TCE Usr Manual
User Manual:
Open the PDF directly: View PDF .
Page Count: 145 [warning: Documents this large are best viewed by clicking the View PDF Link!]
![](asset-1.png)
Customized Parallel Computing group
Department of Pervasive Computing
Tampere University of Technology
TTA-based Co-design Environment v1.14
User Manual
Version: 31
Created: 2006-10-25
Last Modified: 2016-02-23
Document: P-1000
State: complete
TTA-based Co-design Environment v1.14 User Manual
Contents
1 INTRODUCTION 7
1.1 DocumentOverview .................................... 8
1.2 Acronyms, Abbreviations and Definitions . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.3 Typographic Conventions Used in the Document . . . . . . . . . . . . . . . . . . . . . 8
2 PROCESSOR DESIGN FLOW 9
2.1 DesignFlowOverview ................................... 9
2.2 MainFileFormats...................................... 11
2.2.1 Architecture Definition File (ADF) . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2.2 TTA Program Exchange Format (TPEF) . . . . . . . . . . . . . . . . . . . . . . 11
2.2.3 Hardware Database (HDB) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.4 Implementation Definition File (IDF) . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.5 BinaryEncodingMap ............................... 12
2.2.6 Operation Set Abstraction Layer (OSAL) Files . . . . . . . . . . . . . . . . . . 12
2.2.7 Simulation Trace Database . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2.8 Exploration Result Database . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13
2.2.9 Summaryoffiletypes ............................... 13
3 TUTORIALS AND HOW-TOS 14
3.1 TCETour .......................................... 14
3.1.1 TheSampleApplication .............................. 14
3.1.2 Starting Point Processor Architecture . . . . . . . . . . . . . . . . . . . . . . . 15
3.1.3 Compiling and simulating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
3.1.4 Increasing performance by adding resources . . . . . . . . . . . . . . . . . . . . 18
3.1.5 Analyzing the Potential Custom Operations . . . . . . . . . . . . . . . . . . . . 19
3.1.6 Creating the Custom Operation . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.1.7 Use the custom operation in C code. . . . . . . . . . . . . . . . . . . . . . . . . 24
3.1.8 Adding HDL implementation of the FU to the hardware database (HDB). . . . . 26
3.1.9 Generating the VHDL and memory images . . . . . . . . . . . . . . . . . . . . 28
3.1.10 Further acceleration by adding custom operation to large TTA . . . . . . . . . . 30
3.1.11 FinalWords..................................... 31
3.2 From C to VHDL as Quickly as Possible . . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.3 HelloTTAWorld!...................................... 31
3.4 StreamingI/O........................................ 32
3.4.1 Streaming I/O function units . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.5 Implementing Programs in Parallel Assembly Code . . . . . . . . . . . . . . . . . . . . 34
2/145 2016-02-23 version 31
主要文件类型
开发流程图
一个CRC检验的开发例子
开发时的一些
方法。
TTA-BASED CO-DESIGN ENVIRONMENT
3.5.1 Preparations..................................... 34
3.5.2 IntroductiontoDCT ................................ 34
3.5.3 Introduction to TCE assembly . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.5.4 Implementing DCT on TCE assembly . . . . . . . . . . . . . . . . . . . . . . . 35
3.6 Using multiple memories from C code . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.7 RunningTTAonFPGA................................... 37
3.7.1 Simplest example: No data memory . . . . . . . . . . . . . . . . . . . . . . . . 37
3.7.2 Second example: Adding data memory . . . . . . . . . . . . . . . . . . . . . . 39
3.8 How to print from Altera FPGAs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.8.1 HelloWorld2.0................................... 41
3.8.2 FPGAexecution .................................. 43
3.8.3 Caveats in printing from FPGA . . . . . . . . . . . . . . . . . . . . . . . . . . 45
3.8.4 Summary...................................... 45
3.9 Designing Floating-point Processors with TCE . . . . . . . . . . . . . . . . . . . . . . 45
3.9.1 Restrictions..................................... 46
3.9.2 Single-precision Function Units . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.9.3 Half-precision Support . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.9.4 Half-precision Function Units . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
3.9.5 Benchmarkresults ................................. 48
3.9.6 Alternativebitwidths................................ 48
3.9.7 Processor Simulator and Floating Point Operations . . . . . . . . . . . . . . . . 48
3.10Multi-TTADesigns..................................... 49
3.11OpenCLSupport ...................................... 49
3.12 System-on-a-Chip design with AlmaIF Integrator . . . . . . . . . . . . . . . . . . . . . 50
3.12.1 Generate the Processor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.12.2 Create a New Vivado Project . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.12.3 CreateaBlockDesign ............................... 50
3.12.4 Cleanup....................................... 50
3.12.5 Synthesis and Implementation . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4 PROCESSOR DESIGN TOOLS 52
4.1 TTA Processor Designer (ProDe) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.1.1 StartingProDe ................................... 52
4.1.2 Function Unit Operation Dialog . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.2 Operation Set Abstraction Layer (OSAL) Tools . . . . . . . . . . . . . . . . . . . . . . 53
4.2.1 Operation Set Editor (OSEd) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.2.2 Operation Behavior Module Builder (buildopset) . . . . . . . . . . . . . . . . . 56
4.2.3 OSAL Tester (testosal) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
4.3 OSALfiles ......................................... 57
4.3.1 OperationProperties ................................ 58
4.3.2 Operation Input Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.3.3 Operation Output Properties . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
4.3.4 OperationDAG................................... 60
4.3.5 OperationBehavior................................. 60
4.3.6 Behavior Description language . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.4 OSALsearchpaths ..................................... 64
version 31 2016-02-23 3/145
开发工具介绍
TTA-based Co-design Environment v1.14 User Manual
4.5 Processor Generator (ProGe) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
4.5.1 IC/DecoderGenerators............................... 65
4.6 PlatformIntegrator ..................................... 66
4.7 SupportedPlatforms .................................... 67
4.7.1 Alterabaseddevices ................................ 67
4.7.2 AlmaIFIntegrator.................................. 70
4.8 Hardware Database Editor (HDB Editor) . . . . . . . . . . . . . . . . . . . . . . . . . . 71
4.8.1 Usage........................................ 71
4.9 HardwareDatabaseTester ................................. 72
4.9.1 Usage........................................ 73
4.9.2 Testconditions ................................... 73
4.10Processorunittester..................................... 74
4.10.1 Usage........................................ 74
4.11FunctionUnitInterface................................... 74
4.11.1 Operationcodeorder................................ 75
4.11.2 Summary of interface ports . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75
4.11.3 Reserved keywords in generics . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
5 CODE GENERATION TOOLS 77
5.1 TCECompiler........................................ 77
5.1.1 UsageofTCEcompiler .............................. 77
5.1.2 Customoperations ................................. 79
5.1.3 Knownissues.................................... 79
5.2 Binary Encoding Map Generator (BEMGenerator) . . . . . . . . . . . . . . . . . . . . 79
5.2.1 Usage........................................ 80
5.3 Parallel Assembler and Disassembler . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80
5.3.1 UsageofDisassembler............................... 80
5.3.2 UsageofAssembler ................................ 81
5.3.3 MemoryAreas ................................... 81
5.3.4 GeneralLineFormat ................................ 82
5.3.5 Allowedcharacters................................. 82
5.3.6 Literals ....................................... 82
5.3.7 Labels........................................ 83
5.3.8 DataLine...................................... 84
5.3.9 CodeLine...................................... 85
5.3.10 LongImmediateChunk .............................. 86
5.3.11 DataTransport ................................... 86
5.3.12 RegisterPortSpecifier ............................... 86
5.3.13 Assembler Command Directives . . . . . . . . . . . . . . . . . . . . . . . . . . 88
5.3.14 AssemblyFormatStyle............................... 89
5.3.15 ErrorConditions .................................. 91
5.3.16 WarningConditions ................................ 92
5.3.17 DisambiguationRules ............................... 93
5.4 Program Image Generator (PIG) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93
5.4.1 Usage........................................ 94
5.4.2 DictionaryCompressor............................... 94
4/145 2016-02-23 version 31
编译器
TTA-BASED CO-DESIGN ENVIRONMENT
5.5 TPEFDumper(dumptpef) ................................. 96
5.5.1 Usage........................................ 96
6 CO-DESIGN TOOLS 98
6.1 Architecture Simulation and Debugging . . . . . . . . . . . . . . . . . . . . . . . . . . 98
6.1.1 Processor Simulator CLI (ttasim) . . . . . . . . . . . . . . . . . . . . . . . . . 98
6.1.2 Fast Compiled Simulation Engine . . . . . . . . . . . . . . . . . . . . . . . . . 99
6.1.3 RemoteDebugger ................................. 100
6.1.4 Simulator Control Language . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101
6.1.5 Traces........................................ 106
6.1.6 Processor Simulator GUI (Proxim) . . . . . . . . . . . . . . . . . . . . . . . . . 107
6.2 System Level Simulation with SystemC . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6.2.1 Instantiating TTACores . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108
6.2.2 Describing Detailed Operation Pipeline Simulation Models . . . . . . . . . . . . 109
6.3 Processor Cost/Performance Estimator (estimate) . . . . . . . . . . . . . . . . . . . . . 110
6.3.1 Command Line Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
6.4 Automatic Design Space Explorer (explore) . . . . . . . . . . . . . . . . . . . . . . . . 111
6.4.1 Explorer Application format . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
6.4.2 Command Line Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111
6.4.3 Explorer Plugin: ConnectionSweeper . . . . . . . . . . . . . . . . . . . . . . . 112
6.4.4 Explorer Plugin: SimpleICOptimizer . . . . . . . . . . . . . . . . . . . . . . . 112
6.4.5 Explorer Plugin: RemoveUnconnectedComponents . . . . . . . . . . . . . . . . 113
6.4.6 Explorer Plugin: GrowMachine . . . . . . . . . . . . . . . . . . . . . . . . . . 114
6.4.7 Explorer Plugin: ImmediateGenerator . . . . . . . . . . . . . . . . . . . . . . . 114
6.4.8 Explorer Plugin: ImplementationSelector . . . . . . . . . . . . . . . . . . . . . 114
6.4.9 Explorer Plugin: MinimizeMachine . . . . . . . . . . . . . . . . . . . . . . . . 115
6.4.10 Explorer Plugin: ADFCombiner . . . . . . . . . . . . . . . . . . . . . . . . . . 115
7 PROCESSOR TEMPLATE 116
7.1 ArchitectureTemplate ................................... 116
7.1.1 Transport Triggered Architecture . . . . . . . . . . . . . . . . . . . . . . . . . 116
7.1.2 Immediates/Constants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118
7.1.3 Operations, Function Units, and Operand Bindings . . . . . . . . . . . . . . . . 118
7.1.4 Datapath Connectivity Levels . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
7.2 ProgrammerInterface.................................... 119
7.2.1 Default Data Address Space Layout . . . . . . . . . . . . . . . . . . . . . . . . 120
7.2.2 Instruction Address Space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120
7.2.3 Alignment of Words in Memory . . . . . . . . . . . . . . . . . . . . . . . . . . 120
7.2.4 StackFrameLayout ................................ 120
7.2.5 WordByteOrder.................................. 121
7.2.6 Function Calling Conventions . . . . . . . . . . . . . . . . . . . . . . . . . . . 122
7.2.7 Register Context Saving Conventions . . . . . . . . . . . . . . . . . . . . . . . 122
8 PRODUCING EFFICIENT TTA DESIGNS WITH TCE 123
8.1 Registers and Register Files . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123
8.2 InterconnectionNetwork .................................. 123
version 31 2016-02-23 5/145
协同工具
处理器模板
TTA-based Co-design Environment v1.14 User Manual
8.2.1 Negative short immediates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
8.3 OperationSet ........................................ 124
9 TROUBLESHOOTING 125
9.1 Simulation.......................................... 125
9.1.1 Failing to Load Operation Behavior Definitions . . . . . . . . . . . . . . . . . . 125
9.2 Limitations of the Current Toolset Version . . . . . . . . . . . . . . . . . . . . . . . . . 125
9.2.1 IntegerWidth.................................... 125
9.2.2 Instruction Addressing During Simulation . . . . . . . . . . . . . . . . . . . . . 125
9.2.3 Data Memory Addressing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
9.2.4 Ideal Memory Model in Simulation . . . . . . . . . . . . . . . . . . . . . . . . 126
9.2.5 Guards ....................................... 126
9.2.6 Operation Pipeline Description Limitations . . . . . . . . . . . . . . . . . . . . 126
9.2.7 EncodingofXMLFiles .............................. 126
9.2.8 FloatingPointSupport ............................... 126
A FREQUENTLY ASKED QUESTIONS 127
A.1 MemoryRelated ...................................... 127
A.1.1 LoadStoreUnit................................... 127
A.2 ProcessorGenerator..................................... 127
A.3 tcecc............................................. 127
A.4 Hardwarecharacteristics .................................. 128
A.4.1 Interruptsupport .................................. 128
A.5 Misc............................................. 128
A.5.1 FileSearchPaths.................................. 128
B SystemC Simulation Example 129
C Copyright notices 134
C.1 Xerces............................................ 134
C.2 wxWidgets ......................................... 136
C.3 L-GPL............................................ 137
C.4 TCL............................................. 142
C.5 SQLite............................................ 143
C.6 Editline ........................................... 143
BIBLIOGRAPHY 145
6/145 2016-02-23 version 31
故障排除,一些
常见问题
常问问题。
![](asset-7.png)
TTA-BASED CO-DESIGN ENVIRONMENT
Chapter 1
INTRODUCTION
This is the user manual for TTA Co-design Environment (TCE) [TCE] [lpt] The document describes all
the tools in the toolset, and a set of common use cases in the form of tutorials.
Fig.1.1 shows a simplified overview of TCE. TCE supports programs written in C/C++ and OpenCL.
User can easily design new customized Transport-Triggered Architecture (TTA) processors, compile the
program, analyze the performance, and generate HDL implementations of the designed processors. The
generated application-specific processor HDL can be synthesized for example to an FPGA chip. Then,
user can modify the application and upload new program image to continue development.
Application-specific instruction-set processor (ASIP) is a programmable processor which is tailored to
certain application (domain). Hence, it can outperform general-purpose processors in terms of perfor-
mance, area, and power. On the other hand, the programmability and the powerful ASIP tools should
provide increased productivity in comparison to fixed function accelerator implementations, which offer
the highest performance.
With TCE, TTA ASIP configuration, SW compilation, and simulation can be carried out in the order of
minutes and hence most of the time can be reserved to application and architecture co-development. It
is fast to iterate over new TTA configurations, e.g., change the number and type of function units (FU),
configure interconnection network (IC), register files (RF), or experiment with special function units
(custom operations).
Figure 1.1: Overview of TTA Co-design Environment (TCE).
version 31 2016-02-23 7/145
![](asset-8.png)
TTA-based Co-design Environment v1.14 User Manual
Simple TTA designs with support for a couple of parallel arithmetic operations consume in the order of
approx. 1000-5000 LUTs (look-up tables), which means that modern FPGA’s can easily include even
tens of TTA cores. The operating clock frequency can be optimized to match the soft cores offered by
FPGA vendors.
1.1 Document Overview
Chapter 2 provides an overview to the TCE processor design flow, and Chapter 3 contains tutorials for
several common TCE use cases. These should be sufficient for starting to use the TCE toolset. The rest
of the chapters describe the use of each tool separately, and they can be referred to for information on
more advanced usage the tools.
Many people have contributed to this document, most notably Pekka Jääskeläinen, Otto Esko, Heikki
Kultala, Vladimir Guzma and Erno Salminen.
1.2 Acronyms, Abbreviations and Definitions
ADF (Processor/Machine) Architecture Definition File.
BEM Binary Encoding Map. Describes the encoding of instructions.
CLI Command Line Interface
ExpResDB Exploration Result Database.
GUI Graphical User Interface
GPR General Purpose Register
HDB Hardware Database
HDL Hardware Description Language.
HLL High Level (Programming) Language.
IDF (Machine/Processor) Implementation Definition File.
ILP Instruction Level Parallelism.
LLVM Low Level Virtual Machine
MAU Minimum Addressable Unit
PIG Program Image Generator
SQL Structured Query Language.
TCE TTA Co-design Environment.
TPEF TTA Program Exchange Format
TraceDB Execution Trace Database.
TTA Transport Triggered Architecture.
VHDL VHSIC Hardware Description Language.
XML Extensible Markup Language.
1.3 Typographic Conventions Used in the Document
Style Purpose
italic parameter names in running text
[brackets] bibliographic references
‘single quoted’ keywords or literal strings in running text
’file name’ file and directory names in running text
bold Shorthand name of a TCE application.
courier
UNIX commands
8/145 2016-02-23 version 31
![](asset-9.png)
TTA-BASED CO-DESIGN ENVIRONMENT
Chapter 2
PROCESSOR DESIGN FLOW
The main goal for the TTA Co-design Environment (TCE) is to provide a reliable and effective toolset
for designing programmable application specific processors, and generate machine code for them from
applications written in high-level languages, such as C, C++ or OpenCL.
In addition, TCE provides an extensible research platform for experimenting with new ideas for Transport
Triggered Architectures (TTAs), retargetable ILP code generation, and application specific processor
design methodology, among others.
2.1 Design Flow Overview
The main phases in the TCE are illustrated in Figure 2.1 and listed in Table 2.1 The TCE design flow
starts from an application described in a high level language (currently the C language). The initial
TCE
tools
3rd party tools
requirements
user
Processoor designer tool
(prode) HLL
(C, C++, OpenCL)
Retargerable compiler
(tcecc/LLVM)
Retargerable simulator
(proxim)
Processor generator
(proge) Feedback
(cycle count, mem usage,)
Program image generator
(pig, opt. compression)
Platform integrator
(?)
