your code. The OSLOC_Xyz configuration words options not all be the same – for example you can place
ecbs in Bank 2, and tcbs in Bank 3.
If you need to use more than one bank to place Salvo's variables in
RAM, for best performance place them in bank pairs – e.g. in
Banks 2 and 3 only.

Note Your Salvo code will be smallest if you place all of your
Salvo variables in Bank 1 and/or Bank 0. PICC places all auto
variables in Bank 0. Bank switching is minimized by placing
Salvo's variables in the same bank as the auto variables.

Link errors when working with libraries
If you get the following error:
HLINK.EXE::Can't open (error): : No such file or
directory

while working with multiple projects and libraries, it may go away
be simply re-making the project.

Avoiding absolute file pathnames
Use HPDPIC's Abs/Rel path feature when adding source and include files to your project. You'll be able to enter path names much
more quickly.

436

Chapter 11 • Tips, Tricks and Troubleshooting

Salvo User Manual

Compiled code doesn't work
Make sure you're using the latest version of PICC, including any
patches that are available. Check http://www.htsoft.com for version updates.

PIC17CXXX pointer passing bugs
On the 17C756, in certain cases PICC failed to correctly dereference pointers passed as parameters. This affected Salvo's queueing
routines.

Note This was fixed in PICC v7.84.
While() statements and context switches
You may encounter a subtle problem if you use a while() statement immediately following a Salvo context switch, e.g.
...
OS_Delay(5);
while (rxCount) {
...

if rxCount is a banked variable, after optimization the compiler
may fail to set the register page bits properly when accessing the
variable. This will probably lead to incorrect results. A simple
workaround is to add the line
rxCount = rxCount;

between the context switch and the while() statement. This will
"force" the proper RP bits.

Note This was fixed in PICC v7.85.
Library generation in HPDPIC
If you are using HPDPIC projects to compile libraries for use with
PIC processors with different numbers of ROM and RAM banks
(e.g. PIC16C61 and PIC16C77), you may encounter an error when
linking your application(s) to one of those libraries. This is because
the PICC preprocessor CPP.EXE may be fed the wrong processorselection argument if you're switching between projects with different processors.

Salvo User Manual

Chapter 11 • Tips, Tricks and Troubleshooting

437

The solution is to first load a project whose output is a .COD file,
and then load a second project destined for the same type of processor and whose output is a library. Make the library (i.e. make the
second project), then re-load the first project, and make it, linking
to the previously generated library. By loading the first project you
correctly set the processor type for the second project.

Note This was fixed in PICC v7.86.
Problems banking Salvo variables on 12-bit devices
On the 12-bit devices (e.g. PIC16C57), Salvo applications don't
work when Salvo variables are placed in a RAM bank other than
Bank 0. The solution is to upgrade to the latest version of the compiler.

Note This was fixed in PICC v7.86PL4.
Working with Salvo messages
Salvo messages are passed via void pointers. Use the predefined
type definition (typedef) OStypeMsgP when declaring pointers to
messages. This type is defined by default as void *. In PICC a
pointer to a void object points only to RAM. That's fine if your
Salvo application has only messages in RAM. But what if you
want to send messages which point to objects in ROM (e.g. a string
like "STOP' or "GO") as well as RAM?
By changing
OSMESSAGE_TYPE to const messages can now point to objects in
RAM or ROM. This may add 1 extra byte to the size of each event
control block (ecb).
OSMESSAGE_TYPE
must be set to const in your salvocfg.h if you are using messages and/or message queues and you

Note

are accessing message data that's in ROM.
See also Working with Message Pointers in this chapter.

Adding OSTimer() to an Interrupt Service Routine
If you are linking to a freeware or custom Salvo library, or if
timer.c is one of the nodes in your project, and you call
OSTimer() from within an interrupt routine, PICC automatically

438

Chapter 11 • Tips, Tricks and Troubleshooting

Salvo User Manual

assumes the worst case with regard to register usage within
OSTimer() and the functions it may call, and automatically adds a
large number of register save and restores to your interrupt routine.
This makes it large and slow, which is undesirable.
The solution is to change the organization of your source files. Instead of compiling timer.c into a linkable object module, include
it in your source file which contains the call to OSTimer(). For
example, your main.c might now look like this:
…
#include "timer.c"
void interrupt intVector( void )
{
/* handle various interrupts
...

*/

/* this happens every 10ms.
*/
if (TMR1IF) {
/* must clear TMR2 interrupt flag. */
TMR1IF = 0;
/* reload TMR1 while it's stopped. */
TMR1ON = 0;
TMR1 -= TMR1_RELOAD;
TMR1ON = 1;
OSTimer();
}
}

By including timer.c in the same source code file as the interrupt
routine, PICC is able to deduce exactly which temporary registers
must be saved when the interrupt occurs and restored thereafter,
instead of assuming the worst case and saving and restoring all of
them. The resultant savings in code space and improvement in interrupt execution speed are substantial. If your application uses the
Salvo timer, this reorganization is highly recommended.
After including timer.c in your interrupt source code file, you
may want to recompile your custom Salvo library if you are using
one. The Salvo functions will still be able to reference the required
queueing functions – they've simply moved from the library to
your object modules.

Note You may need to add the switch –IPumpkin\Salvo\Src to
PICC's command line in order for the compiler and linker to find
the timer.c source file.

Salvo User Manual

Chapter 11 • Tips, Tricks and Troubleshooting

439

Using the interrupt_level pragma
Whenever you call any Salvo services from both inside an interrupt
and from background code (e.g. from within a task), you must insert the following PICC directive prior to your interrupt routine:
#pragma interrupt_level 0

This alerts the PICC compiler to look for multiple call graphs of
functions called from both mainline and interrupt code. This is
necessary in order to preserve parameters and auto variables.

Note Placing this PICC pragma before an interrupt routine has no
deleterious effects even when multiple call graphs are not generated. Therefore it's recommended that you always do this if you
call any functions from within your interrupt routine.

HI-TECH V8C Compiler
Note Support for the V8C compiler has been discontinued as of
2005.
The initial Salvo port to the VAutomation V8-µRISC™ requires an
updated V8 assembler, ht-v8\bin\asv8.exe, dated 6-21-2001 or
later, along with v7.84 of the compiler. Many of the test programs
(e.g. \salvo\test\t41\sysl) use printf() for run-time output
for use with the simulators.

Note Since the HI-TECH V8C compiler and its HPDV8 IDE are
substantially similar in operation to HI-TECH's PICC compilers
and HPDPIC IDE, refer to HI-TECH PICC Compiler, above, for
related information.

Simulators
Two simulators for the V8-µRISC™ are available – one from HITECH (simv8.exe) and one from VAutomation (v8sim.exe).
Salvo applications run on both.

440

Chapter 11 • Tips, Tricks and Troubleshooting

Salvo User Manual

HI-TECH 8051C Compiler

Problems with static initialization and small and medium memory
models.
When using the small or medium memory models, the compiler
issues the error Can't generate code for this expression
when faced with the declaration
unsigned int counter = 0;

This occurs because initialized objects are in ROM for these models, and therefore cannot be changed. The solution is to either declare the variable as near, or explicitly initialize it elsewhere in
your code.

IAR PICC Compiler

Target-specific header files
The IAR PICC compiler requires a target-specific header file that
contains symbols and addresses for the PICmicro special function
registers (SFRs). These files are located in the inc subdirectory of
the compiler's distribution, and are target-specific.
For example, \iar\ew23\picmicro\inc\io17c756.h is the
header file for the 17C756 PICmicro. By placing
#include "io17C756.h"

in your source files, the compiler will be able to correctly resolve
certain symbols used throughout the Salvo source code.

Interrupts
The vector for each interrupt must be properly defined. Use the
compiler's vector pragma like this:
#pragma vector=0x10
__interrupt void intVector(void)
{
T0IF = 0;
TMR0 -= TMR0_RELOAD;
OSTimer();

Salvo User Manual

Chapter 11 • Tips, Tricks and Troubleshooting

441

}

This will place the TMR0 interrupt vector at 0x10 on a
PIC17C756.