FPGA synthesis tool
(e.g Quartus) HDL simulation tool
(e.g QHDL)
HW database
(hdb)
Platform info
(FPGA-specific)
feedback
Report
(FPGA resources, MHz)
Figure 2.1: Overview of TTA Co-design Environment (TCE).
version 31 2016-02-23 9/145
![](asset-a.png)
TTA-based Co-design Environment v1.14 User Manual
Table 2.1: Summary of tools. Some of the tools have grahical user interface (GUI) and others executed
from the shell command line (CLI).
Tool Purpose Type Output file(s)
ProDe Define FUs, registers, intercconects of TTA core GUI .adf
tcecc Compile software CLI .tpef
Proxim Simulate TTA cores (architectural level) GUI report
ttasim Simulate TTA cores (architectural level) CLI report
OSEd Operation Set Abstraction Layer database editor GUI operation db
HDBEditor HW implementation database editor GUI HDB
ProGe / generateprocessor Generate HDL CLI (+GUI) VHDL or Verilog
PIG / generatebits Generate program image CLI e.g. .mif
Platform integrator Interface with memories, generate IP-XACT CLI e.g. .xml, project files
3rd party Synthesis Convert VHDL to netlist and program FPGA G/C e.g. .sof
3rd party Simulation Simulate VHDL G/C report, .wave
software development phase is intended to be separate from the actual co-design flow of TCE. That is,
the program is expected to be implemented and tested natively (on a workstation PC) before “porting” it
to the TTA/TCE platform. The porting includes ensuring that TTA/TCE runs the program correctly, and
optimizing the hardware together with the software by modifying the resources and architecture of the
processor to fit the application at hand – a process called hardware/software co-design.
Once the user has implemented the software based on requirements, he designs a TTA processor with
a graphical tool, or selects an existing TTA, and compiles and simulates the SW. Both compiler are
simulator are retargetable so they need an architectre definition as input in addition to SW source codes.
The instruction set simulator (Proxim or ttasim) produces statistics about cycle counts and utilization of
function unit, to assist in iteration of the SW of TTA HW until results are satiscatory.
The LLVM compiler framework [llv08] is used to compile the application to ’bitcode’, the intermedi-
ate representation of LLVM. The resulting bitcode is then compiled and scheduled to a particular TTA
processor by TCE. Traditional compiler optimizations are done in LLVM before bitcode generation, so
it is possible to utilize the same bitcode file when exploring different TTA processors for running an
application.
Once the performance requirement is met, the user maps the architectural resources i.e. function units and
register files, to their respective RTL implementations which are stored in Hardware Databases (HDB).
The distinction between architectural models and component implementations makes it possible to have
several implementations, each e.g. targeted for different FPGA families or ASIC technologies, stored in
HDBs. Ideally, changing the target platform might only require changes to the implementation mapping
without any changes to the architecture.
After the implementation mapping, the Processor Generator (ProGe) can produce synthesizable HDL,
which instantiates all FUS and creates the interconnection network and control unit, which is responsible
of fetching, decoding and issuing instructions. ProGe tool also includes a platform integrator feature
which can interface the TTA with memory components of an FPGA board and create synthesis settings.
Moreover, the platform integrator can “package” the TTA (and optionally also the SW) into an easily
re-usable format. In practice, this means generating IP-XACT documents which describe the details in
XML. Synthesing and simulating the HDL are done with 3rd party tools, such as Altera’s Quartus and
Mentor Graphics’ Modelsim.
Finally, Program Image Generator (PIG) is utilized to convert the compiled SW format into actual bina-
ries. PIG supports various different output formats ranging from plain binary format to FPGA vendor
specific embedded RAM formats.
More details about the tools and design phases are given in Chapters 4.1 - 6. The TCE tour tutorial in
section 3.1 provides a hands-on introduction to the design flow.
10/145 2016-02-23 version 31
1.保证能
正确运行
程序。
2.修改处
理器的资
源与结
构,与软
件一起优
化硬件
(软硬件
协同)。
SW:software
HW:hardware
TTA-BASED CO-DESIGN ENVIRONMENT
2.2 Main File Formats
The flow consists of many separate tools and information is exchanged with files. This chapter gives an
overview of 8 file and database types manipulated by TCE applications and accessible to users of the
toolset, architecture definition .adf and SW binary .tpef being the most important ones. Many files are
XML so user can easily access them, if needed.
There a 3 kind of files:
1. HW definitions: adf, hdb, and idf
2. SW definitions: tpef, bem, opp, cc, opb
3. Simulation and exploration data: dsdb, set of the files above
2.2.1 Architecture Definition File (ADF)
Filename extension: .adf
Architecture Definition File (ADF) is a file format for defining target processor architectures, e.g. which
function units are used and how they are connected [CSJ04]. ADF can be created with ProDe manually
or, e.g., by automated design space exploration tools. ADF is a specification of the target “processor
architecture”, meaning that only the information needed to generate valid programs for the processor is
stored, but nothing else. This file is needed by many tools, such as tcecc, ttasim, and ProGe.
The XML snippet below gives an idea of ADF.
<adf version="1.7">
<bus name="B1">
<width>32</width>
<guard>
<always-true/>
...
<socket name="alu_comp_i1">
<reads-from>
<bus>B1</bus>
...
<function-unit name="alu_comp">
<port name="in1t">
<connects-to>alu_comp_i1</connects-to>
<width>32</width>
<triggers/>
...
<operation>
<name>add</name>
<bind name="1">in1t</bind>
...
<register-file name="RF">
<type>normal</type>
...
<address-space name="data">
<width>8</width>
...
2.2.2 TTA Program Exchange Format (TPEF)
Filename extension: .tpef
version 31 2016-02-23 11/145
TTA-based Co-design Environment v1.14 User Manual
TTA Program Exchange Format (TPEF) is a file format for storing compiled TTA programs which can
be either unscheduled, partially scheduled, or scheduled [Cil04]. TPEF supports auxiliary sections for
storing additional information related to the program, such as execution profiles, machine resource data,
and target address space definitions. This file is needed, e.g. by ttasim and Pig.
2.2.3 Hardware Database (HDB)
Filename extension: .hdb
Hardware Database (HDB) is the main SQLite database used by the Processor Generator and the Cost
Estimator. The data stored in HDB consist of hardware description language definitions (HDL) of TTA
components (function units, register files, buses and sockets) and metadata that describe certain parame-
ters used in implementations. In addition, HDB may include data of each implementation needed by the
cost estimation algorithms.
TCE ships with an example HDB that includes implementations for several function units and register
files in VHDL, and cost data for the default interpolating cost estimation plugin.
2.2.4 Implementation Definition File (IDF)
Filename extension: .idf
Describes which implementations to use for each component in the architecture. Hence, this links ar-
chitectural components described in ADF files to implementation alternatives stored in HDB. Using this
information it is possible to fetch correct hardware description language (HDL) files from the hardware
block library for cost estimation and processor generation. This file is needed, by ProGe and PIG.
The XML snippet below gives an idea of IDF.
<adf-implementation>
...
<fu name="LSU">
<hdb-file>asic_130nm_1.5V.hdb</hdb-file>
<fu-id>62</fu-id>
</fu>
...
2.2.5 Binary Encoding Map
Filename extension: .bem
Provides enough information to produce an executable uncompressed bit image from TPEF program data.
2.2.6 Operation Set Abstraction Layer (OSAL) Files
Filename extension: .opp, .cc, .opb
OSAL stores the simulation behavior and static properties of operations that can be added to the function
units in the designed processors.
Simulation behavior of function unit operations is described by implementing simulation functions which
can be plugged in to the simulator run time.
The .opp file is an XML file for defining the static properties of operations (for example, how many inputs
and outputs an operation has). The .cc is the C++ source file that defines the behavior model for a set of
operations. The .opb is the plugin module compiled from the .cc. OSAL is quite complex and further
details are presented in Section 4.3.
12/145 2016-02-23 version 31
![](asset-d.png)
TTA-BASED CO-DESIGN ENVIRONMENT
2.2.7 Simulation Trace Database
Filename extension: .tracedb
Stores data collected from simulations and used by instruction scheduler (profiling data) and cost estima-
tor (utilization statistics, etc.).
2.2.8 Exploration Result Database
Filename extension: .dsdb
Exploration Result Database (ExpResDB) contains the configurations that have been evaluated during
exploration (manual or automatic) and a summary of their characteristics. Files that define each tested
configuration (ADF and IDF) are stored in the database as well.
2.2.9 Summary of file types
Table 2.2 summarizes briefly the most common TCE file types.
Table 2.2: Summary of file types
Postfix Purpose Format
adf Architecture description XML
idf Implementation/microarchitecture description XML
hdb HW database SQLite
tpef TTA program exchange format bin
bem Binary encoding map XML
opp Operation properties XML
cc Simulation behavior of an operation C++
opb Compiled operation behavior bin
tracedb Database of simulation and cost estimation results SQLite
dsdb Database of exploration results and TTA configurations SQLite
version 31 2016-02-23 13/145
TTA-based Co-design Environment v1.14 User Manual
Chapter 3
TUTORIALS AND HOW-TOS
This section presents 11 tutorials covering the basics, printing and streaming I/O, floating-point computa-
tion, parallel assembly, multiple memories, FPGA considerations and programming TTAs using OpenCL
C.
Let’s start with the simplest one.
3.1 TCE Tour
This tutorial goes through most of the tools in TCE using a fairly simple example application. It starts
from C code and ends up with VHDL of the processor and a bit image of the parallel program. This
tutorial will also explain how to accelerate an algorithm by customizing the instruction set, i.e., by using
custom operations. The total time to go through the tutorial is about 2 −3 hours.
The tutorial file package is available at:
http://tce.cs.tut.fi/tutorial_files/tce_tutorials.tar.gz
Fetch and unpack it to a working directory and then enter the directory:
> wget http://tce.cs.tut.fi/tutorial_files/tce_tutorials.tar.gz
> tar -xzf tce_tutorials.tar.gz
> cd tce_tutorials/tce_tour
> ls -la
total 84
drwxr-xr-x 3 tce tce 4096 2010-05-28 11:40 .
drwx------ 7 tce tce 4096 2012-05-18 13:22 ..
-rw------- 1 tce tce 5913 2010-03-08 20:01 crc.c
-rw------- 1 tce tce 1408 2008-11-07 11:35 crc.h
-rw------- 1 tce tce 3286 2008-11-07 11:35 crcTable.dat
-rw-r--r-- 1 tce tce 2345 2010-03-08 13:04 custom_operation_behavior.cc
-rw-r--r-- 1 tce tce 855 2010-05-28 11:41 custom_operations.idf
-rw------- 1 tce tce 1504 2010-03-08 20:01 main.c
-rw-r--r-- 1 tce tce 45056 2010-03-10 16:09 tour_example.hdb
drwxr-xr-x 2 tce tce 4096 2010-05-28 11:40 tour_vhdl
3.1.1 The Sample Application
The test application counts a 32-bit CRC (Cyclic Redundant Check) check value for a block of data, in
this case 10 bytes. The C code implementation is written by Michael Barr and it is published under Public
14/145 2016-02-23 version 31
![](asset-f.png)
TTA-BASED CO-DESIGN ENVIRONMENT
Figure 3.1: Processor Designer (prode) when minimal ADF is opened
Domain. The implementation consists of two different version of crc, but we will be using the fast version
only.
The program consists of two separate files: ’main.c’ contains the simple main function and ’crc.c’ con-
tains the actual implementation. Open ’crc.c’ in your preferred editor and take a look at the code. Algo-
rithm performs modulo-2 division, a byte at a time, reflects the input data (MSB becomes LSB) depending
on the previous remainder, and XORs the remainder with a predefined polynomial (32-bit constant). The
main difference between the crcSlow and crcFast implementations is that crcFast exploits a precalculated
lookup table. This is a quite usual method of algorithm optimization.
3.1.2 Starting Point Processor Architecture
We use the minimal architecture as the starting point. File ’minimal.adf’ describes a minimalistic archi-
tecture containing just enough resources that the TCE compiler can still compile programs for it. Function
units in ’minimal.adf’ are selected from the hardware database (HDB, Section 2.2.3) so we are able to
generate a VHDL implementation of the processor automatically later in the tutorial.
Copy the ’minimal.adf’ included in TCE distribution to a new ADF file which is your starting point
architecture. You can view it using the graphical Processor Designer (ProDe, Section 4.1) tool:
cp $(tce-config --prefix)/share/tce/data/mach/minimal.adf start.adf
prode start.adf &
Fig. 3.1 shows how Prode should look like. There is 1 bus and 5 units: global control unit (GCU), 2
register files , 1 ALU ,and 1 LSU.
You’ll learn how to edit the TTA with ProDe later in this tutorial.
If you want later to simulate the generated processor using GHDL (Section 3.1.9), you should decrease
the amount of memory in the processor already now. Otherwise the GHDL executing the generated
testbench might consume tremendous amount of memory on your computer and even crash. Select Edit
-> Address Spaces in ProDe. Then edit the bit widths of data and instruction address spaces and set them
to 15 bits which should be plenty for our case.
3.1.3 Compiling and simulating
Now we want to know how well the starting point architecture executes our program. We must compile
the source code for this architecture with command:
tcecc -O3 -a start.adf -o crc.tpef -k result main.c crc.c
In addition to source codes, the compiler needs the TTA architecture definition ’start.adf’. It will produce
a parallel program called ’crc.tpef’ which can only be executed on this architecture. The switch −ktells
the compiler to keep the result symbol (variable on line 29 of main.c) also in the generated program in
order to access the variable contents by name in the simulator later on.
version 31 2016-02-23 15/145
大概是对crc这个项目使用start.adf产生一个特定的结构。
![](asset-10.png)
TTA-based Co-design Environment v1.14 User Manual
Figure 3.2: Processor Simulator (proxim) when minimal ADF and crc.tpef are opened and simulated.
Successful compilation does not print anything and we can now simulate the program. Let’s use graphical
user interface version of the simulator called Proxim (Section 6.1.6):
proxim start.adf crc.tpef &
The simulator will load the architecture definition and the program and wait for commands. Proxim shows
one line per instruction. Minimal TTA has only bus and hence there is only instruction slot (column). The
total number of instructions is about 1 200.
Execute the program by clicking “Run” and Proxim should look like Fig. 3.2. Write down this cycle
count in the bottom bar of the simulator, you can use Table 3.1. for future comparison.
You can check the result straight from the processor’s memory by writing this command to the command
line at the bottom of the simulator:
x /u w result
The command specifies the data width (4 bytes) and the result symbol we told the compiler to preserve
before. The correct checksum result is 0x62488e82.
Processor resource utilization data can be viewed with command:
info proc stats
This will output a lot of information like the utilization of transport buses, register files and function units,
for example:
utilizations
------------
buses:
B1 87.7559% (4415 writes)
sockets:
lsu_i1 17.4319% (877 writes)
16/145 2016-02-23 version 31
![](asset-11.png)
TTA-BASED CO-DESIGN ENVIRONMENT
Figure 3.3: Processor Simulator (proxim) shows the utilizations of each FU and bus.
lsu_o1 10.4949% (528 writes)
...
operations executed in function units:
LSU:
...
TOTAL 17.4319% (877 triggers)
ALU:
...
TOTAL 25.2037% (1268 triggers)
...
register accesses:
RF:
0 855 reads, 0 guard reads, 1 writes
...
Your output may vary depending on the compiler version, the optimizations used and your architecture.
For more information on the simulator’s command line tools and syntax, you may use the
help
command.
Proxim can also show othes various pieces of information about the program’s execution and its processor
resource utilization. For example to check the utilization of the resources of our architecture, select
View>Machine Window from the top menu. The parts of the processor that are utilized the most are
visualized with darker red color.
There are 3 basic ways to accelerate computation in TTAs:
Table 3.1: Cycle counts in different TTA configurations. You’ll fill this during the tutorial
Approx. relative
ADF Cycle count cycle count Section(s) Comment
start.adf 1 3.1.3
large.adf 1/2 3.1.4 start + 3 buses
large.adf 1/5 -”- start + 3 buses + RF
large.adf 1/6 -”- start + 3 buses + RF + new ALU
custom.adf 1/5 3.1.5 - 3.1.9
large_custom.adf 1/50 3.1.10
version 31 2016-02-23 17/145
TTA-based Co-design Environment v1.14 User Manual
1. Modify the algorithm itself. Beyond the scope of this tutorial.
2. Add more basic resources (FUs, RFs, buses) to TTA
3. Add custom operation(s) to TTA
3.1.4 Increasing performance by adding resources
As the current architecture is minimalistic we can increase the performance even further by adding re-
sources to the processor.
Transport buses. The architecture has only one transport bus, thus the compiler can’t exploit any in-
struction level parallelism. Let’s start the architecture customization by adding another transport bus.
After this there can be 2 moves per clock cycle. First copy the current architecture:
cp start.adf large.adf
and open the new architecture in ProDe:
prode large.adf &
A new transport bus can be added simply by selecting the current bus and pressing “ctrl+c” to copy
the bus and then pressing “ctrl+v” to paste it. Add 3 more buses. After you have added the buses you
have to connect it to the sockets. Easiest way to do this is to select “Tools->Fully connect IC”. Save the
architecture, recompile the source code for the new architecture:
tcecc -O3 -a large.adf -o crc.tpef -k result crc.c main.c
This time simulate the program using the command line instruction set simulator:
ttasim -a large.adf -p crc.tpef
The command will load the architecture and program and execute it. Once the simulation is finished, you
should see the ttasim command prompt:
> ttasim -a large.adf -p crc.tpef
(ttasim)
Now when you check the cycle count from the simulator:
info proc cycles
You might see a significant drop in cycles. Also check the processor utilization statistics from the simu-
lator with command:
info proc stats
Register files. From the previous simulator statistics you can see from “operations” table that there are
a lot of load and store operations being executed. As the architecture has only 5 general purpose registers
this tells us that there are a lot of register spilling to memory. Let’s try how the amount of registers affect
the cycle count. There are two options how we can add registers. We can either increase the number of
registers in a register file or add a new register file.
Let’s try the latter option because this way we increase the number of registers that can be accessed
simultaneously on one clock cycle. This can be done by selecting the RF and using copy and paste. Then
connect it to the IC. Simulation statistics should indicate performance increase. As expected, the number
of load and store opertations decreased. But notice also that the number of add operations decreased quite
a lot. The reason is simple: addition is used to calculate stack memory addresses for the spilled registers.
Function units. Next subject for the bottleneck is the ALU as now all the basic operations are per-
formed in a single function unit. From the simulator statistics you can see that logical operations and
addition are quite heavily utilized. Instead of duplicating the ALU let’s add more specific FUs from the
Hardware Database. Select “Edit->Add From HDB->Function Unit...”. Select a FU which has operations
18/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
and(1), ior(1), xor(1) and click “Add”. Then select FU with operation add(1) and click “Add”. Close the
dialog, connect the function units and save the architecture. Recompile and simulate to see the effect on
cycle count.
The architecture could be still modified even further to drop the cycle count but let’s settle for this now
and move on to custom operations.
3.1.5 Analyzing the Potential Custom Operations
Custom operations implement application specific functionality in TTA processors. This part of the tu-
torial teaches how to accelerate the CRC computation by adding a custom operation to the starting point
processor design.
First of all, it is quite simple and efficient to implement CRC calculation entirely on hardware. Naturally,
using the whole CRC function as a custom operation would be quite pointless and the benefits of using a
processor based implementation would get smaller. Instead, we will concentrate on trying to accelerate
smaller parts of the algorithm, which is the most useful way for most algorithms.
The steps needed for custom opeation are:
1. Analyze the code for bottlenecks
2. Speculatively add a new module and operation to operation database with OSEd. Define the be-
havior of the new operation (copy-paste from original code and add few macros). Optionally, you
can also simulate the operation behavior. Bind the new operation to a new function unit in ProDe
3. Modify your source code so that it utilizes the new operation and simulate. Go back to step 2 if
you wish to explore other operations
4. Create VHDL for the operation and add it to HDB.
5. Finally, generate HDL implementation of the TTA with ProGe.
Finding the operation to be optimized is quite obvious in this case if you look at function crcFast(). It
consists of a for-loop in which the function reflect() is called through the macro REFLECT_DATA. If
you look at the actual function (shown also in Fig. 3.4 you can see that it is quite simple to implement on
hardware, but requires many instructions if done with basic operations in software. The function “reflects”
the bit pattern around the middle point like a mirror. For example, the bit pattern 1101 0010 becomes
0100 1011. The main complexity of the function is that the bit pattern width is not fixed. Fortunately, the
width cannot be arbitrary. If you examine the crcFast()-function and the reflect macros you can spot that
function reflect() is only called with 8 and 32 bit widths (unsigned char and ’crc’ which is an unsigned
long).
A great advantage of TCE is that the operation semantics, processor architecture, and implementation are
separate abstractions. This simplifies designing custom operations since you can simulate your design by
simply defining the simulation behaviour of the operation and setting the latency of the operation to the
processor architecture definition. This is nice as you do not need an actual hardware implementation of
the operation at this point of the design, but can evaluate different custom operation possibilities at the
architectural level. However, this brings up an awkward question: how to determine the latency of the
operation? Unrealistic or too pessimistic latency estimates can produce inaccurate performance results
and bias the analysis.
One approach to the problem is to take an educated guess and simulate some test cases with different
custom operation latencies. This way you can determine a latency range in which the custom operation
would accelerate your application to the satisfactory level. For example, custom operation with 1 cycle
delay might give 3xspeedup but 10 cycles gives just 3%. After this you can sketch how the operation
could be implemented in hardware, or consult someone knowledgeable in hardware design to figure out
whether the custom operation can be reasonably implemented within the latency constraint.
Another approach is to try and determine the latency by examining the operation itself and considering
how it could be implemented. This approach requires some insight in digital design.
version 31 2016-02-23 19/145
![](asset-14.png)
TTA-based Co-design Environment v1.14 User Manual
Figure 3.4: Reflect function of CRC
Besides latency you should also consider the size of the custom function unit. It will consume extra die
area, but the size limit is always case-specific, especially when area taken by memories is accounted.
Accurate size estimation requires the actual HW implementation and synthesis.
Let us consider the reflect function. If we had fixed width we could implement the reflect by hard wiring
(and registering the output) because the operation only moves bits to other locations in the word. This
could be done easily in one clock cycle, as in right side of Fig. 3.4. But we need two different bit widths
so it is somewhat more complicated. We could design the HW in such way that it has two operations: one
for 8-bit data and another for 32-bit data. One way to implement this is to have 32-bit wide crosswiring
and register the output. The 8-bit value would be reflected to the 8 MSB bits of the 32-bit wiring. Then
we need to move the 8 MSB bits to the LSB end and set rest to zero. This moving can be implemented
using multiplexers. So concerning the latency this can all be done easily within one clock cycle as there
is not much logic needed.
3.1.6 Creating the Custom Operation
Now we have decided the operation to be accelerated and its latency.
Next we will create a function unit implementing the operation and add it to our processor design. First,
a description of the semantics of the new operation must be added at least to Operation Set Abstraction
Layer (Sections 2.2.6 and 4.3). OSAL stores the semantic properties of the operation, which includes the
simulation behavior, operand count etc., but not the latency. OSAL definitions can be added by using the
OSAL GUI, OSEd (Section 4.2.1).
Synthesis or simulation at the VHDL level requires that at least one function unit implementing the
operation must be added to the Hardware Database (Section 2.2.3). In this tutorial we add the FU imple-
mentation for our custom operation so the processor implementation can be generated, but omit the cost
data required for the cost estimation. In general case, cost data should be added to cost database as well.
Using Operation Set Editor (OSEd) to add the operation data. OSEd is started with the command
osed &
Create a new operation module, which is a container for a set of operations. You can add a new module
in any of the predefined search paths, provided that you have sufficient file system access permissions.
For example, choose directory ’/home/user/.tce/opset/custom’, where user is the name of the user account
being used for the tutorial. This directory is intended for the custom operations defined by the current
user, and should always have sufficient access rights.
1. Click the root in the left area of the main window which opens list of paths. Right-click on a path
20/145 2016-02-23 version 31
![](asset-15.png)
TTA-BASED CO-DESIGN ENVIRONMENT
Figure 3.5: Operation Set Editor when adding new operation REFLECT32.
name ’/home/user/.tce/opset/custom’. A drop-down menu appears below the mouse pointer.
2. Select Add module menu item.
3. Type in the name of the module (for example, ‘tutorial’) and press OK. The module is now added
under the selected path.
Adding the new operations. We will now add the operation definitions to the newly created operation
module.
1. Select the module that you just added by right-clicking on its name, displayed in the left area of the
main window. A drop down menu appears.
2. Select Add operation menu item.
3. Type ‘REFLECT8’ as the name of the operation.
4. Add one input by pressing the Add button under the operation input list. Select UIntWord as type.
5. Add one output by pressing the Add button under the operation output list. Select UIntWord as
type.
6. After the inputs and the output of the operation have been added, close the dialog by pressing the
OK button. A confirmation dialog will pop up. Press Yes to confirm the action. The operation
definition is now added to the module.
7. Then repeat the steps for operation ‘REFLECT32’, as shown in Fig. 3.5
Defining the simulation behaviour of the operations The new operations REFLECT8 and REFLECT32
do not yet have simulation behavior models, so we cannot simulate programs that use these operations
with the TCE processor simulator. Open again the operation property dialog by right-clicking RE-
FLECT8, then choosing Modify properties. Now press the Open button at the bottom to open an empty
behavior source file for the module. Copy-paste (or type if you have the time!) the following code in the
editor window:
version 31 2016-02-23 21/145
TTA-based Co-design Environment v1.14 User Manual
#include "OSAL.hh"
OPERATION(REFLECT8)
TRIGGER
unsigned long data = UINT(1);
unsigned char nBits = 8;
unsigned long reflection = 0x00000000;
unsigned char bit;
/*
* Reflect the data about the center bit.
*/
for (bit = 0; bit < nBits; ++bit)
{
/*
* If the LSB bit is set, set the reflection of it.
*/
if (data & 0x01)
{
reflection |= (1 << ((nBits - 1) - bit));
}
data = (data >> 1);
}
IO(2) = static_cast<unsigned> (reflection);
return true;
END_TRIGGER;
END_OPERATION(REFLECT8)
OPERATION(REFLECT32)
TRIGGER
unsigned long data = UINT(1);
unsigned char nBits = 32;
unsigned long reflection = 0x00000000;
unsigned char bit;
/*
* Reflect the data about the center bit.
*/
for (bit = 0; bit < nBits; ++bit)
{
/*
* If the LSB bit is set, set the reflection of it.
*/
if (data & 0x01)
{
reflection |= (1 << ((nBits - 1) - bit));
}
data = (data >> 1);
}
22/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
IO(2) = static_cast<unsigned> (reflection);
return true;
END_TRIGGER;
END_OPERATION(REFLECT32)
This code has the behaviours for the both operations. These behavior definitions reflect the input operand
integer (id 1, assigned to variable ’data’) and writes the result to the ”output operand“ (id 2, assigned
from variable ’reflection’) which is the first output and signals the simulator that all results are computed
successfully.
Open file ’crc.c’ in your preferred editor. Compare the behaviour definition of reflect operations and the
original reflect-function. The function is mostly similar except for parameter passing and adding few
macros added, such as OPERATION(), TRIGGER, IO(2) etc. Custom hardware operation behavior reads
data from the function unit’s input ports and writes to output ports. The value of nBits is determined from
the operation code (REFLECT8 or REFLECT32).
Save the code and close the editor. REFLECT8 and REFLECT32 operations now have TCE simulator
behaviour models.
Compiling operation behavior. REFLECT-operations have been added to the test module. Before we
can simulate the behavior of our operation, the C++-based behavior description must be compiled to a
plugin module that the simulator can call.
1. Right-click on the module name (’tutorial’) displayed in the left area to bring up the drop down
menu.
2. Select Build menu item.
3. Hopefully, no errors were found during the compilation! Otherwise, re-open the behaviour source
file and try to locate the errors with the help of the diagnostic information displayed in the build
dialog.
After the operation simulation model has been added and compiled the operation can be simulated. But
for the sake of speed up we will skip the operation simulation here. Interested readers are referred to
Section 4.2.1 You can close osed now.
Adding a Customized Function Unit to the Architecture. Now the operation definitions of the custom
operations have been added to the Operation Set Abstraction Layer (OSAL) database. Next we need to
add at least one functional unit (FU) which implements these operations so that they can be used in the
processor design. Note the separation between ”operation“ and an ”function unit“ that implements the
operation(s) which allows using the same OSAL operation definitions in multiple FUs with different
latencies.
First, add the architecture of the FU that implements the custom operations to the starting point processor
architecture. Let’s take a copy of the starting point processor design which we can freely modify and still
be able to easily compare the architecture with and without the custom operation support later on:
cp start.adf custom.adf
Open the copy in ProDe:
prode custom.adf &
Then:
1. Add a new function unit to the design, right click the canvas and select: Add>Function Unit. Name
the FU ”REFLECTER“. Add one input port (named as trigger) and an output port (output1) to
the FU in the Function unit dialog. Set the input port (trigger) triggering (Click the port named
trigger->Edit->Check dialog ”triggers“). This port starts the execution of the operation when it is
written to.
version 31 2016-02-23 23/145
![](asset-18.png)
TTA-based Co-design Environment v1.14 User Manual
Figure 3.6: Adding operation REFLECT8 to a new function unit in ProDe
2. Add the operation ”REFLECT8“ we defined to the FU: Add from opset>REFLECT8>OK and set
the latency to 1. Click on the REFLECT8 operation and ensure that the operation input is bound to
the input ports and the output is bound to the output port. Check that the operand usage is in such
a way that input is read at cycle 0 and the result is written at the end of the cycle (can be read from
the FU on the next cycle), as in Fig. 3.6. Thus, the latency of the operation is 1 clock cycles.
3. Repeat the previous step for operation ”REFLECT32“
4. Now an FU that supports the custom operations has been added to the architecture. Next, connect
the FU to the rest of the architecture. This can be most easliy done by selecting Tools->Fully
Connect IC which connects all FUs to all the buses. Save the architecture description by clicking
Save.
3.1.7 Use the custom operation in C code.
To get some benefits from the added custom hardware, we must use it from the C code. This is done by
replacing a C statement with a custom operation invocation.
Let us first backup the original C code.
cp crc.c crc_with_custom_op.c
Then open ’crc_with_custom_op.c’ in your preferred text editor.
1. Add #include “tceops.h” to the top of the file. This includes automatically generated macros which
allow us to use specific operations from C code without getting our hands dirty with inline assembly.
Usage of these macros is as follows:
_TCE_<opName>(input1, ... , inputN, output1, ... , outputN);
where <opName> is the name of the operation in OSAL. Number of input and output operands
depends on the operation. Input operands are given first and they are followed by output operands
if any.
24/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
In our case we need to write operands into the reflecter and read the result from it. We named the
operations “REFLECT8” and “REFLECT32”, thus the macros we are going to use are as follows:
_TCE_REFLECT8(input1, output);
_TCE_REFLECT32(input1, output);
Now we will modify the crcFast function to use the custom op. First declare 2 new variables at the
beginning of the function:
crc input;
crc output;
These will help in using the reflect FU macro.
Take a look at the REFLECT_DATA and REFLECT_REMAINDER macros. The first one has got
a magic number 8 and “variable” X is the data. This is used in the for-loop of crcFast().
The input data of reflect function is read from array message[] in the for-loop the. Let us modify this
so that at the beginning of the loop the input data is read to the input variable. Then we will use the
_TCE_REFLECT8 macro to run the custom operations, and finally replace the REFLECT_DATA
macro with the output variable. After these modifications the body of the for-loop should look like
this:
input = message[byte];
_TCE_REFLECT8(input, output);
data = (unsigned char) output ^ (remainder >> (WIDTH - 8));
remainder = crcTable[data] ^ (remainder << 8);
Next we will modify the return statement. Originally it uses REFLECT_REMAINDER macro
where nBits is defined as WIDTH and data is remainder. Simply use _TCE_REFLECT32 macro
before return statement and replace the original macro with the variable output:
_TCE_REFLECT32(remainder, output);
return (output ^ FINAL_XOR_VALUE);
And now we are ready. Remember to save the file.
2. Compile the custom operation using C code to a parallel TTA program using the new architecture
which includes a FU with the custom operation:
tcecc -O3 -a custom.adf -o crc_with_custom_op.tpef -k result \
crc_with_custom_op.c main.c
3. Simulate the parallel program. This time we will use the command line simulator ttasim. We will
also enable writing of bus trace. It means that the simulator writes a text file containing the bus
values of the processor from every executed clock cycle. This bus trace data will be used to verify
the processor RTL implementation. Start the simulator with command:
ttasim
Then enable the bus trace setting:
setting bus_trace 1
Load architecture and program and run the simulation
mach custom.adf
prog crc_with_custom_op.tpef
run
Verify that the result is the same as before (x /u w result). It should be the same as earlier
(0x62488e82). Check the cycle count info proc cycles and compare it to the cycle count of the
version 31 2016-02-23 25/145
TTA-based Co-design Environment v1.14 User Manual
version which does not use a custom operation. You should see a very noticeable drop compared
to the starting point architecture without the custom operations. Write this cycle count down for a
later step and quit ttasim.
The simulator execution also created a file ’crc_with_custom_op.tpef.bustrace’ which contains the
bus trace.
3.1.8 Adding HDL implementation of the FU to the hardware database (HDB).
Now we have seen that the custom operation accelerates our application. Next we’ll add a VHDL imple-
mentation of the custom FU to Hardware Database (hdb). This way we will be able to generate a VHDL
implementation of our processor.
If you want to skip this phase you can use the given ’tour_example.hdb’ instead of creating it yourself.
Start HDBEditor (see Section 4.8):
hdbeditor &
TCE needs some data of the FU implementation in order to be able to automatically generate processors
that include the FU.
1. function unit from ’custom.adf’ file (edit->add->FU architecture from ADF). You can leave the
checkboxes ”parametrized width“ and ”guard support“ unchecked, which is the default. Then de-
fine implementation for the added function unit entry right click reflect -> Add implementation....
2. Open file ’tour_vhdl/reflect.vhdl’ that was provided in the tutorial package with the editor you
prefer, and take a look. This is an example implementation of a TTA function unit performing the
custom ’reflect8’ and ’reflect32’ operations.
3. The HDB implementation dialog needs the following information from the VHDL:
1. Name of the entity and the naming of the FU interface ports.
Name the implemention after the top level entity: “fu_reflect”.
By examining the VHDL code you can easily spot the clock port (clk), reset (rstx) and global lock
port (glock). Operation code (opcode) port is t1opcode. Write these into the appropriate text boxes.
You do not have to fill the Global lock req. port field because the function unit does not stall, i.e.
does not request a global lock to the processor during its execution.
2. Opcodes.
Check that the operation codes match the top of the vhdl file. REFLECT32 has operation code “0”
and REFLECT8 has operation code “1”.
The operation codes must be always numbered according to the alphabetical order of the OSAL
operation names, starting at 0. For example, in this case REFLECT32 is earlier than REFLECT8
in the alphabetical order (i.e. 3 becomes before 8).
3. Parameters.
Parameters can be found from the VHDL generics. On top of the file there is one parameter: busw.
It tells the width of the transport bus and thus the maximum width of input and output operands.
Thus, add parameter named busw, type it as integer and set the value to 32 in the Parameter dialog.