Mix Power C Compiler
In contrast to usual IBM C call stack programming, which has
positive offsets from BP for function arguments and negative offsets from BP for local variables, the Power C compiler uses positive offsets from BP to access both local variables and function
arguments. This affects the Salvo context switcher for Power C to
the degree that it will only function correctly as long as the call
stack for the task is in its simplest form. The key to compiling
Salvo applications to run on the PC is to guarantee that each task
has the simplest possible Power C entry call stack.
Strict adherence to the Salvo requirement that only static local
variables be used in a task is required to avoid run-time errors. Additionally, there are a few other innocuous things ("gotchas") that
the Power C programmer might do which violate Salvo's requirement that the call stack remain in its simplest form. Those that are
known are outlined below.

Required compile options
When compiling Salvo source code, using the following compile
options for PC.EXE:
/r/2
/mm
Failure to use these options or to use other incompatible options
may prevent your Salvo executable from running properly.
Below is an example line from a makefile:
PCopts = /c /o /w /r- /2 /mm /id:Pumpkin\Salvo\Inc

Application crashes after adding long C source lines to a Salvo
task
If you have source code (e.g. a function with multiple parameters)
within a task that is too long to fit on a single line, you must use

442

Chapter 11 • Tips, Tricks and Troubleshooting

Salvo User Manual

the '\' character to continue on the next line, even if it's not necessary for a successful compile. This is because Mix Power C
changes the task's entry call stack to one that is incompatible with
Salvo's context switcher if the line is not continued with the '\'
character. For example, the call to DispLCD() below
void TaskMsg ( void )
{
while (1) {
...
DispLCD((char *) ((t_dispMsg *)msgP)->strTop,
(char *) ((t_dispMsg *)msgP)->strBot);
OS_Delay((OStypeDelay)
((t_dispMsg *)msgP)->delay);
...
}
}

will compile successfully, but it will cause the PC application to
crash when it runs TaskMsg(). By adding the '\' character to the
DispLCD() line. e.g.
DispLCD((char *) ((t_dispMsg *)msgP)->strTop, \
(char *) ((t_dispMsg *)msgP)->strBot);

the problem is resolved.

Application crashes after adding complex expressions to a Salvo
task
Mix Power C changes the task's entry call stack if the expressions
in a task exceed a certain level of complexity. For example, placing either
char = RxQ[rxHead++];

or
(dummy = dummy);

inside a task will cause problems, whereas replacing them with
char = RxQ[rxHead];
rxHead++;

and
dummy = dummy;

Salvo User Manual

Chapter 11 • Tips, Tricks and Troubleshooting

443

will not.

Application crashes when compiling with /t option
Mix Power C changes the task's call entry stack when trace information for the debugger is enabled via the compiler's /t option.
This change is incompatible with Salvo's context switcher for
Power C. Source code modules which contain Salvo tasks must not
be compiled with the /t option.
One way around this problem is to move functionality that does
not involve context switching out of the module the task is in and
into a separate source code module, and call it as an external function from within the task. A module that does not contain any
Salvo tasks can be compiled with the /t option, and hence debugged using Mix Power Ctrace debugger.

Compiler crashes when using a make system
Make absolutely sure that your DOS command line does not exceed 127 characters in length. If it does, the results can be very unpredictable. Simplify your directory structure to minimize
pathname lengths when invoking any of the Mix Power C executables (e.g. PCL.EXE).

Metrowerks CodeWarrior Compiler

Compiler has a fatal internal error when compiling your source
code
Ensure that you do no use duplicate labels in any single source
code file. This may occur unintentionally if you duplicate labels
for Salvo context-switching macros inside a single function. For
example,
void Task1( void )
{
...
OS_Delay(1);
...
}
void TaskB( void )
{
...

444

Chapter 11 • Tips, Tricks and Troubleshooting

Salvo User Manual

OS_Delay(1);
...
OS_Yield();
...
}

may cause a CodeWarrior exception because of the duplicate label
a in Task2(), whereas
void Task1( void )
{
...
OS_Delay(1);
...
}
void Task2( void )
{
...
OS_Delay(1);
...
OS_Yield();
...
}

may not.

Microchip MPLAB

The Stack window shows nested interrupts
The MPLAB Stack window cannot differentiate between an interrupt and an indirect function call. Because Salvo makes extensive
use of indirect function calls, you may be seeing a combination of
return addresses associated with interrupts and indirect function
call return addresses.

Controlling the Size of your Application
The Salvo source code is contained in several files and is comprised of a large body of functions. Your application is unlikely to
use them all. If you compile and link the Salvo source files along
with your application's source files to form an executable program,
you may inadvertently end up with many unneeded Salvo functions in your application. This may prevent you from fitting your
application into the ROM of your target processor.

Salvo User Manual

Chapter 11 • Tips, Tricks and Troubleshooting

445

The solution is to compile the Salvo source files separately, and
combine them into a single library. You can then link your application to this library in order to resolve all the external Salvo references. Your compiler should extract only those functions that your
application actually uses in creating your executable application,
thus minimizing its size.
You must always recreate the Salvo library in its entirety whenever
you change any of its configuration options.
Refer to your compiler's documentation on how to create libraries
from source files, and how to link to those libraries when creating
an executable.
See Chapter 4 • Tutorial for more information on compiling your
Salvo application.

Working with Message Pointers
If you want to use messages as a means of intertask communications, you'll have to be comfortable using Salvo message pointers.
Salvo provides predefined type definitions (C typedefs) for working with message pointers. The following message pointer declarations are equivalent:
OStypeMsg * messagePointer;

and
OStypeMsgP messagePointer;

but you should always use the latter to declare local or global message pointer variables, both static and auto.
In general, Salvo message pointers are of type void *. However,
you should use the predefined types to avoid problems when a void
pointer is not correct for a message pointer. This occurs mainly
with processors that have banked RAM.
When passing an object that is not already a message pointer,
you'll need to typecast the object to a message pointer in order to
avoid a compiler error. The following two calls to OSSignalMsg()
are equivalent:
OSSignalMsg(MSG1_P, (OStypeMsg *) 1);

446

Chapter 11 • Tips, Tricks and Troubleshooting

Salvo User Manual

and
OSSignalMsg(MSG1_P, (OStypeMsgP) 1);

The typecast above is required because 1 is a constant, not a message pointer. Here are some more examples of passing objects that
are not message pointers:
char letter = ‘c';
OSSignalMsg(MSG_CHAR_VAR_P, (OStypeMsgP) &letter);
const char CARET = ‘^';
OSSignalMsg(MSG_CHAR_CONST_P, (OStypeMsgP)
&CARET);
unsigned int * ptr;
OSSignalMsg(MSG_UINT_P, (OStypeMsgP) ptr);
void Function(void);
OSSignalMsg(MSG_FN_P, (OStypeMsgP) Function);

Once an object has been successfully passed via a message, you
will probably want to extract the object from the message via
OS_WaitMsg().118 When a task successfully waits a message,
Salvo copies the message pointer to a local message pointer (msgP
below) of type OStypeMsgP. To use the contents of the message,
you'll need to properly typecast and dereference it. For the examples above, we have:
char:

* (char *) msgP

const char:

* (const char *) msgP

unsigned int *:

(unsigned int *) msgP

void * (void):

(void * (void)) msgP

Failing to properly typecast an object (e.g. using (char *) instead
of (const char *) when dereferencing a constant) will have unpredictable results. Please see Salvo Application Note Error! Reference source not found. for more information on dereferencing
pointers.

NOTE When working with message pointers, it's very important
to ensure that Salvo's message pointer type OStypeMsgP is properly configured for the kinds of messages you wish to use. On most

118

Salvo User Manual

An exception occurs when you are not interested in the contents of the
message, but only that it has arrived.

Chapter 11 • Tips, Tricks and Troubleshooting

447

targets, the default configuration of void * will suffice … but
there are some exceptions.
For example, the HI-TECH PICC compiler requires 16 bits for
const char pointers, but only 8 bits for char pointers. Therefore
the Salvo code (whether in a library or in a source-code build)
must be configured to handle these larger pointers or else you will
encounter runtime errors.

448

Chapter 11 • Tips, Tricks and Troubleshooting

Salvo User Manual

Appendix A • Recommended
Reading

Salvo Publications
A variety of additional Salvo publications are available to aid you
in using Salvo. They include Application Notes, Where applicable,
some are included in certain Salvo distributions. Application
Notes, Assembly Guides, Compiler Reference Manuals, Conference proceedings & presentations, and others. They are all available online at http://www.pumpkininc.com.

Learning C
K&R
Kernighan, Brian W., and Ritchie, Dennis M., The C Programming
Language, Prentice-Hall, New Jersey, 1978, ISBN 0-13-110163-3.

Of Interest This book is the definitive, original reference for
the C programming language.

C, A Reference Manual
Harbison, Samuel P. and Steele, Guy L., Jr., C, A Reference Manual, Prentice-Hall, NJ, 1995, ISBN 0-13-326224-3.

Of Interest A modern C language reference.
Power C
Mix Software, Power C, The High-Performance C Compiler,
1993.

Salvo User Manual

449

Of Interest Mix Power C is a very inexpensive, full-featured
ANSI-compatible C compiler for use on the PC. Its excellent
600+-page manual contains comprehensive tutorial and reference
sections. Library source code is available.

Real-time Kernels
µC/OS & MicroC/OS-II
Labrosse, Jean J., µC/OS, The Real-Time Kernel, R&D Publications, Lawrence, Kansas, 1992, ISBN 0-87930-444-8.
Labrosse, Jean J., MicroC/OS-II, The Real-Time Kernel, R&D
Books, Lawrence, Kansas, 1999, ISBN 0-87930-543-6.

Of Interest This book and its greatly expanded and wellillustrated successor provide an excellent guide to understanding
RTOS internals. It also demonstrates how even a relatively simple
conventional RTOS requires vastly more memory than Salvo. Its
task and event management is array-based. Source code is included.

CTask
Wagner, Thomas, CTask, A Multitasking Kernel for C, public domain software, version 2.2, 1990, available for download on the
Internet.

Of Interest The author of this well-documented kernel takes a
very hands-on approach to describing its internal workings. CTask
is geared primarily towards use on the PC. As such, it is not a realtime kernel. Its task and event management is primarily queuebased. Source code is included.

Embedded Programming
Labrosse, Jean J., Embedded Systems Building Blocks, R&D Publications, Lawrence, Kansas, 1995, ISBN 0-13-359779-2.

450

Appendix A • Recommended Reading

Salvo User Manual

Of Interest This book provides canned routines in C for a variety of operations (e.g. keypad scanning, serial communications and
LCD drivers) commonly encountered in embedded systems programming. RTOS- and non-RTOS-based approaches are covered.
The author also provides an excellent bibliography. Source code is
included.
LaVerne, David, C in Embedded Systems and the Microcontroller
World, National Semiconductor Application Note 587, March
1989, http://www.national.com.

Of Interest The author's comments on the virtues of C programming in embedded systems are no less valid today than they
were in 1989.

RTOS Issues

Priority Inversions
Kalinsky, David, "Mutexes Prevent Priority Inversions," Embedded Systems Programming, Vol. 11 No. 8, August 1998, pp.76-81.

Of Interest An interesting way of solving the priority inversion
problem.

Microcontrollers
PIC16
Microchip, Microchip PIC16C6X Data Sheet, Section 13.5, Interrupts, 1996.

Of Interest A special method for disabling the global interrupt
bit

required on the PIC16C61/62/64/65. Set
OSPIC16_GIE_BUG to TRUE when using these and certain other
processors. The later versions (e.g. PIC16C65A) do not require this

Salvo User Manual

GIE

is

Appendix A • Recommended Reading

451

fix. Below is a response from Microchip to a customer query on
this issue:
The GIE issue is not a 'bug' in the part it relates more to an operational consideration when the GIE bit is handled in software to disable the interrupt system
and the fact that during execution of that operation it is possible for an interrupt
to occur. The nature of the MCU core operation means that whilst the current
instruction is flowing through the device an asynchronous interrupt can occur.
The result of this is that the processor will vector to the ISR disable GIE, handle
the Interrupt and then enable GIE again. The result of this is of course that the
instruction to disable GIE has been overridden by the processor vectoring to the
interrupt and disabling then enabling the interrupt. This is a very real possibility
and AN576 is explaining a method to ensure that, in the specific instance where
you wish to disable GIE in software during normal execution that your operation
has not been negated by the very action you wish to stop.
The app note is related to the disabling of GIE in software. The disabling and reenabling of GIE when an interrupt occurs is performed in hardware by the processor and the execution of the RETFIE instruction. The GIE check is a safeguard
to ensure your expected/desired operation has occurred and your program can
then operate as expected/desired without the unexpected occurrence of an interrupt. This issue remains on the current range of parts since it is related to the
operation of the core when the user wishes to take control of the interrupt system
again.
BestRegards,
UK Techhelp

452

Appendix A • Recommended Reading

Salvo User Manual

Appendix B • Other Resources

Web Links to Other Resources
Here are some web sites for information and products related to
Salvo and its use:
•
•
•

•
•

•
•
•
•
•
•
•
•

• http://www.atmel.com/ – Atmel Corporation,
supplier of 8051 architecture and AVR 8-bit
RISC microcontrollers
• http://www.circuitcellar.com/, "The magazine
for Computer Applications," – lots of
information on computer and embedded
computer programming
• http://www.cygnal.com/ – Cygnal Integrated
Products, supplier of advanced in-system
programmable, mixed-signal System-on-Chip
products
• http://www.embedded.com/ – Home of
Embedded Systems Programming magazine
• http://www.gnu.org/ – The Free Software
Foundations GNU119 project web server
• http://www.htsoft.com/ – HI-TECH Software
LLC, home of the PICC, PICC Lite, PICC-18
and V8C compilers.
• http://www.iar.com/ – IAR Systems, makers of
embedded computing tools including C
compilers, Embedded Workbench IDE and CSPY debugger

•

119

Salvo User Manual

GNU is a recursive acronym for ``GNU's Not Unix''; it is pronounced "guhNEW".

453

•
•
•

•
•
•
•
•
•
•
•

•
•
•
•

•
•
•
•
•
•

120

454

• http://www.imagecraft.com/ – ImageCraft,
makers of ANSI C tools combined with a
modern GUI development environment
• http://www.keil.com/ – Keil Software, makers
of C compilers, macro assemblers, real-time
kernels, debuggers, simulators, integrated
environments, and evaluation boards for the
8051
• http://www.metrowerks.com/ – Metrowerks
Corporation, home of the CodeWarrior compiler
and integrated development environment
• http://www.microchip.com/ – Microchip
Corporation, supplier of PIC microcontrollers
• http://www.mixsoftware.com/ – Mix Software,
Inc., home of the Power C compiler
• http://www.motorola.com/ – Motorola, Inc.,
makers of M68HCxx single-chip
microcontrollers and providers of the
Metrowerks CodeWarrior IDE
• http://www.mixsoftware.com/ – Mix Software,
Inc., home of the Power C compiler
• http://www.quadravox.com/ – Quadravox,
Inc., makers the AQ430 Development Tools for
TI's MSP430 line of ultra-low-power
microcontrollers
• http://www.redhat.com/ – Provider of a wellknown Linux distribution, and also home of the
Cygwin120 project.
• http://www.rowley.co.uk.com/ – Rowley
Associates, makers development tools for TI's
MSP430
• http://www.ti.com/ – Texas Instruments,
makers of the TMS320C family of DSPs as well

Search site for "Cygwin".

Appendix B • Other Resources

Salvo User Manual

as the MSP430 line of ultra-low-power
microcontrollers
•
•

Salvo User Manual

• http://www.vautomation.com/ – VAutomation,
Inc., home of the V8-µRISC™ synthesizeable
8-bit core

Appendix B • Other Resources

455

456

Appendix B • Other Resources

Salvo User Manual

Appendix C • File and Program
Descriptions

Overview
Each Salvo distribution contains a variety of files in different formats. Most (e.g. Salvo libraries and project files) are intended for
use with a particular set of tools and on a particular target, although some – e.g. the Salvo source code – are often universal.
Each distribution has an organized file hierarchy. Directories (i.e.
folders) include subdirectories (i.e. subfolders), etc. Files that are
higher up in a particular directory tree are more general, and those
towards the bottom are more specific for a particular target, compiler and / or Salvo distribution.
If you have installed only one Salvo distribution on your PC, it will
contain files for just your compiler and / or target processor. If you
have installed multiple Salvo distributions, all of their files will be
distributed in subdirectories below the root Salvo directory.