4. Architecture ports.
These settings define the connection between the ports in the architectural description of the FU
and the VHDL implementation. Each input data port in the FU is accompanied with a load port
that controls the updating of the FU input port registers.
Choose a port in the Architecture ports dialog and click edit. Name of the architecture port p1 is
t1data (as in VHDL) and associated load port is t1load. Width formula is the parameter busw. No
guard port is needed in this case.
26/145 2016-02-23 version 31
![](asset-1b.png)
TTA-BASED CO-DESIGN ENVIRONMENT
Figure 3.7: Adding the implementation of custom operations to HW database
version 31 2016-02-23 27/145
TTA-based Co-design Environment v1.14 User Manual
Name the output port (p2) to r1data and the width formula is now busw because the output port
writes to the bus. The output port does not have a load port.
5. Add VHDL source file.
Add the VHDL source file into the Source code dialog. Notice that the HDL files must be added
in the compilation order (see section 4.8). But now we have only one source file so we can simply
add it without considering the compilation order (Add -> Browse -> tour_vhdl/reflect.vhdl).
Now you are done with adding the FU implementation, see Fig. 3.7. Click OK.
3.1.9 Generating the VHDL and memory images
In this step we generate the VHDL implementation of the processor, and the bit image of the parallel
program.
Select Function Unit Implementations You can either use the given ’custom_operations.idf’ included
in the tutorial files or select the implementations yourself. If you use the given file replace ’custom.idf’
with ’custom_operations.idf’ in the following commands.
Next, we must select implementations for all components in the architecture. Each architecture compo-
nent can be implemented in multiple ways, so we must choose one implementation for each component
to be able to generate the implementation for the processor.
This can be done in the ProDe tool:
prode custom.adf
Then we’ll select implementations for the FUs which can be done in Tools>Processor Implementation....
Note that the selection window is not currently very informative about the different implementations, so
a safe bet is to select an implementation with parametrizable width/size.
1. Selecting implementations for register files, immediate units and function units can be done two
different ways.
(a) Manual selection - user chooses the implementations from HDB files.
i. Select implementation for RF: Click the RF name, ’Select Implementation’, find the
TCE’s default HDB file (PREFIX/share/tce/hdb/asic_130nm_1.5V.hdb) from your tce
installation path and select an implementation for the RF from there.
ii. Next select implementation for the boolean RF like above. But this time select an imple-
mentation which is guarded i.e. select an implementation which has word “guarded_0”
in its name.
iii. Similarly, select implementations for the function units from TCE’s default HDB. Notice
that it is vital that you choose the implementation for LSU from the asic_130nm_1.5V.hdb.
Then select implementation for the reflecter but this time you have to use the ’tour.hdb’
created earlier to find the FU we added that supports the REFLECT custom operations.
(b) Automatic selection - ProDe chooses the implementations from HDB files.
i. Select implementations for register files and function units all at once: Click Auto Se-
lect Implementations, find the TCE’s default HDB file from your TCE installation path
(PREFIX/share/tce/hdb/asic_130nm_1.5V.hdb), make sure ’Register Files’ and ’Func-
tion Units’ checkboxes are checked and press ’Find’. A dialog should appear saying 4
implementations were found; 2 for register files and 2 for function units. Click ’Yes’.
If the number of found implementations was under 4 in your case, refer to the manual
implementation selection above.
ii. Browse to Function Units page if you were not on it already. You should see, that reflecter
FU is still without implementation (because it is not defined in asic_130nm_1.5V.hdb).
Select Auto Select Implementations again, find ’tour.hdb’ as the HDB file, and press
’Find’. 1 FU implementation should be found (the reflecter), click ’Yes’. Now all your
register files and function units should have implementations.
28/145 2016-02-23 version 31
![](asset-1d.png)
TTA-BASED CO-DESIGN ENVIRONMENT
Figure 3.8: Generating TTA VHDL with ProGe
2. Next select the IC/Decoder generator plugin used to generate the decoder in the control unit and
interconnection network: Browse... (installation_path)/share/tce/icdecoder_plugins/base/ Default-
ICDecoderPlugin.so>OK. This should be selected by default.
3. Enable bus tracing from the Implementation-dialog’s IC / Decoder Plugin tab. Set the bustrace
plugin parameter to “yes” and the bustracestartingcycle to “5”. The IC will now have a component
which writes the bus value from every cycle to a text file. Notice that this option cannot be used if
the processor is synthesized.
You do not have to care about the HDB file text box because we are not going to use cost estimation
data.
4. Click “Save IDF...”
Generate the VHDL for the processor using Processor Generator (ProGe). You can start proces-
sor generation from ProDe’s implementation selection dialog: Click “Generate Processor”. For Binary
Encoding Map: Select the “Generate new”, see Fig. 3.8
In the target directory click “Browse” and create a new directory ’proge-output’ and select it. Then click
OK to create the processor.
Or alternatively execute ProGe from command line:
generateprocessor -t -i custom.idf -o proge-output custom.adf
Flag −tgenerates a testbench, −idefines the implementation file, and −othe output directory. Now
directory ’proge-output’ includes the VHDL implementation of the designed processor except for the
instruction memory width package which will be created by Program Image Generator. You can take
a look what the directory includes, how the RF and FU implementations are collected up under ’vhdl’
subdir and the interconnection network has been generated to connect the units (the ’gcu_ic’ subdir). The
’tb’ subdir contains testbench files for the processor core. Moreover, there are 4 shell scripts for compiling
and simulating the VHDL codes in Modelsim and GHDL.
Generate instruction memory bit image using Program Image Generator. Finally, to get our shiny
new processor some bits to chew on, we use generatebits to create instruction memory and data memory
images:
generatebits -d -w 4 -p crc_with_custom_op.tpef -x proge-output custom.adf
Flag −dcreates data images, −w4 set data memory width 4 MAUs (Minimum Addressable Units, bytes
in this case), −pdefines the program, −xdefines the HDL directory, and adf is given without a flag.
Now the file ’crc_with_custom_op.img’ includes the instruction memory image in “ascii 0/1” format.
Each line in that file represents a single instruction. Thus, you can get the count of instructions by
counting the lines in that file:
version 31 2016-02-23 29/145
TTA-based Co-design Environment v1.14 User Manual
wc -l crc_with_custom_op.img
Accordingly, the file ’crc_with_custom_op_data.img’ contains the data memory image of the processor.
Program Image Generator also created file ’proge-output/vhdl/tta0_imem_mau_pkg.vhdl’ which contains
the width of the instruction memory (each designed TTA can have a different instruction width). The
_imem_mau_pkg.vhdl file is appended with the top level entity name, which in this case is “tta0”.
Simulation and verification If you have GHDL installed you can now simulate the processor VHDL.
First cd to proge-output directory:
cd proge-output
Then compile and simulate the testbench:
./ghdl_compile.sh
./ghdl_simulate.sh
Compilation gives a couple of warnings from Synopsys’ std_arith package but that should do no harm.
Simulation will take some time as the bus trace writing is enabled. If you did not change the memory
widths in Prode (Section 3.1.2 ), the simulation will crash and print “Killed”. There will be many warnings
at 0ns due uninitialized signals but you can ignore them.
Aftern simulation, you should see a message
./testbench:info: simulation stopped by -stop-time
The simulation produces file “bus.dump” which looks like this:
0 | 0 |
1 | 4 |
2 | 32760 |
3 | 24 |
...
As the testbench is ran for constant amount of cycles we need to get the relevant part out of the bus dump
for verification. This can be done with command:
head -n <number of cycles> bus.dump > sim.dump
where the <number of cycles> is the number of cycles in the previous ttasim execution (same as line
ocunt in crc_with_custom_op.tpef.bustrace). Then compare the trace dumps from the VHDL simulation
and the architecture simulation:
diff -u sim.dump ../crc_with_custom_op.tpef.bustrace
If the command does not print anything the dumps were equal and GHDL simulation matches the ttasim.
Now you have succesfully added a custom operation, verified it, and gained notably performance increase.
Well done!
3.1.10 Further acceleration by adding custom operation to large TTA
As a final step, let us add the custom operation to the large TTA.
cp large.adf large_custom.adf
prode large_custom.adf
Add the reflecter FU from the tour.hdb like done earlier. Fully connect the IC, as in Fig.3.9 and save.
tcecc -O3 -a large_custom.adf -o crc_with_custom_op.tpef -k result crc_with_custom_op.c
main.c
ttasim -a large_custom.adf -p crc_with_custom_op.tpef
info proc cycles
Voila! Now you have a lightning fast TTA.
30/145 2016-02-23 version 31
![](asset-1f.png)
TTA-BASED CO-DESIGN ENVIRONMENT
Figure 3.9: Processor Designer (prode) with large_custom.adf
3.1.11 Final Words
This tutorial is now finished. Now you should know how to customize the processor architecture by
adding resources and custom operations, as well as generate the processor implementation and its in-
struction memory bit image.
This tutorial used a “minimalistic” processor architecture as our starting point. The machine had only
1 transport bus and 5 registers so it could not fully exploit the parallel capabilities of TTA. Then we
increased resources in the processor and the cycle count dropped to half. Adding 2 simple custom op-
erations to the starting point architecture gives a similar speedup and in large machine the speedup in
cyclecount is about 50x.
3.2 From C to VHDL as Quickly as Possible
This tutorial introduces a fast way to generate a processor VHDL model from C source code using the
Design Space Explorer.
If you haven’t already downloaded the tutorial file package, you can get it from:
http://tce.cs.tut.fi/tutorial_files/tce_tutorials.tar.gz
Unpack it to a working directory and then cd to tce_tutorials/c2vhdl/
A script named c2vhdl should generate all the needed files from the C source code. The script is a bash
script using various TCE command line tools, and can be inspected for detailed instructions to generate
the VHDL and other files manually.
c2vhdl application1/complex_multiply.c
Now directory called “proge-output” should include the VHDL implementation of the processor. The
current working directory should also contain the instruction memory bit image file and the instruction
encoding (bem) file along with the processor architecture definition and implementation definition files.
Passing "-e" as the first parameter to c2vhdl script tells it to print estimation data, like area and energy
requirements for the processor it generates.
Tutorial is now finished and you can simulate the generated VHDL implementation of the processor with
a VHDL simulator and synthesize it.
3.3 Hello TTA World!
What would a tutorial be without the traditional “Hello World!” example? Interestingly enough, printing
out “Hello World” in a standalone (operating system free) platform like the TTA of TCE is not totally
straightforward. That is the reason this tutorial is not the first one in this tutorial chapter.
The first question is: where should I print the string? Naturally it is easy to answer that question while
working with the simulator: to the simulator console, of course. However, when implementing the final
version 31 2016-02-23 31/145
TTA-based Co-design Environment v1.14 User Manual
hardware of the TTA, the output is platform dependent. It can be a debugging interface, serial port output,
etc.
In order to make printing data easier from TTAs designed with TCE, we have included an operation
called STDOUT in the “base operation set” that is installed with the TCE. The simulation behavior of the
operation reads its input, expects it to be a 8bit char and writes the char verbatim to the simulator host’s
standard output. Of course, in case the designer wants to use the operation in his final system he must
provide the platform specific implementation for the function unit (FU) implementing the operation, or
just remove the FU after the design is fully debugged and verified.
The default implementation of printf() in TCE uses the STDOUT operation to print out data. Therefore,
implementing a “Hello World” application with TCE is as simple as adding an FU that implements the
STDOUT to the processor design and then calling printf() in the simulated program. The simulator should
print out the greeting as expected.
Here is a simple C code (hello.c) that prints the magic string:
#include <stdio.h>
int main() {
printf("Hello TTA World!");
return 0;
}
Next, compile it to an architecture that implements the STDOUT. TCE ships with a minimal architecture
that also includes the STDOUT function unit, which you can use:
cp $(tce-config --prefix)/share/tce/data/mach/minimal_with_stdout.adf .
tcecc --swfp -O0 hello.c -a minimal_with_stdout.adf -o hello.tpef
It should compile without errors. Beware: the compilation can take a while on a slower machine. This is
because printf() is actually quite a large function and the compiler is not yet optimized for speed.
Finally, simulate the program to get the greeting:
ttasim -a minimal_with_stdout.adf -p hello.tpef --no-debugmode
Hello TTA World!
That’s it. Happy debugging!
3.4 Streaming I/O
Because TTA/TCE is an environment without operating system, there is also no file system available for
implementing file-based I/O. Therefore, one popular way to get input and output to/from the TTA is using
shared memory for communicating the data. For stream processing type of applications, one can also use
an I/O function unit that implements the required operations for streaming.
TCE ships with example operations for implementing stream type input/output. These operations can be
used to read and write samples from streams in the designed TTA processor. The basic interface of the
operations allows reading and writing samples from the streams and querying the status of an input or
output stream (buffer full/empty, etc.). The status operations are provided to allow the software running
in the TTA to do something useful while the buffer is empty or full, for example switch to another thread.
Otherwise, in case one tries to read/write a sample from/to a stream whose buffer is empty/full, the TTA
is locked and the cycles until the situation resolves are wasted.
The example streaming operations in the base operation set are called STREAM_IN, STREAM_OUT,
STREAM_IN_STATUS, and STREAM_OUT_STATUS. These operations have a simulation behavior
definition which simulates the stream I/O by reading/writing from/to files stored in the file system of the
simulator host. STREAM_IN and STREAM_OUT operations allow defining in/out files using TTASIM_STREAM_IN_FILE
and TTASIM_STREAM_OUT_FILE environment variables respectively.
Here is an example C code that implements streaming I/O with the operations:
32/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
#include "tceops.h"
int main()
{
char byte;
int status;
while (1)
{
_TCE_STREAM_IN_STATUS(0, status);
if (status == 0)
break;
_TCE_STREAM_IN(0, byte);
_TCE_STREAM_OUT(byte);
}
return 0;
}
This code uses the TCE operation invocation macros from tceops.h to read bytes from the input stream
and write the same bytes to the output stream until there is no more data left. This situation is indicated
with the status code 0 queried with the STREAM_IN_STATUS operation. The value means the stream
buffer is empty, which means the file simulating the input buffer has reached the end of file.
You can test the code by creating a file <FU_NAME>.in with some test data, compiling the code and
simulating it. The code should create a copy of that file to the stream output file <FU_NAME>.out.
These files should reside in the directory where the simulator is started. You can get the FU_NAME from
the streaming function units in the adf file.
3.4.1 Streaming I/O function units
VHDL implementations and HDB entries for streaming I/O have been included in TCE since version 1.4.
The stream.hdb -file in the TCE installation directory contains one function unit with STREAM_IN and
STREAM_IN_STATUS -operations, and another FU contains the STREAM_OUT and STREAM_OUT_
STATUS -operations. These FUs can directly be instantiated in ProDe and synthesized on an FPGA along
with the rest of the processor.
When these function units are instantiated in ProDe, they are realized with specific operation latencies.
STREAM_IN and STREAM_OUT take 3 clock cycles, and STREAM_IN_STATUS and STREAM_OUT
_STATUS execute in 1 clock cycle.
The Stream In -FU has three external ports. ext_data is 8 bits wide and is used to communicate the data
byte from the external stream source to the TTA FU. When the TTA application program invokes the
STREAM_IN -operation, the ext_data signal is sampled by the FU and the Stream In -FU automatically
sends an acknowledge-signal back to the stream source through the ext_ack port of the FU. The external
stream source is supposed to provide the next stream data value to ext_data upon noticing a rising edge
in ext_ack. The three cycle latency of the operation allows some time for this to happen. Finally, the TTA
application program can query the availability of stream data by invoking the STREAM_IN_STATUS
-operation. This command reads the value of the ext_status port of the Stream In FU and the external
stream device is expected to keep the signal high in this port when there is still data available, and low
when the stream device has run out of data. The application program sees the status as a numerical value
’1’ or ’0’.
The Stream Out -FU works in a similar fashion. The ext_data port is now an output and provides the data
to the external stream sink. The external stream sink is expected to sample the value of ext_data when
ext_dv (data valid) is high. ext_dv is automatically operated by the FU when the application program
invokes the operation STREAM_OUT. STREAM_OUT has also a latency of 3 clock cycles to allow the
version 31 2016-02-23 33/145
![](asset-22.png)
TTA-based Co-design Environment v1.14 User Manual
external stream sink to take care of the data sample. Invoking the STREAM_OUT_STATUS operation
samples the signal in the ext_status -port, which the external stream sink is expected to keep high if there
is still space in the stream sink. When the stream sink is full, the signal in the ext_status -port must be
pulled low by the stream sink.
Having several distinct stream sources or sinks must at the moment be realized by manually copying the
FUs along with their HDB and VHDL parts. The operations must have distinct names so that the compiler
is explicitly instructed to read from a specific source or write to a specific sink.
3.5 Implementing Programs in Parallel Assembly Code
This tutorial will introduce you to TTA assembly programming. It is recommended that you go through
this tutorial because it will certainly familiarize you with TTA architecture and how TTA works.
3.5.1 Preparations
For the tutorial you need to download file package from
http://tce.cs.tut.fi/tutorial_files/
tce_tutorials.tar.gz
and unpack it to a working directory. Then cd to parallel_assembly-directory.
The first thing to do is to compile the custom operation set called cos16 shipped within the parallel_assembly-
directory. The easiest way to do this is:
buildopset cos16
This should create a file named ’cos16.opb’ in the directory.