Online File Locations
Salvo Distributions
Unless otherwise noted, each complete Salvo distribution is distributed as a Windows self-extracting executable,121 and is available
online
exclusively
at
Pumpkin’s
website
http://www.pumpkininc.com for download. Salvo Lite is freely
downloadable. Salvo LE and Salvo Pro are only available to those
Salvo customers who have purchased the corresponding Salvo license(s).

121

Salvo User Manual

In Salvo v3 and earlier, Salvo’s installer was built using MindVision’s
Installer VISE product. As of Salvo 4, Salvo installers are built using the
NSIS system.

457

Local/User File Locations
By default, Salvo is always installed in the \Pumpkin directory of
the user’s hard disk. Therefore Salvo’s various installation directo\Pumpkin\Salvo,
\Pumpkin\Salvo\Doc,
ries
are
\Pumpkin\Salvo\Inc, etc.

Note Due to a variety of problems that may be encountered122 if
installing Salvo to the Windows Program Files directory (due to
the space character in said directory’s name), installation to a root
directory other than \Pumpkin\Salvo is not recommended.

Salvo Uninstaller(s)
The Salvo uninstaller(s) are located in the Salvo directory. There
is an uninstaller for each unique Salvo distribution installed on the
user PC.

Salvo Documentation
Salvo documents – when included in a Salvo distribution – are located in Salvo\Doc.

Note Not all Salvo documents are included in every Salvo distribution. For example, the Salvo User Manual is not included, due to
its size. An alias file with a link to the on-line (and therefore most
up-to-date) version of the Salvo User Manual is included in each
Salvo distribution.

Salvo Header Files
Salvo's header files are located in the Salvo\Inc directory. All
Salvo header files (*.h) are in written in C.

Salvo Source Files
Salvo's source files located in the Salvo\Src directory. Most of
Salvo's source files (*.c) are in written in C. The remaining
source files (*.asm, *.s, etc.) are written in target- and com-

122

458

E.g. problems for Salvo Pro users with Salvo makefiles. Also, certain
compilers cannot properly handle spaces in e.g. the names of include paths.

Appendix C • File and Program Descriptions

Salvo User Manual

piler-specific assembly language, and are located in designated
subdirectories.

Salvo Libraries
Salvo's target- and compiler-specific libraries (*.lib, *.a, etc.) are
located in the Salvo\Lib directory. Where compiler versions impact the format of libraries, there may be multiple directories of
libraries for a particular series of compilers.

Salvo Applications
Depending on the particular distribution, a Salvo installation may
include applications related to Salvo’s tutorials, examples, test
code or other applications. They are located in the Salvo\Example
directory, and are normally in the form of projects (see Projects,
below) for the associated software toolset.

Salvo Graphics Files
The Salvo installers require some graphics files. These are located
in Salvo\Gfx.

Other Pumpkin Products
Salvo is just one of Pumpkin's software products involving. Other
Pumpkin products will usually be installed alongside Salvo under
the \Pumpkin directory.

Target and Compiler Abbreviations
Salvo employs a shorthand notation when referring to files that are
specific to a particular target and compiler combination. These abbreviations are usually a combination of an abbreviation of the
toolset vendor name and of the target’s name. The implied compiler and target are usually self-explanatory.

Salvo User Manual

Appendix C • File and Program Descriptions

459

Projects
Nomenclature
Nearly all Salvo applications are built using projects.123 Usually
the project type is the one native to the tool being used, e.g. Microchip MPLAB projects (*.mcp) or Keil µVision2 (*.uV2) projects.
Programs can be built using Salvo libraries or Salvo source code.
Projects follow the naming convention shown below:
•
•
•
•
•
•

•
•

projectnamelite.*:
(freeware)

uses Salvo Lite
libraries

projectnamele.*:

•

uses Salvo LE or Pro
(standard) libraries

projectnamepro.*:
uses source code from
a
•
Salvo Pro
distribution
projectnamepro-lib.*: uses Salvo LE or Pro
•
(standard) libraries
with
•
embedded
debugging
•
information

•
Each project has a single, unique salvocfg.h configuration file
associated with it.
Wherever possible, relative pathnames have been used for maximum installation flexibility.
In general, projects designed for a particular target and compiler
system can be easily modified to work with other, similar target
processors. For example, a project for the NXP LPC2129
ARM7TDMI-based MCU could be rebuilt for the NXP LPC2106
with minor changes.
123

460

Some applications may be built via simple makefiles and via the command
line.

Appendix C • File and Program Descriptions

Salvo User Manual

Project Files
The source files for a project are generally unique to a project,
though they may be substantially similar to those of similar projects. The only target-specific code contained in a project’s source
code is unique to the intended target.124
Wherever possible, the projects used to generate the applications
have organized the project’s files into abstracts, help files, source
files and libraries that are unique to the project; Salvo help files,
configuration file, source files and libraries; and other files (e.g.
map files, hex files, etc.).
Additionally, where several projects are grouped together (e.g. the
Salvo Lite, LE and Pro versions of the tutorial project tut5), files
that are common to all of the projects are located in the parent directory of the project files.

124

Salvo User Manual

This methodology differs substantially from that used in Salvo v3 and earlier.
In the earlier projects, target-specific code for the intended target was enabled
via the preprocessor, and all other target-specific code was ignored. This
created substantial confusion among Salvo users, to the point where it was
deemed detrimental to overall comprehension of the Salvo applications.

Appendix C • File and Program Descriptions

461

462

Appendix C • File and Program Descriptions

Salvo User Manual

Index

µ
µC/OS .................................................................. See MicroC/OS-II

A
additional documentation
application notes ............ xxvi, 82, 86, 91, 93, 197, 447, 449, 451
compiler reference manuals ... 51, 82, 93, 94, 103, 104, 105, 106,
107, 111, 168, 196, 197, 369, 391, 395, 403, 428
porting manual ........................................................................ 425
assembly language ...................................................................... xxv
portability.................................................................................. 25

B
build process
library build..................................... 93, 94, 96, 98, 111, 406, 430
source-code build .......... 93, 96, 98, 200, 206, 208, 403, 429, 448

C
C compiler................................................................................... 432
C programming language............................................................ 449
portability.................................................................................. 26
compiler
recompile (re-make)................................................................ 431
required features.......................................................................... 7
search paths............................................................................. 428
complex expressions in Power C ................................................ 443
complexity
application........................................................................... 11, 88
managing................................................................................. 192
scheduler ................................................................................... 19
size vs. speed........................................................................... 166
configuration options
OS_MESSAGE_TYPE........................................................... 164
OSBIG_SEMAPHORES ........................ 113, 185, 187, 281, 341
OSBYTES_OF_COUNTS.............. 114, 152, 185, 187, 319, 385