3.5.2 Introduction to DCT
Now you will be introduced to TCE assembler language and assembler usage. Your task is to write TCE
assembly code for 2-Dimensional 8 times 8 point Discrete Cosine Transform (DCT_8x8). First take a
look at the C code of DCT_8x8 ’dct_8x8_16_bit_with_sfus.c’. The code is written to support fixed point
datatype with sign plus 15 fragment bits, which means coverage from −1to 1−215. The fixed point
multiplier, function mul_16_fix, and fixed point adder, function add_16_fix, used in the code scale inputs
automatically to prevent overflow. Function cos16 takes x(2i+1)as input and returns the corresponding
cosine value cosx(2i+1)π
16 . The code calculates following equations:
F(x) = C(x)
2
7
∑
i=0f(i)cosx(2i+1)π
16 (3.1)
F(y) = C(y)
2
7
∑
i=0f(i)cosy(2i+1)π
16 (3.2)
C(i) = 2
√2,i=0
1,else .(3.3)
F(x,y) = F(x)F(y)(3.4)
3.5.3 Introduction to TCE assembly
First take a look at assembly example in file ’example.tceasm’ to get familiar with syntax. More help can
be found from section 5.3
Compilation of the example code is done by command:
tceasm -o example.tpef dct_8x8_16_bit_with_sfus.adf example.tceasm
The assembler will give some warnings saying that “Source is wider than destination.” but these can be
ignored.
34/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
The compiled tceasm code can be simulated with TCE simulator, ttasim or proxim(GUI).
ttasim -a dct_8x8_16_bit_with_sfus.adf -p example.tpef
, or
proxim dct_8x8_16_bit_with_sfus.adf example.tpef
It is recommended to use proxim because it is more illustrating to track the execution with it. Especially
if you open the Machine Window (View -> Machine Window) and step through the program.
Check the result of example code with command (you can also write this in proxim’s command line at
the bottom of the main window):
x /a IODATA /n 1 /u b 2
.
the output of x should be 0x40.
3.5.4 Implementing DCT on TCE assembly
Next try to write assembly code which does the same functionality as the C code. The assembly code
must be functional with the given machine ’dct_8x8_16_bit_with_sfus.adf’. Take a look at the processor
by using prode:
prode dct_8x8_16_bit_with_sfus.adf &
The processor’s specifications are the following:
Supported operations
Operations supported by the machine are: mul,mul_16_fix,add,add_16_fix,ldq,ldw,stq,stw,shr,shl,
eq,gt,gtu,jump,cos16 and immediate transport.
When you program using TTA assembly you need to take into account operation latencies. The jump
latency is four clock cycles and load latencies (ldq and ldw) are three cycles. Latency for multiplications
(mul and mul_16_fix) are two clock cycles.
Address spaces
The machine has two separate address spaces, one for data and another for instructions. The data memory
is 16-bit containing 128 memory slots and the MAU of data memory is 16-bits. The instruction memory
has 1024 memory slots which means that the maximum number of instructions of 1024.
Register files
The machine contains 4 register files, each of which have 4 16-bit registers, leading to total of 16 16-bit
registers. The first register file has 2 read ports.
Transport buses
The machine has 3 16-bit buses, which means maximum of 3 concurrent transports. Each bus can contain
a 8-bit short immediate.
Immediates
Because the transport buses can only contain 8-bit short immediates you must use the immediate unit if
you want to use longer immediates. The immediate unit can hold a 16-bit immediate. There is an example
of immediate unit usage in file ’immediate_example.tceasm’. Basically you need to transfer the value to
the immediate register. The value of immediate register can be read on the next cycle.
The initial input data is written to memory locations 0-63 in the file ’assembler_tutorial.tceasm’. Write
your assembly code in that file.
3.5.4.1 Verifying the assembly program
The reference output is given in ’reference_output’. You need to compare your assembly program’s
simulation result to the reference output. Comparision can be done by first dumping the memory contents
in the TCE simulator with following command:
x /a IODATA /n 64 /u b 0
The command assumes that output data is stored to memory locations 0-63.
The easiest way to dump the memory into a text file is to execute ttasim with the following command:
ttasim -a dct_8x8_16_bit_with_sfus.adf -p assembler_tutorial.tpef < input_command.txt
> dump.txt
version 31 2016-02-23 35/145
TTA-based Co-design Environment v1.14 User Manual
After this you should use sed to divide the memory dump into separete lines to help comparison between
your output and the reference output. Use the following command to do this (there is an empty space
between the first two slashes of the sed expression):
cat dump.txt | sed ’s/ /
\
n/g’ > output.txt
And then compare the result with reference:
diff -u output.txt reference_output
When the TCE simulator memory dump is the same as the reference output your assembly code works and
you have completed this tutorial. Of cource you might wish to improve your assembly code to minimize
cycle count or/and instruction count.
You should also compile the C program and run it because it gives more detailed information which can
be used as reference data if you need to debug your assembly code.
To compile the C code, enter:
gcc -o c_version dct_8x8_16_bit_with_sfus.c
If you want the program to print its output to a text file, you can use the following command:
./c-version > output.txt
To get some idea of the performance possibilities of the machine, one assembly code has 52 instructions
and it runs the DCT8x8 in 3298 cycles.
3.6 Using multiple memories from C code
TCE supports accessing multiple address spaces from C code via the __attribute__((address_space(N))
extension. The numerical id from the attribute is used to connect the memory reference to the ADF
address space using the XML element numerical-id. There can be one or more numerical ids mapped to
a single ADF address space.
The following is from an ADF with three separate data address spaces with numerical ids 0, 1, and 2. The
numerical id 0 is reserved for the default data address space which stores the stack, heap and the global
constant data.
<address-space name="data">
<width>8</width>
<min-address>0</min-address>
<max-address>16777215</max-address>
<numerical-id>0</numerical-id>
</address-space>
<address-space name="data1">
<width>8</width>
<min-address>0</min-address>
<max-address>65535</max-address>
<numerical-id>1</numerical-id>
</address-space>
<address-space name="data2">
<width>8</width>
<min-address>0</min-address>
<max-address>16383</max-address>
<numerical-id>2</numerical-id>
</address-space>
The address spaces can be referred to from the C program with code like this:
#include <stdio.h>
36/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
#define ASIZE 4
volatile int space0[ASIZE] __attribute__((address_space(0)));
volatile int space0_1[ASIZE] __attribute__((address_space(0)));
volatile int space1[ASIZE] __attribute__((address_space(1)));
volatile int space1_1[ASIZE] __attribute__((address_space(1)));
volatile int space2[ASIZE] __attribute__((address_space(2)));
volatile int space2_1[ASIZE] __attribute__((address_space(2)));
int main() {
int i = 0;
/* The start addresses of the tables should overlap as
they are allocated in different address spaces. */
iprintf("space0@%x space0_1@%x space1@%x space1_1@%x space2@%x space2_1@%x\n",
(unsigned)space0, (unsigned)space0_1, (unsigned)space1, (unsigned)space1_1,
(unsigned)space2, (unsigned)space2_1);
return 0;
}
The effect of having multiple disjoint address spaces is visible when simulating the code as the arrays
allocated to the different memories share the same addresses but different storage:
space0@a84 space0_1@a94 space1@4 space1_1@14 space2@4 space2_1@14
The space0 array is stored in the default address space. In the beginning of the default address space there
is usually some global data for some C library functions such as the iprintf or the emulation library, thus
ending up with the start address of 0xa84 for the array0. For the extra address spaces 1 and 2, the arrays
start from 4 which is the first valid storage address in those address spaces. Storing data to address 0 is
not done to avoid problems with NULL pointers pointing to valid data (in this case space1 and space2
would have been stored at address 0 = NULL).
When using custom operation calls to manually execute operations that access memory on processors that
have multiple data address spaces, it should be done using _TCEAS_OPERATIONNAME macro instead
of using _TCE_OPERATIONNAME macro. See section 5.1.2 for more information about these custom
operation call macros.
3.7 Running TTA on FPGA
This tutorial illustrates how you can run your TTA designs on a FPGA board. Tutorial consists of two
simple example sections and a more general case description section.
Download the tutorial file package from:
http://tce.cs.tut.fi/tutorial_files/tce_tutorials.tar.gz
Unpack it to a working directory and cd to tce_tutorials/fpga_tutorial
3.7.1 Simplest example: No data memory
3.7.1.1 Introduction
This is the most FPGA board independent TTA tutorial one can make. The application is a simple led
blinker which has been implemented using register optimized handwritten TTA assembly. In other words
the application doesn’t need a load store unit so there is no need to provide data memory. In addition the
instruction memory will be implemented as a logic array.
version 31 2016-02-23 37/145
TTA-based Co-design Environment v1.14 User Manual
3.7.1.2 Application
The application performs a 8 leds wide sweep in an endless loop. Sweep illuminates one led at a time and
starts again from first led after reaching the last led. There is also a delay between the iterations so that
the sweep can be seen with human eye.
As stated in the introduction the application is coded in assembly. If you went through the assembly
tutorial the code is probably easy to understand. The code is in file ’blink.tceasm’. The same application
is also written in C code in file ’blink.c’.
3.7.1.3 Create TTA processor core and instruction image
The architecture we’re using for this tutorial is ’tutorial1.adf’. Open it in ProDe to take a look at it:
prode tutorial1.adf
As you can see it is a simple one bus architecture without a LSU. There are also 2 “new” function
units: rtimer and leds. Rtimer is a simple tick counter which provides real time clock or countdown
timer operations. Leds is function unit that can write ’0’ or ’1’ to FPGA output port. If those ports are
connected to leds the FU can control them.
Leds FU requires a new operation definition and the operation is defined in ’led.opp’ and ’led.cc’. You
need to build this operation defintion:
buildopset led
Now you can compile the assembly code:
tceasm -o asm.tpef tutorial1.adf blink.tceasm
If you wish you can simulate the program with proxim and see how it works but the program runs in
endless loop and most of the time it stays in the “sleep” loop.
Now you need to select implementations for the function units. This can be done in ProDe. See TCE
tour section 3.1.9 for more information. Implementations for leds and rtimer are found from the fpga.hdb
shipped with the tutorial files. Notice that there are 2 implementations for the rtimer. ID 3 is for 50 MHz
clock frequency and ID 4 for 100 MHz. All other FUs are found from the default hdb.
Save the implementation configuration to ’tutorial1.idf’.
Next step is to generate the VHDL implementation of the processor:
generateprocessor -i tutorial1.idf -o asm_vhdl/proge-output tutorial1.adf
Then create the proram image:
generatebits -f vhdl -p asm.tpef -x asm_vhdl/proge-output tutorial1.adf
Notice that the instruction image format is “vhdl” and we request generatebits to not create data image at
all. Now, move the generated ’asm_imem_pkg.vhdl’ to the asm_vhdl directory and cd there.
mv asm_imem_pkg.vhdl asm_vhdl/
cd asm_vhdl
3.7.1.4 Final steps to FPGA
We have successfully created the processor core and instruction memory image. Now we need an instruc-
tion memory component that can use the generated image. Luckily you don’t have to create it as it is
shipped with the tutorial files. The component is in file ’inst_mem_logic.vhd’ in asm_vhdl directory and
it can use the generated ’asm_imem_pkg.vhdl’ without any modifications.
Next step is to connect TTA toplevel core to the memory component and connect the global signals out
from that component. This has also been done for you in file ’tutorial_processor1.vhdl’. If you are curious
how this is done open the file with your preferred text editor. All the signals coming out of this component
are later connected to FPGA pins.
Now you need to open your FPGA tool vendor’s FPGA design/synthesis program and create a new project
for your target FPGA. Add the three files in asm_vhdl-directory (toplevel file ’tutorial_processor1.vhdl’,
38/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
’inst_mem_logic.vhd’ and ’asm_imem_pkg.vhdl’) and all the files in proge-output/gcu_ic/ and proge-
output/vhdl directories to the project. The toplevel entity name is ’tutorial_processor1’.
Then connect the toplevel signals to appropriate FPGA pins. The pins are most probably described in
the FPGA board’s user manual. Signal ’clk’ is obviously connected to the pin that provides clock signal.
Signal ’rstx’ is the reset signal of the system and it is active low. Connect it to a switch or pushbutton
that provides ’1’ when not pressed. Signal bus ’leds’ is 8 bits wide and every bit of the bus should be
connected to an individual led. Don’t worry if your board doesn’t have 8 user controllable leds, you can
leave some of them unconnected. In that case all of the leds are off some of the time.
Compile and synthesize your design with the FPGA tools, program your FPGA and behold the light
show!
3.7.2 Second example: Adding data memory
In this tutorial we will implement the same kind of system as above but this time we include data memory
and use C coded application. Application has the same functionality but the algorithm is a bit different.
This time we read the led pattern from a look up table and to also test store operation the pattern is stored
back to the look up table. Take a look at file ’blink_mem.c’ to see how the timer and led operations are
used in C code.
3.7.2.1 Create TTA processor core and binary images
The architecture for this tutorial is ’tutorial2.adf’. This architecture is the same as ’tutorial1.adf’ with the
exception that now it has a load store unit to interface it with data memory.
You need to compile the operation behaviour for the led function unit if you already haven’t done it:
buildopset led
Then compile the program:
tcecc -O3 -a tutorial2.adf -o blink.tpef blink_mem.c
Before you can generate processor vhdl you must select implementations for the function units. Open the
architecture in ProDe and select Tools->Processor Implementation...
prode tutorial2.adf
It is important that you choose the implementation for LSU from the fpga.hdb shipped with the tutorial
files. This implementation has more FPGA friendly byte enable definition. Also the implementations for
leds and timer FUs are found from fpga.hdb. As mentioned in the previous tutorial, timer implementation
ID 3 is meant for 50 MHz clock frequency and ID 4 for 100 MHz clock. Other FUs are found from the
default hdb.
Generate the processor VHDL:
generateprocessor -i tutorial2.idf -o c_vhdl/proge-output tutorial2.adf
Next step is to generate binary images of the program. Instruction image will be generated again as
a VHDL array package. But the data memory image needs some consideration. If you’re using an
Altera FPGA board the Program Image Generator can output Altera’s Memory Initialization Format
(mif). Otherwise you need to consult the FPGA vendor’s documentation to see what kind of format is
used for memory instantiation. Then select the PIG output format that you can convert to the needed
format with the least amount of work. Of course you can also implement a new image writer class to
PIG. Patches are welcome.
Image generation command is basically the following:
generatebits -f vhdl -d -w 4 -o mif -p blink.tpef -x c_vhdl/proge-output tutorial2.adf
Switch ’-d’ tells PIG to generate data image. Switch ’-o’ defines the data image output format. Change
it to suit your needs if necessary. Switch ’-w’ defines the width of data memory in MAUs. By default
MAU is assumed to be 8 bits and the default LSU implementations are made for memories with 32-bit
data width. Thus the width of data memory is 4 MAUs.
Move the created images to the vhdl directory:
version 31 2016-02-23 39/145
TTA-based Co-design Environment v1.14 User Manual
mv blink_imem_pkg.vhdl c_vhdl/
mv blink_data.mif c_vhdl/
3.7.2.2 Towards FPGA
Go to the vhdl directory:
cd c_vhdl
TTA vhdl codes are in the proge-output directory. Like in the previous tutorial file ’inst_mem_logic.vhd’
holds the instruction memory component which uses the created ’blink_imem_pkg.vhdl’. File ’tuto-
rial_processor2.vhdl’ is the toplevel design file and again the TTA core toplevel is connected to the in-
struction memory component and global signals are connected out from this design file.
Creating data memory component
Virtually all FPGA chips have some amount of internal memory which can be used in your own designs.
FPGA design tools usually provide some method to easily create memory controllers for those internal
memory blocks. For example Altera’s Quartus II design toolset has a MegaWizard Plug-In Manager
utility which can be used to create RAM memory which utilizes FPGA’s internal resources.
There are few points to consider when creating a data memory controller:
1. Latency. Latency of the memory should be one clock cycle. When LSU asserts a read command
the result should be readable after one clock cycle. This means that the memory controller shouldn’t
register the memory output because the registering is done in LSU. Adding an output register would
increase read latency and the default LSU wouldn’t work properly.
2. Address width. As stated before the minimal addressable unit from the TTA programmer’s point
of view is 8 bits by default. However the width of data memory bus is 32 bits wide in the default
implementations. This also means that the address bus to data memory is 2 bits smaller because it
only needs to address 32-bit units. To convert 8-bit MAU addresses to 32-bit MAU addresses one
needs to leave the 2 bits out from LSB side.
How this all shows in TCE is that data memory address width defined in ADF is 2 bits wider than
the actual address bus coming out of LSU. When you are creating the memory component you
should consider this.
3. Byte enable. In case you were already wondering how can you address 8-bit or 16-bit wide areas
from a 32-bit addressable memory the answer is byte enable (or byte mask) signals. These signals
can be used to enable individual bytes from 32-bit words which are read from or written to the
memory. And those two leftover bits from the memory address are used, together with the memory
operation code, to determine the correct byte enable signal combination.
When you are creating the memory controller you should add support for byte enable signals.
4. Initialization. Usually the internal memory of FPGA can be automatically initialized during FPGA
configuration. You should find an option to initialize the memory with a specific initialization file.
Connecting the data memory component
Next step is to interface the newly generated data memory component to TTA core. LSU interface is the
following:
fu_lsu_data_in : in std_logic_vector(fu_lsu_dataw-1 downto 0);
fu_lsu_data_out : out std_logic_vector(fu_lsu_dataw-1 downto 0);
fu_lsu_addr : out std_logic_vector(fu_lsu_addrw-2-1 downto 0);
fu_lsu_mem_en_x : out std_logic_vector(0 downto 0);
fu_lsu_wr_en_x : out std_logic_vector(0 downto 0);
fu_lsu_bytemask : out std_logic_vector(fu_lsu_dataw/8-1 downto 0);
Meanings of these signals are:
40/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
Signal name Description
fu_lsu_data_in Data from the memory to LSU
fu_lsu_data_out Data from LSU to memory
fu_lsu_addr Address to memory
fu_lsu_mem_en_x Memory enable signal which is active low. LSU asserts this signal to ’0’ when
memory operations are performed. Otherwise it is ’1’. Connect this to memory
enable or clock enable signal of the memory controller.
fu_lsu_wr_en_x Write enable signal which is active low. During write operation this signal
is ’0’. Read operation is performed when this signal ’1’. Depending on the
memory controller you might need to invert this signal.
fu_lsu_bytemask Byte mask / byte enable signal. In this case the signal width is 4 bits and each
bit represents a single byte. When the enable bit is ’1’ the corresponding byte
is enabled and value ’0’ means that the byte is ignored.
Open file ’tutorial_processor2.vhdl’ with your preferred text editor. From the comments you can see
where you should add the memory component declaration and component instantiation. Notice that those
LSU signals are connected to wires (signals with appendix ’_w’ in the name). Use these wires to connect
the memory component.
Final steps
After you have successfully created the data memory component and connected it you should add the
rest of the design VHDL files to the design project. All of the files in proge-output/gcu_ic/ and proge-
output/vhdl/ directories need to be added.
Next phase is to connect toplevel signals to FPGA pins. Look at the final section of the previous tutorial
for more verbose instructions how to perform pin mapping.
Final step is to synthesize the design and configure the FPGA board. Then sit back and enjoy the light
show.
3.7.2.3 More to test
If you simulate the program you will notice that the program uses only STW and LDW operations. Reason
for this can be easily seen from the source code. Open ’blink_mem.c’ and you will notice that the look
up table ’patterns’ is defined as ’volatile unsigned int’. If you change this to ’volatile unsigned char’ or
’volatile unsigned short int’ you can test STQ and LDQU or STH and LDHU operations. Using these
operations also means that the LSU uses byte enable signals.
Whenever you change the source code you need to recompile your program and generate the binary
images again. And move the images to right folder if it’s necessary.
In addition you can compile the code without optimizations. This way the compiler leaves function calls
in place and uses stack. The compilation command is then:
tcecc -O0 -a tutorial2.adf -o blink.tpef blink_mem.c
3.8 How to print from Altera FPGAs
TCE comes with a special function unit which enables character output from Altera FPGAs. This SFU
utilizes Altera JTAG UART IP core and nios2-terminal host program for this functionality. Neither of
them are included in TCE so you need the appropriate Altera software and libraries (namely Quartus II
and Nios II EDS) in order to use the printing capability.
3.8.1 Hello World 2.0
If you haven’t already downloaded the tutorial file package, you should download it now from:
http://tce.cs.tut.fi/tutorial_files/tce_tutorials.tar.gz
Then unpack it to a working directory and cd to tce_tutorials/fpga_stdout
version 31 2016-02-23 41/145
TTA-based Co-design Environment v1.14 User Manual
Let’s begin by examining our tutorial architecture. Open the given test.adf architecture in ProDe:
prode test.adf &
As you can see, the architecture includes IO function unit which implements the STDOUT operation. In
addition the architecture also has a timer function unit which we will also be used in this tutorial.
3.8.1.1 Examine the first version
Open source code file ’std_print.c’ in your preferred text editor. As you can see, the code includes ’stdio.h’
and uses printf() for printing “Hello World!”. Furthermore, operation RTC is used to measure how long
the printing takes and this time is then printed at the end of the program. Writing value “0” to the RTC
resets the real time clock in timer FU. When the RTC is invoked with a non-zero value, the current time is
written to the given variable (in this case to variable “timestamp”). RTC counts the time in milliseconds
and by default, the instruction set simulator assumes clock frequency of 100 MHz.
Now compile the source code with the following command:
tcecc -O0 --swfp -a test.adf -o std_print.tpef std_print.c
The compilation command contains a few flags that should be explained in more detail. Let’s first look
into the --swfp flag: this tells the compiler to link the program with the floating point emulation library.
Floating point support is needed because printf() function includes support for printing floating point
values and our architecture does not contain floating point function units. To emphasize the lesson of
this tutorial it is important that you compile the code without optimizations i.e. with -O0 flag. Otherwise
our witty compiler will optimize the first printf() call (which just prints a constant character string) into
inlined _TCE_STDOUT operations.
After compiling, execute the program in the instruction set simulator:
ttasim -a test.adf -p std_print.tpef
You should see the printed text in your terminal. By the time of writing this tutorial, the duration of the
“Hello World” printing with printf() took 61 ms (time may vary depending on the TCE version).
But speaking of the lesson, execute the following command to examine the minimum memory consump-
tion of this program:
dumptpef -m std_print.tpef
As you should notice, the instruction count is high (around 45 000 at the time of writing this tutorial) on
this simple sequential TTA architecture with optimizations disabled. Major part of the instructions are
spent on the printf() and floating point emulation functions.
First step in reducing the instruction count is to use iprintf() instead of printf(). Function iprintf() is
a simplified version of printf() which drops the support for printing floating points. In our test case we
don’t need to print floats, so open the source code file ’std_print.c’ in your preferred text editor and change
the two printf() calls to iprintf(). Alternatively, you can do this with program sed:
sed -i ’s/printf/iprintf/g’ std_print.c
Now recompile the program and check the instruction count:
tcecc -O0 -a test.adf -o std_print.tpef std_print.c
dumptpef -m std_print.tpef
You should see a significant drop in the instruction count. If you simulate the new program, you notice
no difference in the behavior (except that the measured time might be a bit lower).
3.8.1.2 Light weight printing
TCE includes a Light Weigth PRinting (lwpr) library to provide small and simple functions for printing
strings and integers for further reducing the instruction count overhead of print support. Take a look at
the source code file ’lightweight_print.c’ to see how these functions are used. The library is included
with header ’lwpr.h’. Function lwpr_print_str() is utilized to output strings and function lwpr_print_int()
is used for printing integers. There is also function for printing integers in hexadecimal format called
lwpr_print_hex(), but it is not used in this tutorial.
42/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
Compile the new code to see the difference of using lwpr:
tcecc -O0 -llwpr -a test.adf -o lightweight_print.tpef lightweight_print.c
Notice the new compilation flag -llwpr for including the light weight printing library.
First, check the instruction count with:
dumptpef -m lightweight_print.tpef
You should notice that the instruction count has dropped dramatically and also the initialized data memory
is a lot smaller that previously.
Next, simulate program:
ttasim -a test.adf -p lightweight_print.tpef
Printed text is the same as previously except that the measured duration has dropped significantly.
3.8.2 FPGA execution
Next step is to get the program running on an FPGA. Prerequisite for this step is that you have Altera
Quartus II and nios2-terminal programs installed and you have a Stratix II DSP FPGA board. But don’t
worry if you don’t have this specific board, this tutorial can be completed with other Altera FPGA boards
as well. In case you are using alternative board you must do a few manual changes before synthesizing
and executing the design. Disclaimer: we assume no liability in case you fry your FPGA :)
3.8.2.1 Preparations
Before starting to generate the processor we first must adjust the address space sizes of the architecture.
In order to do this, open the architecture in ProDe and open the address space dialog (Edit -> Address
Spaces...).
prode test.adf &
Adjust both the data and instruction address space to be 10 bits wide. This should be enough for our
application according to the dumptpef output.
Next recompile the application so it adapts to the new address space sizes:
tcecc -O0 -llwpr -a test.adf -o lightweight_print.tpef lightweight_print.c
If you wish you can execute the program with ttasim to verify that it still works after the address space
sizes changed.
Before generating the processor, you must also select the RF and FU implementations. If you wish to
skip this step, you can use the given IDF by renaming it:
mv preselected.idf test.idf
Then move on to section 3.8.2.2. If you choose to learn and do this step manually, keep following the
instructions. In case you don’t already have the architecture open, do it now:
prode test.adf &
Then select Tools -> Processor Implementation... to open the implementation dialog. First select the
Register Files implementations from the ’asic_130nm_1.5V.hdb’. It doesn’t matter which implementation
you choose for this tutorial.
After selecting RFs, click open the Function Unit tab. Now you must choose the implementations care-
fully:
1. LSU: Change the HDB to ’stratixII.hdb’ (don’t worry if you are using another Altera FPGA) and
select the fu_lsu_with_bytemask_always_3.
2. ALU: Change the HDB back to ’asic_130nm_1.5V.hdb’ and select any available implemenation.
3. IO: Change the HDB to ’altera_jtag_uart.hdb’ and select altera_jtag_uart_stdout_always_1 as the
implementation.
version 31 2016-02-23 43/145
TTA-based Co-design Environment v1.14 User Manual
4. TIMER: Change the HDB again to ’stratixII.hdb’ and select the timer implementation with id 5.
This implementation is for 100 MHz clock frequency (you can check it by opening the stratixII.hdb
with hdbeditor and examining the generic parameters).
Rememeber to save IDF before closing the dialog.
3.8.2.2 Generate the processor
Now we will use the Platform Integrator feature (see section 4.6 for more information) of ProGe. Execute
the following command to generate the processor implementation:
generateprocessor -i test.idf -o proge-out -g Stratix2DSP -d onchip -f onchip
-p lightweight_print.tpef -e hello_tta test.adf
The new flag -g commands to use Stratix II DSP board Platform Integrator and -d and -f tells the Platform
Integrator to implement instruction and data memory as FPGA onchip memory. Flag -p defines the name
of the program we want to execute on the processor (it is just needed for naming the memory initialization
files) and flag -e defines the toplevel entity name for our processor.
Next, generate the memory images from the tpef program:
generatebits -d -w 4 -f mif -o mif -x proge-out -p lightweight_print.tpef -e hello_tta
test.adf
Important flags to notice here are the -x which tells where the HDL files generated by ProGe are stored
and -e which defines the toplevel entity name (must be the same you gave to ProGe). Data and instruction
memory image formats are defined as mif (Altera’s memory initialization format for onchip memories).
3.8.2.3 Modifications for using alternative Altera FPGA boards
If you are using the Stratix II DSP FPGA board you can skip this section. Otherwise you must complete
the tasks described here to get the processor running on your alternative FPGA.
HDL changes Open file ’proge-out/platform/hello_tta_toplevel.vhdl’ in your preferred text editor. Change
the value of the toplevel generic dev_family_g according to your FPGA. For example if you are using Al-
tera DE2 board with a Cyclone II FPGA, change the string from “Stratix II” to “Cyclone II”. Save the file
before exit.
Changes to Quartus II project files Open file ’hello_tta_toplevel.qsf’ in your preferred text editor.
You must locate and change the following settings according to the FPGA board you are using (all the
examples are given for the DE2 board for illustrative purposes):
1. FAMILY: Set the FPGA device family string according to your FPGA. Example: “Cyclone II”
2. DEVICE: Set the specific FPGA device name according to your FPGA. Example: EP2C35F672C6
3. FMAX_REQUIREMENT: Set the target clock frequency according to the oscillator you are using
to drive clock signal. Example: “50 MHz”
4. Pin assignments: Change the pin assignments for clk and rstx accorind to your FPGA board. Reset
signal rstx is active low so take this into consideration in the pin mapping. Example: PIN_N2 -to
clk (50 MHz oscillator to clk signal) and PIN_G26 -to rstx (push button KEY0 to rstx)
3.8.2.4 Synthesize and execute
Synthesize the processor simply by executing the generated script:
./quartus_synthesize.sh
Assuming that your FPGA is set up, on and properly connected via USB Blaster to your PC, you can
program the FPGA with the following script:
44/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
./quartus_program_fpga.sh
Or if you prefer, you can also use the graphical Quartus programmer tool for FPGA programming.
After the programmer has finished, open nios2-terminal program in order to capture the printed charac-
ters:
nios2-terminal -i 0
After the connection is open and program starts to run, you will the see the characters printed to the
terminal. Notice that the measured printing time may vary according to your FPGA board and clock
frequency as the timer implementation was set for 100 MHz clock.
3.8.3 Caveats in printing from FPGA
There are few caveats in printing from FPGA you should be aware of. First of all the transfer speed
between the host PC and FPGA is finite and in order to avoid extra stalls the STDOUT FU uses a character
buffer. When this buffer is full, the TTA processor pipeline is stalled until there is space again in the buffer.
This happens especially when the nios2-terminal is not open i.e. there’s no host process to clear the buffer.
In otherwords, if your application uses STDOUT you must open nios2-terminal with the correct instance
number to avoid the execution getting stuck.
Because of the stalling behavior you should avoid print during profiling as the results may be affected.
You should not print during the execution of timing critical code as the real time characteristics cannot be
guaranteed due to the possibility of stalls.
3.8.4 Summary
This tutorial illustrated how to print from Altera FPGAs. In addition this tutorial discussed different ways
to print from C code and demonstrated their impact on the instruction count.
For demonstrational purposes all the compilations were done without optimizations. However, in case
optimizations are applied, printf() or iprintf() may sometimes produce the smallest code overhead when
printing only constant strings. This is possible when the compiler is able to reduce the function call into
direct _TCE_STDOUT macro calls and fill NOP slots with the STDOUT operations. But in general it is
advisable to use light weight printing library functions whenever applicable.
3.9 Designing Floating-point Processors with TCE
TCE supports single and half precision floating-point calculations. Single-precision calculations can be
performed by using the float datatype in C code, or by using macros from tceops.h, such as _TCE_ADDF.
If the compilation target architecture does not support these operations, they can be emulated using integer
arithmetic in software. Passing the switch –swfp to tcecc enables the software emulation library linkage.
A set of floating-point FU implementations is included with TCE, in a HDB file named fpu_embedded.hdb,
which can be found at PREFIX/share/tce/hdb/fpu_embedded.hdb. The FUs operate with 32-bit, single-
precision floating point numbers. Supported operations include addition, subtraction, negation, absolute
value, multiplication, division, square root, conversion between floats and integers, and various compar-
isons.
The FUs are based on the VHDL-2008 support library (http://www.vhdl.org/fphdl/), which is in public
domain. Changes include:
•Full pipelining.
•Radix-2 division changed to Radix-4.
•Simple newton’s iteration square root (with division in each pass) replaced by Hain’s algorithm
from paper "Fast Floating Point Square Root" by Hain T. and Mercer D.
version 31 2016-02-23 45/145
TTA-based Co-design Environment v1.14 User Manual
The FUs are optimized for synthesis on Altera Stratix II FPGA’s, and they have been benchmarked both
on a Stratix II EP2S180F1020C3, and a Stratix III EP3SL340H1152C2. They have maximum frequencies
between 190-200 MHz on the Stratix II, and between 230-280 MHz on the Stratix III. Compared to an
earlier implementation based on the Milk coprocessor (coffee.cs.tut.fi), they are between 30% and 200%
faster.
3.9.1 Restrictions
The FUs are not IEEE compliant, but instead comply to the less strict OpenCL Embedded Profile standard,
which trades off accuracy for speed. Differences include:
•Instead of the default rounding mode round-to-nearest-even, round-to-zero is used.
•Denormal numbers as inputs or outputs are flushed to zero.
•Division may not be correctly rounded, but should be accurate within 4 ulp.
The TCE Processor Simulator uses IEEE-compliant floats. With a processor simulated on GHDL or
synthesized on actual hardware, the calculation results are thus slightly different from the ones from
Processor Simulator.
3.9.2 Single-precision Function Units
The emphfpu_embedded and emphfpu_half function units are described in detail below.
fpu_sp_add_sub Supported operations: addf, subf
Latency: 5
Straightforward floating-point adder.
fpu_sp_mul Supported operations: mulf
Latency: 5
Straightforward floating-point multiplier.
fpu_sp_div Supported operations: divf
Latency: 15 (mw/2+3)
Radix-4 floating-point divider.
fpu_sp_mac Supported operations: macf, msuf
Latency: 6
Single-precision fused multiply-accumulator.
Parameters are ordered so that MACF(a,b,c,d) is equal to d=a+b*c and MSUF(a,b,c,d) to d=a-b*c.
Special case handling is not yet supported.
fpu_sp_mac_v2 Supported operations: macf, msuf, addf, subf, mulf
Latency: 6
Single-precision fused multiply-accumulator. Performs addition/subtraction by multiplying by 1,
and multiplication by adding 0. fpu_sp_mac_v2 will replace fpu_sp_mac completely if bench-
marking shows it to be reasonably fast.
Parameters are ordered so that MACF(a,b,c,d) is equal to d=a+b*c and MSUF(a,b,c,d) to d=a-b*c.
fpu_sp_sqrt Supported operations: sqrtf
Latency: 26 (mw+3)
Floating-point square root FU, using Hain’s algorithm.
Note that the C standard function sqrt does not take advantage of hardware acceleration; the
_TCE_SQRTF macro must be used instead.
46/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
fpu_sp_conv Supported operations: cif, cifu, cfi, cfiu
Latency: 4
Converts between 32-bit signed and unsigned integers, and single-precision floats. OpenCL em-
bedded allows no loss of accuracy in these conversions, so rounding is to nearest even.
fpu_sp_compare Supported operations: absf, negf, eqf, nef, gtf, gef, ltf, lef
Latency: 1
A floating-point comparator. Also includes the cheap absolute value and negation operations.
3.9.3 Half-precision Support
A set of half-precision arithmetic units is included with tce in PREFIX/share/tce/hdb/fpu_half.hdb. In C
and C++, half-precision operations an only be invoked with tceops.h macros. It may be helpful to define
ahalf class with overloaded operators to wrap the macros. The test case testsuite/systemtest/proge/hpu is
written using such a class. There is ongoing work to add acceleration for the half datatype in OpenCL.
Like their single-precision counterparts, the half-precision FPUs round to zero and lack support for de-
normal numbers. In addition, they do not yet handle special cases such as INFs and NaNs.
3.9.4 Half-precision Function Units
The emphfpu_half function units are described in detail below.
fpu_chf_cfh Supported operations: cfh, chf
Latency: 1
Converter between half-precision and single-precision floating points.
fpadd_fpsub Supported operations: addh, subh
Latency: 1
Straightforward half-precision floating-point adder.
fpmul Supported operations: mulh, squareh
Latency: 2
Straightforward half-precision floating-point multiplier. Also supports a square-taking operation.
fpmac Supported operations: mach, msuh
Latency: 3
Half-precision fused multiply-accumulator.