Salvo User Manual

463

OSBYTES_OF_DELAYS... 87, 89, 90, 115, 117, 124, 125, 176,
185, 187, 208, 210, 211, 253, 301, 319, 353, 385, 386, 407
OSBYTES_OF_EVENT_FLAGS. 102, 116, 135, 185, 187, 257,
275
OSBYTES_OF_TICKS . 117, 125, 160, 176, 185, 187, 208, 211,
212, 245, 297, 299, 331, 333, 353
OSCALL_OSCREATEEVENT .... 118, 119, 120, 121, 122, 186,
188, 271, 275, 277, 279, 281
OSCALL_OSGETPRIOTASK............................................... 121
OSCALL_OSGETSTATETASK ........................................... 121
OSCALL_OSMSGQCOUNT ........................ 121, 186, 188, 303
OSCALL_OSMSGQEMPTY......................... 121, 186, 188, 305
OSCALL_OSRETURNEVENT.... 118, 122, 136, 137, 186, 188,
307, 309, 311, 313, 315, 355, 357, 359, 361
OSCALL_OSSIGNALEVENT ..... 118, 122, 186, 188, 269, 326,
335, 337, 339, 341
OSCALL_OSSTARTTASK................................... 122, 186, 188
OSCLEAR_GLOBALS.................................. 123, 185, 187, 301
OSCLEAR_UNUSED_POINTERS ............... 124, 185, 187, 321
OSCLEAR_WATCHDOG_TIMER().................... 208, 431, 432
OSCOLLECT_LOST_TICKS................................ 125, 185, 187
OSCOMBINE_EVENT_SERVICES .... 126, 186, 187, 241, 269,
271, 275, 277, 279, 281, 325, 335, 337, 339, 341
OSCOMPILER ....... 100, 109, 123, 164, 185, 189, 376, 397, 430
OSCTXSW_METHOD .......................... 127, 172, 186, 188, 189
OSDISABLE_ERROR_CHECKING ............ 129, 133, 185, 353
OSDISABLE_FAST_SCHEDULING ................... 130, 186, 188
OSDISABLE_TASK_PRIORITIES...... 131, 251, 283, 289, 291,
327, 329
OSENABLE_BINARY_SEMAPHORES..... 101, 132, 135, 141,
142, 146, 185, 187, 255, 271, 307, 335, 355
OSENABLE_BOUNDS_CHECKING........................... 133, 174
OSENABLE_CYCLIC_TIMERS . 134, 186, 188, 273, 285, 317,
323, 343, 347, 367
OSENABLE_EVENT_FLAGS.. xxviii, 101, 102, 116, 132, 135,
141, 142, 146, 187, 257, 269, 275, 309, 325
OSENABLE_EVENT_READING 136, 137, 185, 187, 307, 309,
311, 313, 315, 355, 357, 359, 361
OSENABLE_EVENT_TRYING ................... 136, 137, 185, 187
OSENABLE_FAST_SIGNALING ........................ 138, 185, 187
OSENABLE_IDLE_COUNTER............................ 139, 185, 187
OSENABLE_IDLING_HOOK ..... 139, 140, 185, 186, 187, 215,
380
OSENABLE_INTERRUPT_HOOKS.................................... 378
OSENABLE_MESSAGE_QUEUES .... 101, 108, 132, 135, 141,
142, 146, 185, 187, 263, 279, 303, 305, 313, 339, 359

464

Index

Salvo User Manual

OSENABLE_MESSAGES 89, 90, 101, 132, 135, 141, 142, 146,
185, 187, 261, 311, 337, 357
OSENABLE_OSSCHED_DISPATCH_HOOK .... 143, 186, 382
OSENABLE_OSSCHED_ENTRY_HOOK........... 144, 186, 382
OSENABLE_OSSCHED_RETURN_HOOK........ 145, 186, 382
OSENABLE_SCHEDULER_HOOK..................................... 186
OSENABLE_SEMAPHORES ...... 101, 132, 135, 141, 142, 146,
185, 187, 207, 265, 281, 315, 341, 361
OSENABLE_STACK_CHECKING..... 123, 147, 152, 157, 185,
187, 201, 243, 245, 247, 251, 253, 255, 257, 261, 263, 265,
269, 271, 275, 277, 279, 281, 283, 287, 289, 291, 293, 295,
297, 299, 301, 319, 321, 325, 327, 329, 331, 333, 335, 337,
339, 341, 345, 349, 353
OSENABLE_TCBEXT0|1|2|3|4|5 .......... 148, 177, 186, 188, 365
OSENABLE_TIMEOUTS .... 124, 125, 151, 158, 185, 211, 255,
257, 261, 265, 371, 372
OSEVENT_FLAGS........................ 101, 102, 135, 275, 376, 398
OSEVENTS . 88, 89, 90, 101, 110, 132, 135, 141, 142, 146, 157,
174, 185, 187, 227, 255, 257, 261, 263, 265, 269, 271, 275,
277, 279, 280, 281, 301, 307, 309, 311, 313, 315, 325, 335,
337, 339, 341, 355, 357, 359, 361, 376, 398
OSGATHER_STATISTICS .. 114, 139, 147, 152, 158, 161, 185,
187, 207
OSINTERRUPT_LEVEL............................................... 153, 186
OSLIBRARY_CONFIG 103, 104, 105, 106, 107, 111, 186, 188,
397, 400, 408, 430
OSLIBRARY_GLOBALS .... 103, 104, 105, 106, 107, 111, 186,
188, 399
OSLIBRARY_OPTION ................. 103, 104, 105, 106, 107, 111
OSLIBRARY_TYPE..... 103, 104, 105, 106, 107, 111, 186, 188,
397, 399, 408, 430
OSLIBRARY_VARIANT..... 103, 104, 105, 106, 107, 111, 186,
188, 397, 401, 430
OSLOC_ALL.......................................... 154, 156, 186, 188, 203
OSLOC_COUNT.... 154, 156, 157, 158, 159, 160, 186, 188, 387
OSLOC_CTCB ....................................... 154, 157, 186, 188, 387
OSLOC_DEPTH..................................... 154, 157, 186, 188, 387
OSLOC_ECB.................... 89, 154, 157, 184, 186, 188, 386, 387
OSLOC_EFCB ....................................................................... 157
OSLOC_ERR.......................................... 154, 158, 186, 188, 387
OSLOC_GLSTAT .......................................................... 158, 387
OSLOC_LOGMSG................................. 154, 158, 186, 188, 387
OSLOC_LOST_TICK ............................................ 158, 186, 188
OSLOC_MQCB.............. 108, 154, 159, 186, 188, 280, 386, 387
OSLOC_MSGQ.............. 108, 154, 159, 186, 188, 280, 386, 387
OSLOC_PS ............................................. 154, 159, 186, 188, 387

Salvo User Manual

Index

465

OSLOC_SIGQ ........................................ 154, 160, 186, 188, 387
OSLOC_TCB.................. 148, 154, 160, 184, 186, 188, 386, 387
OSLOC_TICK ........................................ 154, 160, 186, 188, 387
OSLOG_MESSAGES .... 158, 159, 161, 162, 163, 185, 187, 188
OSLOGGING 152, 161, 162, 163, 185, 187, 188, 201, 243, 245,
253, 255, 257, 261, 263, 265, 269, 271, 275, 277, 279, 281,
283, 301, 321, 325, 335, 337, 339, 341, 345
OSMESSAGE_QUEUES ...... 101, 108, 142, 159, 185, 227, 279,
280, 376, 398
OSMESSAGE_TYPE..................................... 185, 187, 385, 438
OSMPLAB_C18_LOC_ALL_NEAR ............ 155, 165, 186, 188
OSOPTIMIZE_FOR_SPEED......................... 166, 185, 187, 321
OSPIC16_GIE_BUG .............................................................. 451
OSPIC18_INTERRUPT_MASK.................................... 167, 168
OSRPT_HIDE_INVALID_POINTERS. 169, 170, 171, 185, 187
OSRPT_SHOW_ONLY_ACTIVE......... 169, 170, 171, 185, 187
OSRPT_SHOW_TOTAL_DELAY........ 169, 170, 171, 185, 187
OSRTNADDR_OFFSET................................ 127, 172, 186, 188
OSSCHED_RETURN_LABEL()........................................... 173
OSSET_LIMITS ..................................................... 133, 174, 399
OSSPEEDUP_QUEUEING ................................... 175, 185, 187
OSTARGET.................................... 100, 109, 185, 189, 397, 430
OSTASKS 66, 87, 89, 90, 96, 101, 110, 185, 186, 214, 219, 220,
301, 376, 397, 398, 430
OSTIMER_PRESCALAR89, 115, 117, 176, 185, 186, 187, 209,
210, 211, 212, 353, 407
OSUSE_EVENT_TYPES ..... 179, 185, 187, 269, 271, 275, 277,
279, 281, 319, 325, 335, 337, 339, 341
OSUSE_INLINE_OSSCHED ........ 180, 181, 186, 188, 201, 321
OSUSE_INLINE_OSTIMER . 180, 182, 186, 188, 201, 232, 353
OSUSE_INSELIG_MACRO.................................. 180, 183, 240
OSUSE_LIBRARY . 94, 103, 104, 105, 106, 107, 111, 186, 188,
396, 397, 408, 430
OSUSE_MEMSET ................................................. 184, 186, 188
OSUSTOM_LIBRARY_CONFIG ......... 128, 186, 188, 407, 408
conflicts
deadlock .................................................................................... 38
priority inversion............................................................... 39, 451
context switch ............................................................................... 12
critical section ............................................................................... 18
CTask .......................................................................................... 450
custom libraries..............................................................See libraries

D
debugging.................................................................................... 427