Parameters are ordered so that MACH(a,b,c,d) is equal to d=a+b*c and MSUH(a,b,c,d) to d=a-b*c.
fpmac_v2 Supported operations: mach, msuh, addh, subh, mulh
Latency: 3
Half-precision fused multiply-accumulator. Performs addition/subtraction by multiplying by 1, and
multiplication by adding 0. fpmac_v2 will replace fpmac completely if benchmarking shows it to
be reasonably fast.
Parameters are ordered so that MACH(a,b,c,d) is equal to d=a+b*c and MSUH(a,b,c,d) to d=a-b*c.
fp_invsqrt Supported operations: invsqrth
Latency: 5
Half-precision fast inverse square root using Newton’s iteration.
fpu_hp_compare Supported operations: absh, negh, eqh, neh, gth, geh, lth, leh
Latency: 1
Half-precision floating-point comparator. Also includes the absolute value and negation operations.
version 31 2016-02-23 47/145
![](asset-30.png)
TTA-based Co-design Environment v1.14 User Manual
mul add_sub sqrt conv comp div baseline
Comb ALUTs 1263 1591 4186 1500 1012 2477 907
Total regs 892 967 2444 917 669 1942 567
DSP blocks 8 0 0 0 0 0 0
Fmax(MHz)196.39 198.81 194.78 191.5 192.2 199.32 222.82
Latency 5 5 26 4 1 15 -
Table 3.2: Synthesis results for Stratix II EP2S180F1020C3
mul add_sub sqrt conv comp div baseline
Comb ALUTs 1253 1630 4395 1507 1002 2597 1056
Total regs 819 1007 2401 997 665 2098 710
DSP blocks 4 0 0 0 0 0 0
Fmax(MHz)272.03 252.4 232.07 232.88 244.32 260.82 286.45
Latency 5 5 26 4 1 15 -
Table 3.3: Synthesis results for Stratix III EP3SL340H1152C2
3.9.5 Benchmark results
Most of the single-precision FPUs have been benchmarked on the FPGAs Stratix II EP2S180F1020C3
and Stratix III EP3SL340H1152C2. As a baseline, a simple TTA processor was synthesized that had
enough functionality to support an empty C program. After this, each of the FPUs was added to the
baseline processor and synthesized. The results are shown below in Tables 3.2 and 3.3.
3.9.6 Alternative bit widths
The fpu_embedded Function Units have mantissa width and exponent width as generic parameters, so
they can be used for float widths other than the IEEE single precision. The FPUs are likely prohibitively
slow for double-precision computation, and the fpu_half units should be better fit for half-precision.
The parameters are mw (mantissa width) and ew (exponent width) for all FUs. In addition, the float-int
converter FU fpu_sp_conv has a parameter intw, which decides the width of the integer to be converted.
Use of these parameters has the following caveats:
•The TCE-CC compiler converts floating-point literals into 32-bit floats, so they have to be entered
some other way, f.ex. by casting integer bitpatterns to floats, or with a cif operation.
•TCE does not include a HDB file for alternative bit widths
•Mantissa width affects the latency of the divider and square root FUs. The divider FU’s latency is
(mw/2) + 3, and the square root FU’s latency is mw +3.
•Bit widths other than single-precision have not been exhaustively tested. Half-precision floats
appear to work in a simple test case.
3.9.7 Processor Simulator and Floating Point Operations
Designers of floating point TTAs should note that ttasim uses the simulator host’s floating point (FP)
hardware to simulate floating point operations (for speed reasons). Thus, it might or might not match the
FP implementation of the actual implemented TTA as it depends on the standard compliance, the default
rounding modes, and other differences between floating point implementations.
48/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
3.10 Multi-TTA Designs
In order to add multiple TTA cores to your system level design, there are two useful switches in the
Processor Generator and Program Image Generator tools.
One of them is –shared-files-dir which can be used to define a directory to which HDL files that can be
shared between multiple TTAs are copied. This avoids name clashes, e.g. on FU names when two or
more TTAs use the same implementation.
The second useful switch is –entity-name. This switch allows naming the generated core in the HDL
files with a custom name. The name string is also inserted to all the core-specific component names to
make them unique to avoid name clashes in HDL compilation of the system. It should be noted that the
Program Image Generator (generatebits) also supports the –entity-name switch and the same entity name
should be given for both tools to produce functioning TTA cores.
Using these switches, generating cores to designs with multiple TTAs can be done by using a shared
HDL directory for all the cores (–shared-files-dir) and then naming each core with a different entity name
(–entity-name) to avoid name clashes. How the cores are connected and the generation of the shared
memory hierarchy is currently out of scope of TCE. While these switches assist in HDL generation of
multi-TTA designs, the designer can also simulate multi-TTA designs at system level by using the TCE
SystemC API (see Section 6.2).
3.11 OpenCL Support
TCE has experimental support for running both OpenCL C kernels and the host runtime in a same stati-
cally linked program. This can be used to benefit from parallelizable kernels written in OpenCL C without
requiring a host processor with an OpenCL C compiler.
The OpenCL support uses the Portable OpenCL (pocl) project. In order to add OpenCL support to tcecc,
you should download and install the latest version of pocl. At the end of the configure run, the script
mentions whether the OpenCL support was enabled or not. After building TCE You can check if the
OpenCL C support is enabled via “tcecc –supported-languages” switch.
It must be emphasized that the OpenCL support is a work in progress and does not provide full support
for the standard. The missing APIs are implemented “as needed”.
The “statically compiled OpenCL C” support works by relying on the reqd_work_group_size kernel at-
tributes. It uses the work group dimensions defined with the attributes to statically parallelize multiple
work items (WI) in a single work group (WG) to produce highly parallel code for the compiler instruction
scheduler.
For example: the dot_product example must be written as follows to produce four parallel WIs per WG:
/* dot.cl */
__attribute__((reqd_work_group_size(4, 1, 1)))
kernel void
dot_product (global const float4 *a,
global const float4 *b,
global float *c) {
int gid = get_global_id(0);
c[gid] = dot(a[gid], b[gid]);
}
Currently the host API must know about the WG dimensions and use the same ones when launching the
kernel, otherwise undefined behavior occurs. For the other parts, the OpenCL host runtime API can be
used similarly as in the “regular mode“ where the kernels are (or can be) compiled and linked at runtime.
This enables easier porting of OpenCL programs to the TCE standalone OpenCL mode. In the standalone
mode, the compiler invokation APIs of the OpenCL host runtime are implemented as dummy functions
to produce source code level compatibility with most of the OpenCL programs.
version 31 2016-02-23 49/145
TTA-based Co-design Environment v1.14 User Manual
For example, the host program for invoking the dot_product code can be written identically as it was done
in the original OpenCL example. The host program can invoke the compiler etc. to make the program
run in the regular OpenCL environments while in TCE standalone mode the kernel is compiled offline,
linked to the main program, and the compiler calls are no-operations.
The command line to compile both the host code and the kernel to a single program is as follows:
tcecc -a mytta.adf -O3 host.c dot.cl -o dot_product -loclhost-sa
The resulting program can be simulated and executed like any other TTA program produced by tcecc.
The oclhost-sa library provides the more-or-less dummy implementations of the host runtime APIs to
enable launching the kernel and moving the buffers etc.
3.12 System-on-a-Chip design with AlmaIF Integrator
This tutorial will cover TTA core generation and integration with a Zynq-7000 series SoC, but the work-
flow should adapt quite well for other AXI4-capable Xilinx SoCs.
3.12.1 Generate the Processor
The command to generate a processor core with an AlmaIF wrapper is as follows:
generateprocessor -d onchip -f onchip -e tta_core -i mach.idf -g AlmaIFIntegrator
-o proge-out -p program.tpef mach.adf
generatebits -e tta_core -x proge-out mach.adf
3.12.2 Create a New Vivado Project
To create a new project in Vivado, select File>New project from the menu bar. You will be presented with
a window, where you can configure the project to be created. Click Next and set the project name and
location to your preference. On the next screen, pick RTL Project and click Next. Select the proge-out
folder and verify the Include files from subdirectories checkbox is checked. No action is needed for the
next two screens, so you can click past them. Select the part you intend to use for the project from the
list, click Next and Finish.
You should now have a fresh project with the TTA sources included.
3.12.3 Create a Block Design
Navigate to Flow>Create Block Design and name it toplevel. Right click on the design, select Add IP
and add a ZYNQ7 Processing System block. Next, add the TTA cores. Again, right click on the design,
but this time choose Add block.... Since our entity name was tta_core, the AlmaIF wrapper was named
tta_core_toplevel. Add that, and repeat once.
You should now have two TTA cores and one ZYNQ7 Processing System in your block design. We will
need to connect these together. Luckily, Vivado will recognize the AXI4 bus on our AlmaIF wrapper,
and will offer to connect it for us. Right click on an empty part of the block design and select Run block
automation, Click OK. Right click on the block design again and select Run connection automation.
Check the All Automation checkbox and click OK.
After this, Vivado has picked address spaces for the cores. You can see these from the Address Editor tab.
You will need these addresses to control the cores from the hard processor cores.
3.12.4 Cleanup
You will notice that the TTA cores’ asynchronous reset has been left unconnected. We can actuate a
synchronous reset through AlmaIF, so we can connect these to a constant high signal. Right click on the
50/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
design, select Add IP and add the Constant IP block. Right click on the newly created block and select
Customize IP. Set both Width and Value to 1. Connect the output to both cores’ asynchronous reset pin.
You may customize the TTA cores’ clock frequency by right clicking on the ZYNQ7 Processing System
block and selecting Customize IP. The FPGA fabric clock frequencies can be set under Clock Configura-
tion.
3.12.5 Synthesis and Implementation
Our block design is ready, but we’ll have to create a wrapper for it before it can be synthesized. Select
Flow>Project Manager from the menu bar. You can see the toplevel block design in the sources window.
Right click on it and select Create HDL Wrapper. The default setting works for us, click OK. Right click
on the newly created toplevel_wrapper and select Set as Top.
You can now select Flow>Generate Bitstream from the menu bar. This will run the design through
synthesis and implementation, and give you a bitstream file with which you can program the FPGA
fabric on your SoC.
version 31 2016-02-23 51/145
TTA-based Co-design Environment v1.14 User Manual
Chapter 4
PROCESSOR DESIGN TOOLS
4.1 TTA Processor Designer (ProDe)
Processor Designer (ProDe) is a graphical application mainly for viewing, editing and printing processor
architecture definition files. It also allows selecting implementation for each component of the processor,
and generating the HDL implementation of the processor. The application is very easy to use and intuitive,
thus this section provides help only for the most common problematic situations encountered while using
the toolset.
Input: ADF
Output: ADF, VHDL
The main difficulty in using the tool is to understand what is being designed, that is, the limitations placed
by the processor template. Details of the processor template are described in [CSJ04].
4.1.1 Starting ProDe
Processor Desinger can simply be executed from command line with:
prode
4.1.2 Function Unit Operation Dialog
The most complex part of ProDe is the operation property dialog which pops up when editing the opera-
tions inside a function unit.
The upper right side of the dialog contains the function unit resource table. The resource table is used to
describe the pipeline resource usage of the operation inside the function unit. When the function unit is
fully pipelined, there are no limitations on executing of successive instructions, thus it should be empty
in that case.
If there is some limitation, for example pipeline component, which is used in some clock cycles of
the operation execution, so that it limits the execution of successive operations, one ore more pipeline
resources should be added to model this component. Each operation that uses the resources (that can be
arbitrarily named), should include timing information by marking the clock cycles the resource is used
from the start of the operation.
For example, in a multiplication function unit, which has a 4-cycle latency, and is not pipelined at all;
the resource called, for example, “multiplier” should be created, by pressing the add button. All the
operations which use the multiplier component should have the first 4 clock cycles marked to use this
resource in the resource usage table. Pressing the “add” button adds row to the resource usage table.
The lower right side of the dialog contains the operand I/O timing information. This describes when the
operands are read or written from/to the function unit ports. R letter here means reading an input operand
and W means writing a result to an output operand.
52/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
Note: the instruction scheduler of TCE cannot yet take advantage of scheduling input operands after the
triggering operand, so all the input operands should be read at cycle 0.
The writes of the results effectively mark the latency of the operation. Important: The cycles from this
table for the result writes start one cycle after the trigger of the operand. So write in cycle 0 means that
the result is ready on the cycle following the triggering of the operation, not the same cycle as the trigger
is executed.
If an operation produces multiple results, they can be written at different clock cycles.
4.2 Operation Set Abstraction Layer (OSAL) Tools
Input: OSAL definitions
Output: OSAL definitions
4.2.1 Operation Set Editor (OSEd)
Operation Set Editor (OSEd) is a graphical application for managing the OSAL (Section 2.2.6) operation
database. OSEd makes it possible to add, simulate, edit and delete operation definitions.
4.2.1.1 Capabilities of the OSEd
OSEd is capable of the following operations:
1. All operations found in pre-defined search paths (see Section 4.4) are organised in a tree-like
structure which can be browsed.
2. Operation properties can be examined and edited.
3. Operations with a valid behavior model can be tested (simulated).
4. New operation modules can be added to search paths.
5. Operation definitions can be added to a module.
6. Modules containing operation behaviors can be compiled, either all at once, or separately.
7. Modules can be removed.
8. Contents of the memory can be viewed and edited.
4.2.1.2 Usage
This chapter introduces the reader to the usage of OSEd. Instructions to accomplish the common tasks
are given in detail.
Operation Set Editor can simply be executed from command line with:
osed
The main window is split in two areas. The left area always displays a tree-like structure consisting of
search paths for operation definition modules, operation modules, and operations. The right area view
depends on the type of the item that is currently selected in the left area. Three cases are possible.
1. If the selected item is a search path, the right area shows all operation modules in that path.
2. If the item is a module, the right area shows all the operations defined in the module.
3. If the item is an operation, the right area displays all the properties of the operation.
version 31 2016-02-23 53/145
![](asset-36.png)
TTA-based Co-design Environment v1.14 User Manual
Figure 4.1: OSEd Main window.
Figure 4.2: Operation property view.
Figure 4.1 shows an example of the second situation, in which the item currently selected is a module.
The right area of the window shows all the operations in that module. If an operation name is shown in
bold text, it means that the operation definition is “effective”, that is, it will actually be used if clients of
OSAL request an operation with that name. An operation with a given name is effective when it is the
first operation with that name to be found in the search paths. Other operations with the same name may
be found in paths with lower search priority. Those operations are not effective.
Figure 4.2 shows an example of an operation property view, that is shown in the right side when an
operation is selected on the left side.
Editing Static Operation Properties Figure 4.3 shows the dialog for editing the static properties of an
operation.
Operation inputs and outputs (henceforth, “terminal” is used to denote both) can be deleted by selecting
an item from the list and clicking the Delete button. New terminals can be added by clicking the Add
54/145 2016-02-23 version 31
![](asset-37.png)
TTA-BASED CO-DESIGN ENVIRONMENT
Figure 4.3: Operation property window
button, and can be modified by clicking the Modify button. The order of the terminals can be changed
by selecting a terminal and pushing on of the arrow buttons. By pushing the downward arrow button, the
terminal is moved one step downwards in the list; by pushing the upward arrow button, it is moved one
step up on the list.
Operand properties can be modified by clicking on the check boxes. The set of operands that can be
swapped with the operand being edited are shown as a list of references to operands (input indentification
numbers). A reference to an operand can be removed by selecting the corresponding item in the ‘can
swap’ list and clicking the Delete button. A new reference to another operand can be added by selecting
an item from the choice list and clicking the Add button.
Operation Behaviour Model Behaviour models of operations are stored in separate source files. If the
operation definition includes a behaviour model, the behaviour source file can be opened in a text editor
of choice by clicking on the Open button. If the operation does not have behavior source file, clicking
Open will open an empty file in an editor. The text editor to use can be defined in the options dialog. All
changes to operation properties are committed by clicking the OK button and canceled by clicking the
Cancel button.
Operation Directed Acyclic Graph By treating each operation as a node and each input-output pair
as an directed arc, it is possible to construct operation’s Directed Acyclic Graph (DAG) presentation. For
primitive operations which do not call any other operations, this graph is trivial; one node (operation
itself) with input arcs from root nodes and output arcs to leafs. With OSAL DAG language, it is possible
to define operation behavior model by composing it from multiple operations’ respective models.
Operation’s OSAL DAG code sections can be edited by pressing the Open DAG button, which opens the
DAG editor window. Code section shows the currently selected DAG code from the list box below. A
new DAG section can be created either by selecting New DAG list box item or pressing the New button.
By pressing the Undo button, it is possible to revert changes to current code from the last saved position.
DAG can be saved to operation’s definition file by pressing the Save button. Unneccessary DAG sections
can be deleted by pressing the Delete button.
If code section contains valid OSAL DAG code, then the editor window shows a DAG presentation of that
version 31 2016-02-23 55/145
![](asset-38.png)
TTA-based Co-design Environment v1.14 User Manual
Figure 4.4: Dialog for adding new operation module.
Figure 4.5: Result of module compilation
code. In order to view the graph, a program called ’dot’ must be installed. This is included in Graphviz
graph visualization software package and can be obtained from www.graphviz.org.
Operation Modules Figure 4.4 shows the dialog for adding a new operation module to a search path.
The name of the module can be entered into the text input field.
Operation modules may consist also of a behaviour source file. Before operation behaviour modules
can be simulated, it is necessary to compile the source file. Figure 4.5 shows a result dialog of module
compilation.
Data Memory Simulation Model The contents of the data memory simulation model used to simulate
memory accessing operations can be viewed and edited. Figure 4.6 shows the memory window. Memory
can be viewed as 1, 2, 4, or 8 MAUs. The format of the data is either in binary, hexadecimal, signed
integer, unsigned integer, float, or double format. The contents of the memory can be changed by double
clicking a memory cell.