466

Index

Salvo User Manual

breakpoints.............................................................................. 431
delay..................................................................................... See task

E
event flags ............................................................................. 13, 223
events ............................................................................................ 13
response time ............................................................................ 20
examples
how to
allow access to a shared resource........................................ 271
ascertain which event flag bit(s) are set .............................. 310
avoid overfilling a message queue.............................. 304, 306
build a library without command-line tools........................ 429
change a cyclic timer's period on-the-fly............................ 324
change a task’s priority on-the-fly ...................................... 252
change a task's priority from another task .......................... 330
check a message before signaling ....................................... 312
clear an event flag after successfully waiting it .................. 270
context-switch outside a task's infinite loop ....................... 334
context-switch unconditionally........................................... 268
count interrupts ................................................................... 379
create a task......................................................................... 284
create an 8-bit event flag..................................................... 276
define a null function .......................................................... 429
destroy a task....................................................................... 288
detect a timeout ........................................................... 371, 373
directly read the system timer ............................................. 298
directly write the system timer............................................ 332
dispatch most eligible task .................................................. 322
display Salvo status............................................................. 320
generate a single pulse ........................................................ 336
get current task's taskID ...................................................... 300
get current task's timestamp................................................ 300
get system ticks ................................................................... 298
initialize a ring buffer.......................................................... 282
initialize an LCD controller without delay loops........ 244, 246
initialize Salvo .................................................................... 302
manage access to a shared resource .................................... 338
measure run-time context switching performance.............. 383
obtain a message from within an ISR ................................. 358
obtain the current task's priority.................. 290, 292, 294, 296
pass a keypress in a message .............................................. 278
pass raw data using messages ............................................. 230
phase-shift a task................................................................. 352
preserve a task's timestamp................................................. 334

Salvo User Manual

Index

467

print the version number ..................................................... 375
process a buffer only when it is non-empty ........................ 266
protect a service called from foreground and background.. 370
protect Salvo variables against power-on reset........... 203, 213
read a binary semaphore's value ......................................... 308
read a semaphore's value..................................................... 316
repeatedly invoke a function with a cyclic timer ................ 274
replace one task with another using only one taskID ......... 250
reset a binary semaphore by reading it ............................... 356
restart a cyclic timer............................................................ 318
reuse a taskID...................................................................... 248
rotate a message queue's contents ....................................... 360
run a task for a one-time event.................................... 256, 258
run a task only once ............................................................ 254
run an idling function alongside Salvo ............................... 364
run incompatible code alongside Salvo .............................. 364
run OSTimer() from an interrupt ........................................ 354
set a task's timestamp when it starts.................................... 334
set system ticks ................................................................... 332
share a tcb between a cyclic timer and a task ..................... 286
start a task ........................................................................... 346
start and stop a cyclic timer ................................................ 344
stop a cyclic timer ....................................................... 348, 368
stop a task............................................................................ 350
test a message in a message queue...................................... 313
toggle a port bit when idling ............................................... 381
use the persistent type qualifier........................................... 156
vary a task's priority based on global variable .................... 328
wait for a keypress in a message......................................... 262
wake another task................................................................ 342
wake two tasks simultaneously........................................... 326
of
different task structures......................................................... 21
multiple delays in a task.......................................................... 4
non-reentrant function behavior............................................ 15
specifying register bank 0 in Hi-Tech PICC............... 154, 156
using #define to improve legibility..................... 70, 74, 78, 88

F
foreground / background systems ..................................... 11, 14–15
freeware version of Salvo ... xxvi, xxvii, 60, 99, 191, 192, 197, 210,
395, 409

468

Index

Salvo User Manual

H
Harbison, Samuel P..................................................................... 449

I
idle task ............................................................................... 187, 215
priority..................................................................................... 218
idling ............................................................................................. 14
installation
avoiding long pathnames .......................................................... 54
directories
demos .......................................................................... 207, 320
include files........... 85, 390, 398, 399, 406, 407, 408, 428, 442
libraries ............................................... 398, 399, 405, 408, 409
source files ................ 85, 94, 96, 390, 391, 404, 409, 410, 439
test programs............................................................... 207, 440
tutorials ..................................... 63, 78, 83, 86, 89, 90, 91, 207
license agreement...................................................................... 52
multiple distributions ................................................................ 60
non-Wintel platforms ................................................................ 57
on a network.............................................................................. 56
uninstalling............................................................. See uninstaller
interrupt service routine (ISR) ................................................ 12, 14
calling Salvo services from..................................................... 231
compiler-generated context saving ......................................... 209
OSTimer() ......................................................... 76, 208, 212, 353
priorities .................................................................................. 221
requirements.............................................................................. 17
response times........................................................................... 20
restrictions on calling Salvo services ...................................... 216
salvocfg.h ................................................................................ 226
stack depth .............................................................................. 201
static variables......................................................................... 217
use in
foreground / background systems ......................................... 15
intertask communications ..................................................... 13
interrupt_level pragma (HI-TECH PICC compiler) ........... 120, 440
interrupts .....12, 14–15, 230–32. See interrupt service routine (ISR)
avoiding problems with reentrancy........................................... 16
calling Salvo services from..................................................... 227
debugging........................................................................ 431, 432
effect on performance ............................................................. 199
in cooperative multitasking................................................. 20–21
in preemptive multitasking ................................................. 18–20
interrupt level #pragma ........................................................... 440
latency............................................................................... 18, 210
Salvo User Manual

Index

469

periodic ....................................................................... 25, 76, 209
polling ..................................................................................... 197
recovery time ............................................................................ 20
response time ............................................................................ 20
Salvo configuration options .................................................... 186
using OSTimer() without ........................................................ 213
intertask communication............................................................... 13

K
Kalinsky, David .......................................................................... 451
kernel....................................................................................... 13, 16
Kernighan, Brian W. ................................................................... 449

L
Labrosse, Jean J. ......................................................................... 450
LaVerne, David........................................................................... 451
libraries
configurations ......................................................................... 400
custom ............... 94, 128, 206, 231, 403, 405, 406, 407, 408, 409
salvoclcN.h configuration file............................... 94, 406, 409
global variables ....................................................................... 399
memory models....................................................................... 399
options..................................................................................... 399
overriding default RAM settings ............................................ 397
rebuilding ................................................................................ 403
bash shell and GNU make................................................... 404
specifying the compiler version .......................................... 405
types ................................................................................ 395, 399
using........................................................................................ 396
variants.................................................................................... 401
Linux / Unix......................................... xxvi, 59, 404, 433, 453, 454
Cygwin Unix environment for Windows........ 405, 406, 433, 454
MinGW Unix environment for Windows ............................... 405

M
make utility ................................................................................... 83
message queues....................................................................... 13, 37
messages ................................................................................. 13, 35
receiving.................................................................................... 36
signaling.................................................................................... 36
use in place of binary semaphores ............................................ 37
MicroC/OS-II.............................................................................. 450

470

Index

Salvo User Manual

multitasking............................................................................. 16, 21
event-driven .............................................................................. 28
mutexes ....................................................................................... 451
mutual exclusion ........................................................................... 16

O
operating system (OS)................................................................... 14

P
persistent type qualifier............................................................... 203
pointer ........................................................................................... 35
declaring multiple ................................................................... 388
dereferencing............................................................................. 35
null ............................................................................................ 36
runtime bounds checking ........................................................ 133
predefined constants...................................... 66, 127, 172, 189, 219
OSCALL_OSCREATEEVENT
OSFROM_ANYWHERE ................... 118, 119, 120, 189, 369
OSFROM_BACKGROUND...................................... 118, 119
OSFROM_FOREGROUND............................... 118, 119, 189
OSCALL_OSXYZ
OSFROM_ANYWHERE ........................................... 118, 189
OSFROM_BACKGROUND.............................................. 118
OSFROM_FOREGROUND............................................... 118
OSCOMPILER
OSAQ_430.......................................................................... 189
OSHT_8051C ..................................................................... 189
OSHT_PICC ....................................................................... 189
OSHT_V8C......................................................................... 189
OSIAR_ICC........................................................................ 189
OSKEIL_C51...................................................................... 189
OSMIX_PC......................................................................... 189
OSMPLAB_C18 ................................. 155, 165, 186, 188, 189
OSMW_CW........................................................................ 189
OSCTXSW_METHOD
OSRTNADDR_IS_PARAM ...................................... 127, 189
OSRTNADDR_IS_VAR .................................... 127, 172, 189
OSLOGGING
OSLOG_ALL ............................................................. 162, 188
OSLOG_ERRORS...................................................... 162, 188
OSLOG_NONE .......................................................... 162, 188
OSLOG_WARNINGS................................................ 162, 188
OSStartCycTmr()
OSDONT_START_CYCTMR........................................... 274

Salvo User Manual

Index

471

OSStartTask()
OSDONT_START_CYCTMR................................... 274, 343
OSDONT_START_TASK ................... 66, 219, 249, 283, 346
OSTARGET
OSMSP430 ......................................................................... 189
OSPIC12 ............................................................................. 189
OSPIC16 ..................................................................... 189, 451
OSPIC17 ............................................................................. 189
OSPIC18 ............................................................. 167, 168, 189
OSX86................................................................................. 189
OSVERSION .......................................................................... 374
preemption .................................................................................... 12
printf() ................................................................... 15, 162, 319, 432
program counter ...................................................................... 16, 17

R
RAM
reducing freeware library requirements .................................. 205
real-time operating system (RTOS) .............................................. 14
reentrancy...................................................................................... 15
resources
managing via semaphores ......................................................... 33
Ritchie, Dennis M. ...................................................................... 449
round-robin ........................................................................... 22, 218
rules
#2
where context switches may occur ..................................... 236
#3
persistent local variables ..................................................... 237

S
salvo.h .... 3, 63, 64, 65, 68, 70, 74, 76, 78, 84, 85, 94, 96, 100, 109,
207, 243, 245, 247, 249, 251, 253, 255, 257, 261, 263, 265, 267,
363, 365, 371, 372, 374, 376, 378, 390, 428
including ................................................................................... 84
locating...................................................................................... 85
salvocfg.h xxviii, 84, 85, 86, 87, 88, 89, 90, 91, 94, 96, 98, 99, 100,
103, 104, 105, 106, 107, 109, 111, 132, 134, 135, 136, 137, 141,
142, 146, 150, 154, 156, 189, 194, 196, 201, 205, 206, 207, 208,
212, 220, 226, 275, 280, 396, 397, 398, 399, 400, 401, 407, 408,
428, 429, 430, 432, 438, 460
default ..................................................................................... 206
default values ............................................................................ 89
including ................................................................................... 84

472

Index

Salvo User Manual

leaving a configuration option undefined ................................. 88
locating...................................................................................... 85
specifying the number of events ............................................... 88
specifying the number of tasks ................................................. 87
scheduling ......................................................................... 13, 16, 24
semaphores.............................................................................. 13, 29
shared resources ............................................................................ 16
stack ........................................................................................ 12, 19
overcoming limitations ........................................................... 233
role in reentrancy ...................................................................... 16
saving context ........................................................................... 17
Steele, Guy L., Jr......................................................................... 449
superloop................... 11, 14. See foreground / background systems
synchronization
conjunctive............................................................ See event flags
disjunctive ............................................................. See event flags
system response ............................................................................ 15
system timer ....................................................................... See timer

T
task ................................................................................................ 12
association with events ............................................................. 29
behavior
due to context switch ............................................................ 17
during interrupts.............................................................. 17–18
in cooperative multitasking............................................. 20–21
in preemptive multitasking ............................................. 18–20
context................................................................................. 12, 17
delay.............................................................................. 13, 24–26
in-line loop ............................................................................ 25
maximum .............................................................................. 25
using timer ............................................................................ 26
preemption ................................................................................ 12
priority....................................................................................... 12
dynamic................................................................................. 22
importance thereof .............................................................. 198
static ...................................................................................... 22
priority-based execution............................................................ 22
relationship to events ................................................................ 13
round-robin execution............................................................... 22
running ...................................................................................... 13
state ............................................................................... 13, 23–24
transition ............................................................................... 23
structure............................................................................... 21–22
suspending and resuming.......................................................... 12

Salvo User Manual

Index

473

switch ...............................................................See context switch
synchronization ......................................................................... 31
timeouts......................................................................................... 13
breaking a deadlock with .......................................................... 38
timer .............................................................................................. 13
accuracy .................................................................................... 26
resolution................................................................................... 26
system tick ................................................................................ 25
system tick rate ......................................................................... 25
using OSTimer() without interrupts........................................ 213
tools
HI-TECH Software
HPDPIC integrated development environment . 434, 436, 437,
440
mouse problems .............................................................. 434
running in DOS window ................................................. 434
running under Windows 2000......................................... 434
HPDV8 integrated development environment.................... 440
PICC compiler ..... 89, 118, 119, 120, 153, 154, 155, 156, 164,
173, 178, 189, 200, 203, 233, 319, 396, 432, 434, 435, 436,
437, 438, 439, 440, 441, 448, 453
PICC-18 compiler ....................................... 119, 153, 155, 453
IAR Systems
C-SPY Debugger ................................................................ 453
in-circuit debugger (ICD) ....................................................... 431
in-circuit emulator (ICE)......................................................... 431
Keil
Cx51 Compiler.................................................... 154, 155, 178
make utility ............................................................... 83, 404, 433
makefile........................................................... 404, 409, 410, 442
Makefile .................................................... 98, 404, 405, 408, 409
Metrowerks
CodeWarrior C compiler............. 100, 429, 444, 445, 453, 454
Microchip
MPLAB integrated development environment...... 86, 91, 155,
165, 167, 388, 431, 435, 445, 460
MPLAB-C18 C compiler.................................... 155, 165, 167
MPLAB-ICD in-circuit debugger ....................................... 431
MPLAB-ICE in-circuit emulator ........................................ 431
PICMASTER in-circuit emulator ....................................... 431
Microchip, Inc.
MPLAB-C18 compiler................................................ 155, 165
Mix Software
Power C compiler 100, 217, 442, 443, 444, 449, 450, 454, 455
Power C debugger............................................................... 444
Quadravox

474

Index

Salvo User Manual

AQ430 Development Tools ................................ 407, 408, 454
tutorial ....................... 63, 78, 83, 86, 89, 90, 91, 196, 206, 450, 461
typecasting ............................................................ 80, 226, 446, 447
types
predefined ................................See variables:Salvo defined types

U
uninstaller...................................................................................... 58
user macros
OSECBP()...... 70, 74, 78, 88, 174, 202, 256, 258, 270, 271, 276,
278, 280, 282, 304, 306, 313, 326, 336, 338, 376, 377
OSEFCBP()............................................................. 275, 276, 376
OSMQCBP()........................................................... 279, 280, 376
OSTCBP().. 4, 65, 66, 68, 70, 74, 78, 87, 88, 149, 202, 215, 218,
219, 220, 246, 250, 252, 274, 284, 286, 292, 296, 318, 322,
324, 344, 346, 348, 366, 368, 376, 377, 398
user services
events
OS_WaitBinSem()71, 72, 73, 74, 75, 132, 223, 239, 255, 256,
271, 307, 318, 335, 336, 350, 352, 355, 356, 372, 373
OS_WaitEFlag() 135, 189, 257, 258, 259, 260, 269, 270, 276,
309, 310, 325, 326
OS_WaitMsg()78, 80, 141, 179, 212, 222, 223, 226, 228, 244,
261, 262, 264, 277, 278, 311, 338, 357, 372, 373, 447
OS_WaitMsgQ() 108, 142, 263, 264, 280, 303, 305, 306, 313,
339, 359, 372, 373
OS_WaitSem()... 101, 146, 202, 211, 225, 265, 266, 281, 315,
341, 361, 372, 373
OSClrEFlag() ..... 122, 135, 258, 259, 260, 269, 270, 309, 326,
392
OSCreateBinSem() 71, 75, 119, 120, 126, 132, 256, 271, 272,
307, 308, 335, 336, 355, 392
OSCreateEFlag() 102, 135, 258, 259, 269, 275, 276, 309, 326,
392
OSCreateMsg() ...... 79, 80, 141, 226, 262, 271, 277, 278, 311,
338, 357, 373, 392
OSCreateMsgQ() 108, 126, 142, 264, 279, 280, 303, 306, 313,
339, 359, 392
OSCreateSem() .. 146, 202, 225, 229, 265, 281, 282, 315, 316,
341, 342, 350, 361, 362, 377, 392
OSMsgQCount() ......................................................... 121, 303
OSMsgQEmpty() ................................ 121, 305, 306, 392, 402
OSReadBinSem() 136, 256, 271, 307, 308, 335, 355, 392, 402
OSReadEFlag() .. 122, 136, 258, 269, 276, 309, 310, 326, 392,
402

Salvo User Manual

Index

475

OSReadMsg() ..... 136, 262, 277, 311, 312, 338, 357, 392, 402
OSReadMsgQ().. 136, 264, 280, 303, 306, 313, 314, 339, 359,
392, 402
OSReadSem() ..... 136, 265, 281, 315, 316, 341, 361, 392, 402
OSSetEFlag() ...................... 122, 135, 258, 259, 260, 325, 326
OSSignalBinSem()...... xxvi, 71, 72, 73, 74, 75, 132, 138, 232,
255, 256, 271, 307, 335, 336, 349, 355, 369, 370, 393, 402
OSSignalMsg() . xxvi, 78, 79, 80, 81, 122, 126, 141, 179, 201,
222, 226, 227, 228, 231, 262, 277, 278, 311, 337, 338, 339,
357, 385, 393, 402, 446, 447
OSSignalMsgQ()....... xxvi, 142, 264, 280, 303, 305, 306, 313,
339, 340, 359, 360, 393, 402
OSSignalSem() . xxvi, 146, 179, 203, 225, 231, 265, 281, 315,
341, 342, 361, 393, 400, 402
OSTryBinSem() .................. 137, 256, 271, 307, 335, 355, 356
OSTryMsg()........................ 137, 262, 277, 311, 338, 357, 358
OSTryMsgQ() ..... 137, 264, 280, 303, 306, 313, 339, 359, 360
OSTrySem()........ 122, 137, 265, 281, 315, 341, 361, 362, 371
general
OSInit() 4, 63, 64, 65, 68, 71, 75, 79, 123, 149, 180, 181, 203,
220, 225, 245, 246, 250, 284, 297, 301, 302, 321, 322, 331,
345, 392, 430, 431
OSSched() .... 4, 64, 65, 68, 69, 71, 75, 79, 125, 139, 140, 143,
144, 145, 147, 149, 157, 173, 180, 215, 218, 220, 226, 232,
246, 250, 284, 302, 321, 322, 346, 360, 363, 364, 366, 380,
382, 392, 430, 431
hooks
OSDisableIntsHook().................................................. 378, 379
OSEnableIntsHook()................................................... 378, 379
OSIdlingHook() .................................. 140, 215, 380, 381, 429
monitor
OSRpt()147, 161, 169, 170, 171, 215, 220, 319, 320, 392, 427
other
OSCreateCycTmr() .... 273, 274, 285, 286, 317, 323, 343, 347,
367
OSCycTmrRunning().. 274, 285, 317, 323, 343, 347, 367, 368
OSDestroyCycTmr()... 274, 285, 286, 317, 323, 343, 347, 367
OSProtect() ................................................. 119, 120, 369, 370
OSResetCycTmr()....... 274, 285, 317, 318, 323, 343, 347, 367
OSSetCycTmrPeriod()................ 274, 285, 323, 324, 343, 347
OSStartCycTmr()........ 274, 285, 317, 323, 343, 344, 347, 367
OSStopCycTmr() 274, 285, 317, 323, 343, 344, 347, 348, 367
OSTimedOut() ............ 151, 211, 224, 225, 363, 371, 372, 373
OSUnprotect()............................................. 119, 120, 369, 370
OSVersion() ................................................................ 374, 375
tasks

476

Index

Salvo User Manual

OS_Delay() ...... 4, 26, 75, 77, 79, 90, 115, 210, 216, 218, 222,
235, 237, 239, 240, 241, 243, 244, 245, 246, 247, 248, 250,
253, 260, 278, 286, 288, 300, 308, 312, 314, 328, 330, 333,
334, 338, 342, 344, 400, 429, 430, 437, 443, 444, 445
OS_DelayTS()..... 243, 245, 246, 299, 300, 333, 334, 351, 352
OS_Destroy() ...................................................... 247, 248, 287
OS_Prio() .................................................... 219, 220, 327, 329
OS_Replace().............................................................. 249, 250
OS_SetPrio() ................... 79, 80, 251, 252, 289, 290, 291, 327
OS_Stop() ................................... 243, 246, 253, 254, 256, 349
OS_Yield(). 4, 65, 66, 67, 68, 69, 71, 72, 75, 77, 78, 127, 149,
172, 213, 214, 216, 217, 237, 251, 252, 267, 268, 284, 334,
346, 356, 366, 430, 445
OSCreateTask().... 4, 65, 66, 67, 68, 71, 75, 79, 110, 149, 199,
216, 218, 219, 220, 226, 246, 247, 248, 249, 250, 251, 252,
254, 268, 283, 284, 286, 287, 288, 321, 322, 330, 342, 345,
346, 350, 366, 377, 392, 398, 430, 431
OSDestroyTask() ........................................ 249, 287, 288, 392
OSGetPrio() ........................ 121, 251, 289, 290, 291, 327, 392
OSGetPrioTask()......... 121, 289, 291, 292, 327, 329, 392, 402
OSGetState() ....................................... 121, 293, 294, 295, 392
OSGetStateTask() ....................... 121, 293, 295, 296, 392, 402
OSGetTS() .......................... 246, 299, 300, 333, 334, 351, 392
OSSetEFlag() ..... 122, 135, 258, 259, 260, 269, 309, 325, 326,
392
OSSetPrio() .... 79, 80, 219, 220, 251, 289, 291, 327, 328, 329,
392
OSSetPrioTask() ......................... 289, 291, 327, 329, 330, 393
OSStartTask() ...... 66, 110, 122, 199, 219, 253, 283, 284, 321,
345, 346, 349, 393, 402, 427
OSStopTask()...................................... 253, 284, 349, 350, 393
OSSyncTS() ................................ 246, 299, 333, 351, 352, 393
timer
OSGetTicks() ...... 117, 125, 211, 212, 297, 298, 331, 334, 392
OSSetTicks()............... 117, 125, 211, 212, 297, 331, 332, 393
OSSetTS() ................................... 246, 299, 333, 334, 351, 393
OSTimer() ...... 76, 77, 115, 117, 125, 159, 176, 182, 199, 208,
209, 210, 212, 213, 219, 228, 231, 232, 243, 246, 353, 354,
393, 438, 439, 441

V
va_arg() ....................................................................................... 126
variables
declaring.................................................................................. 385
errors when dereferencing....................................................... 227

Salvo User Manual

Index

477

global, shared ............................................................................ 68
initializing globals to zero....................................................... 123
local..................................................................................... 16, 19
locating in memory ..................................................... 89, 154–60
Salvo defined types ................................................................. 384
static ................................................................................ 164, 217

W
Wagner, Thomas ......................................................................... 450
watchdog timer............................................................................ 431

Y
Y2K compliance ......................................................................... 195

478

Index

Salvo User Manual

Notes

Salvo User Manual

479

480

Notes

Salvo User Manual

Salvo User Manual

Notes

481

482

Notes

Salvo User Manual

Salvo User Manual

Notes

483



---

Source Exif Data:

File Type                       : PDF
File Type Extension             : pdf
MIME Type                       : application/pdf
PDF Version                     : 1.2
Linearized                      : Yes
Encryption                      : Standard V1.2 (40-bit)
User Access                     : Print, Copy, Annotate, Fill forms, Extract, Assemble, Print high-res
Producer                        : Acrobat Distiller 4.05 for Windows
Creator                         : Microsoft Word 9.0
Keywords                        : Salvo, Pumpkin, RTOS, User, Manual, Installation, Reference, Tutorial, Fundamentals, Configuration, FAQ, Troubleshooting
Modify Date                     : 2010:06:03 16:38:11-07:00
Subject                         : Salvo The RTOS that runs in tiny places
Create Date                     : 2010:06:03 16:35:05
Title                           : Salvo User Manual
Author                          : Andrew E. Kalman
Page Count                      : 529
Page Mode                       : UseNone

EXIF Metadata provided by EXIF.tools