Simulating Operation Behavior The behavior of operation can be simulated using the dialog in Fig-
ure 4.7. Input values can be edited by selecting a input from the input list and typing the new value in a
text field below the input list. Change can be committed by pushing Update button. Trigger command
and advance clock command are executed by pushing Trigger and Advance clock buttons. The format of
the inputs and outputs can be modified by selecting a new format from the choice list above the Trigger
button.
4.2.2 Operation Behavior Module Builder (buildopset)
The OSAL Builder is an external application that simplifies the process of compiling and installing new
(user-defined) operations into the OSAL system.
The OSAL Builder is invoked with the following command line:
buildopset <options> operation_module
where operation_module is the name of the operation module. Operation module is the base name of a
definition file, e.g., the module name of base.opp is ‘base’. The operation_module can also be a full path,
e.g., ‘/home/jack/.tce/opset/custom/mpeg’.
56/145 2016-02-23 version 31
![](asset-39.png)
TTA-BASED CO-DESIGN ENVIRONMENT
Figure 4.6: Memory window
Figure 4.7: Operation Simulation
The behavior definition source file is searched in the directory of the operation_module. The directory of
the operation_module is by default the current working directory. User may also enter the directory of the
source file explicitly with switch ‘-s’. The suffix of the behavior definition source file is ‘.cc’. If no options
are given, the output file is a dynamic module and is stored in the directory of the operation_module.
The OSAL Builder accepts the following command line options:
Short
Name
Long Name Description
b
ignore
boolean Ignores the case whereby the source file containing operation behav-
ior model code are not found. By default, the OSAL Builder aborts
if it cannot build the dynamic module. This option may be used
in combination with install option to install XML data files before
operation behavior definitions are available.
s
source-dir
directory Enter explicit directory where the behavior definition source file to
be used is found.
Third-party libraries can be used in the operation behavior model. buildopset uses the environment vari-
ables CPPFLAGS,CXXFLAGS and LDFLAGS to search for the external library. E.g.:
CPPFLAGS=-I/opt/mpeg/inlucde LDFLAGS=-lmpeg buildopset /home/jack/.tce/opset/custom/mpeg
4.2.3 OSAL Tester (testosal)
The OSAL Tester is a small external application meant for debugging operation behavior models. The
Tester lets user to specify lists of operations and constant input values and the outputs the corresponding
sequence of result values.
The OSAL Tester can be run in interactive mode or in batch mode (like a script interpreter). The batch
mode is especially useful to create input data sets of regression tests.
For all operations having state, a single operation state instance is constructed the first time the operation
is simulated. This state instance is then used through the testing session by all operations that share the
same operation state.
When started, the OSAL Tester enters interactive mode and prompts the user for operations. Any
nonempty line entered by the user is interpreted as one operation (e.g. “ADD 1 2“), unless the char-
acter ‘!’ is given at the beginning of the string. This character (which is not allowed in operation names)
introduces meta-commands that are recognized by the interpreter and are not treated as potential opera-
tions in the architecture repertoire. For example, the OSAL Tester can be quit with the meta-command
!quit. See !help for a list of meta commands and !help <cmdname> for their usage.
Single operands of operations may only be constants and are separated by blanks.
For any line entered by the user, the Tester responds with a text line containing the results computed by
the operation.
4.3 OSAL files
Operations are divided in “operation modules”. For example, ‘base’ module, included in the TCE distri-
bution, contains all the operations available to the front end compiler’s code generation.
An operation in OSAL is defined by its properties and its behavior. The properties defined by OSAL
do not restrict in any way the hardware implementation. For example, latency, bit width or the fact that
reading the result of an operation can lock the processor are properties of the implementation, and are not
defined in OSAL module.
version 31 2016-02-23 57/145
TTA-based Co-design Environment v1.14 User Manual
4.3.1 Operation Properties
The following properties define a TTA operation in TCE:
•name (unique string)
•description (string)
•number of inputs (integer)
•number of outputs (integer)
•accesses memory (yes/no)
•has side effects (yes/no)
•clocked (yes/no)
•affected-by (set of operations)
•affects (set of operations)
operation name The operation name is a string of characters starting with a character in set [A-Z_] and
followed by one or more character in set [0-9A-Z_] (i.e. lowercase letters are not allowed). All names of
operations of the database must be unique. Different data bases can contain operations with equal names.
operation description Optional description of the operation.
inputs Number of inputs of the operation. The number of inputs is a nonnegative integer. It must be
positive if the number of outputs is zero.
outputs Number of outputs of the operation. The number of outputs is a nonnegative integer. It must
be positive if the number of inputs is zero.
reads/writes-memory Indicates that this operation can access memory. Normally, memory access is
also implied from the properties ‘mem-address’ and ‘mem-data’ of operation inputs (or the ‘mem-data’
property of operation outputs). However, it is possible to define operations that perform invisible accesses
to memory, whereby no input or output is related to the memory access itself. That is, neither the address
nor the data moved into or out of memory is explicitly specified by the operation. In these operations,
memory accesses occur as a side effect, and none of the inputs or outputs have memory-related properties.
To avoid potential errors, the memory property must be specified explicitly even when it is implied by
some of the inputs or outputs of the operation. See sections on input and output declarations, below.
clocked Clocked attribute indicates that the operation can change its state synchronously with clock
signal and independently from its input.
side-effect Indicates that two subsequent executions of this operation with the same input values may
generate different output values. An operation marked with “side-effect” is an operation which writes
some state data, affecting further executions of the same operation and other operations sharing that same
state data. The other operations that read or write the same state data written by an operation marked with
“side-effect” must be marked “affected-by” that operation (see later for an example).
Note: only operations that write the state should marked to have “side-effects”. Operations that only read
the state data do not have side effects and may be reordered more freely by the compiler.
58/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
affected-by In case an operation reads or writes state data written by another operation sharing the same
state data (that is marked with the “side-effect” property), the operation should be marked “affected-by”
that operation.
This property restricts the compiler’s reordering optimizations from moving operations that read or write
state data above an operation that also writes the same data, which would result in potentially producing
wrong results from the execution.
affects This is an optional convenience property that allows defining the state data dependency the other
way around. If an operation is listed in the “affects” list it means that the operation is writing state data
that is read or written by the affected operation.
Note: it is not necessary that, if operation A ‘affects’ operation B, then B must contain A in its ‘affected-
by’ list. Vice versa, if A is ‘affected-by’ B, it is not needed that B must contain A in its ‘affects’ list. This
allows, for example, a user to add new operations that share state with the base operations shipped with
TCE, without needing to modify the base operations.
An example of defining multiple operations that share the same state. A common use case for
multiple operations that share state data is a case where one or more operations initialize an internal
register file in an FU and one or more operations use the data in the register file to compute their results.
For example, INIT_RF could initialize the internal register file with the given number. This operation
should be marked “side-effects”. Let’s say that another two operations COMPUTE_X and COMPUTE_Y
only read the internal RF data, thus they can be freely reordered by the computer with each other in the
program code in case there are no other dependencies. As they only read the state data, they don’t have
and visible side effects, thus they should not be marked with the “side-effects” property. However, as they
read the data written by the INIT_RF operation, both operations should be marked to be “affected-by”
the INIT_RF.
4.3.2 Operation Input Properties
Each input of an operation requires an independent declaration of its properties. An operation input is
completely defined by the following properties:
•identification number (integer)
•memory address (yes/no)
•memory data (yes/no)
•can be swapped (set of integers)
identification number Integer number in the range [1,N] where N is the number of inputs as defined
in section 4.3.1 of operation declaration. If N is zero, then no input declarations can be specified. TCE
does not currently allow operations with zero inputs to be defined.
can-swap A list of identification numbers. All the inputs listed can be swapped with this input. The
identification number of this input definition is not allowed in the list. The can-swap property is commu-
tative, thus any of the listed inputs is implicitly ‘can-swap’ with this input and all the other inputs listed.
The can-swap declaration need not be symmetrical, but it is not an error if the implied declaration is also
specified.
mem-address Optional. Indicates that the input is used to compute (affects the value of) the memory
address accessed by this operation.
mem-data Optional. Indicates that the input is used to compute (affects the value of) the data word
written to memory by this operation. This property implies that the operation writes to memory.
version 31 2016-02-23 59/145
TTA-based Co-design Environment v1.14 User Manual
4.3.3 Operation Output Properties
Note: it is not an error if a program, before instruction scheduling, contains an operation where one of
the output moves is missing. If all output moves of an operation are missing, then the only useful work
that can be performed by the operation is state change.
mem-data Optional. Indicates that the output contains a value that depends on a data word read from
memory by this operation. This property implies that the operation reads data from memory.
4.3.4 Operation DAG
The semantics of an operation may be modeled with a simple dag-based language. This can be used to
both simulate the operation without writing the simulation code, and to allow the compiler to automati-
cally use custom operations.
The optional field to describe the Operation DAGs is trigger-semantics. It contains description of the
operation semantics as a directed acyclic graph (DAG).
The OperationDAG language description is given as the content of this element. The OperationDAG
language has the following syntax:
All commands end in semicolon.
SimValue name{, name2, ... } Creates a temporary variables with given name. Usage examples:
SimValue temp;
SimValue temp1, temp2;
EXEC_OPERATION(operationname,input1, ..., inputN,output1, ..., outputN)Calls another operation
with the given operand values.
The input values can be either temporary variables, inputs to the operation whose semantics is
being modeled, or integer constants.
The output value can be either temporary variables or outputs of the operation whose semantics is
being modelled.
Operands of the modeled operation are referred with name IO(<number>) where 1 is first input
operand, 2 second input operand, and after input operands come output operands.
Temporary values are referred by their name, and integer constants are given in decimal format.
Example OperationDAG of the ADDSUB operation:
<trigger-semantics>
EXEC_OPERATION(add, IO(1), IO(2), IO(3));
EXEC_OPERATION(sub, IO(1), IO(2), IO(4));
</trigger-semantics>
4.3.5 Operation Behavior
To be complete, the model of an operation needs to describe the behavior of the operation. The behavior
is specified in a restricted form of C++, the source language of the TCE toolset, augmented with macro
definitions. This definition is used for simulating the operation in the instruction set simulator of TCE.
Definition of Operation Behavior Operation behavior simulation functions are entered inside an op-
eration behavior definition block. There are two kinds of such blocks: one for operations with no state,
and one for operations with state.
OPERATION(operationName)Starts an operation behavior definition block for an operation with name
operationName. Operations defined with this statement do not contain state. Operation names must
be written in upper case letters!
60/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
END_OPERATION(operationName)End an operation behavior definition block for an operation with
no state. operationName has to be exactly the same as it was entered in the block start statement
OPERATION()
.
OPERATION_WITH_STATE(operationName,stateName)Starts an operation behavior definition block
for an operation with state. operationName contains the name of the operation, stateName name of
the state. DEFINE_STATE() definition for the stateName must occur before this statement in the
definition file. Operation and state names must be written in upper case letters!
END_OPERATION_WITH_STATE(operationName)Ends an operation behavior definition block for
an operation with state. operationName has to be exactly the same as it was entered in the block
start statement
OPERATION_WITH_STATE()
.
The main emulation function definition block is given as:
TRIGGER ... END_TRIGGER; Main emulation function.
The bodies of the function definitions are written in the operation behavior language, described in Sec-
tion 4.3.6.
Operations with state. To define the behavior of an operation with state it is necessary to declare
the state object of the operation. An operation state declaration is introduced by the special statement
DEFINE_STATE()
. See Section 4.3.6 for a description of this and related statements. State must be
declared before it is used in operation behavior definition.
A target processor may contain several implementations of the same operation. These implementations
are called Hardware Operations and are described in [CSJ04]. Each Hardware Operation instance belongs
to a different function unit and is independent from other instances. When an operation has state, each
of its Hardware Operations uses a different, independent instance of the state class (one for each function
unit that implements that operation).
An operation state object is unambiguously associated with an operation (or a group of operations, in case
the state is shared among several) by means of its name, which should be unique across all the operation
definitions.
Operation state can be accessed in the code that implements the behavior of the operation by means of
a
STATE
expression. The fields of the state object are accessed with the dot operator, as in C++. See
Section 4.3.6 for a complete description of this statement.
Main emulation function. The behavior model of an operation must include a function that, given a set
of input operand values and, optionally, an operation state instance, produces one or more output values
that the operation would produce.
The definition of an emulation function is introduced by the statement
TRIGGER
and is terminated by the
statement
END_TRIGGER;
.
An emulation function is expected to read all the inputs of its operation and to update the operation
outputs with any new result value that can be computed before returning.
4.3.6 Behavior Description language
The behavior of operations and the information contents of operation state objects are defined by means
of the behavior description language.
The emulation functions that model operation behavior are written in C++ with some restrictions. The
OSAL behavior definition language augments the C++ language with a number of statements. For exam-
ple, control may exit the definition body at any moment by using a special statement, a set of statements
is provided to refer to operation inputs and outputs, and a statement is provided to access the memory
model.
version 31 2016-02-23 61/145
TTA-based Co-design Environment v1.14 User Manual
Base data types. The behavior description language defines a number of base data types. These types
should be used to implement the operation behavior instead of the C base data types, because they guar-
antee the bit width and the format.
IntWord Unsigned integer 32-bit word.
FloatWord Single-precision (32-bit) floating-point word in IEEE-754 format.
DoubleWord Double-precision (64-bit) floating-point word in IEEE-754 format.
HalfWord Half-precision (16-bit) floating-point word in IEEE-754 format.
Access to operation inputs and outputs. Inputs and outputs of operations (henceforth referred to as
terminals, when a distinction is not needed) are referred to by a unique number. The inputs are assigned
a number starting from 1 for the first input. The first output is assigned the number n+1, where nis the
number of inputs of the operations, the second n+2, and so on.
Two sets of expressions are used when accessing terminals. The value of an input terminal can be read
as an unsigned integer, a signed integer, a single precision floating point number, or a double precision
floating point number using the following expressions:
UINT(number)Treats the input terminal denoted by number as a number of type IntWord, which is an
unsigned integer of 32 bits maximum length.
INT(number)Treats the input terminal denoted by number as a number of type SIntWord, which is a
signed integer of 32 bits maximum length.
FLT(number)Treats the input terminal denoted by number as a number of type FloatWord.
DBL(number)Treats the input terminal denoted by number as a number of type DoubleWord.
Output terminals can be written using the following expression:
IO(number)Treats the terminal denoted by number as an output terminal. The actual bit pattern (signed,
unsigned or floating point) written to the output terminal is determined by the right hand expression
assigned to the IO() expression.
Since the behavior of certain operations may depend in non-trivial ways on the bit width of the terminals
of a given implementation, it is sometimes necessary to know the bit width of every terminal. The
expression
BWIDTH(number)
returns the bit width of the terminal denoted by number in the implementation of the calling client.
Bit width of the operands can be extended using two different expressions.
SIGN_EXTEND(integer,sourceWidth)Sign extends the given integer from sourceWidth to 32 bits.
Sign extension means that the sign bit of the source word is duplicated to the extra bits provided by
the wider target destination word.
For example a sign extension from 1001b (4 bits) to 8 bits provides the result 1111 1001b.
ZERO_EXTEND(integer,sourceWidth)Zero extends the given integer from sourceWidth to 32 bits.
Zero extension means that the extra bits of the wider target destination word are set to zero.
For example a zero extension from 1001b (4 bits) to 8 bits provides the result 0000 1001b.
Example. The following code implements the behavior of an accumulate operation with one input and
one output, where the result value is saturated to the “all 1’s” bit pattern if it exceeds the range that can
be expressed by the output:
STATE.accumulator += INT(1);
IntWord maxVal = (1 << BWIDTH(2)) - 1;
IO(2) = (STATE.accumulator <= maxVal ? STATE.accumulator : maxVal);
62/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
Definition of operation state. Operation state consists of a data structure. Its value is shared by one or
more operations, and it is introduced by the statement
DEFINE_STATE(name)
where name is a string that identifies this type of operation state. This statement is followed by a list of
data type fields. The state name string must be generated with the following regular expression:
[A-Z][0-9A-Z_]*
Note that only upper case letters are allowed.
A state definition block is terminated by the statement
END_DEFINE_STATE
Example. The following declaration defines an operation state class identified by the name string “BLISS”,
consisting of one integer word, one floating-point word and a flag:
DEFINE_STATE(BLISS)
IntWord data1;
FloatWord floatData;
bool errorOccurred;
END_DEFINE_STATE;
Some operation state definitions may require that the data is initialized to a predefined state, or even that
dynamic data structures are allocated when the operation state object is created. In these cases, the user
is required to provide an initialization definition inside the state definition block.
INIT_STATE(name)Introduces the code that initializes the operation state.
END_INIT_STATE Terminates the block that contains the initialization code.
Some state definitions may contain resources that need to be released when the state model is destroyed.
For example, state may contain dynamically allocated data or files that need to be closed. In these cases,
the user must define a function that is called when the state is deallocated. This function is defined by a
finalization definition block, which must be defined inside the state definition block.
FINALIZE_STATE(name)Introduces the code that finalizes the operation state, that is, deallocates the
dynamic data contained in an operation state object.
END_FINALIZE_STATE Terminates the block that contains finalisation code.
The state model provides two special definition blocks to support emulation of operation behaviour.
ADVANCE_CLOCK . . . END_ADVANCE_CLOCK In case the model of operations state is synchronous,
this definition can be used to specify activity that occurs “in the raising edge of the clock signal”,
that is, at the end of a simulation cycle. The C ‘return’ statement can used to return from this
function.
Access to operation state. Operation state is denoted by a unique name string and is accessed by means
of the statement
STATE
Typically, an operation state data structure consists of several fields, which are accessed using the dot
operator of C++. For example, the expression
STATE.floatData
in a simulation function refers to the
field
floatData
of the state object assigned to the operation being defined. The precise format of an
operation state structure is defined by means of the
DEFINE_STATE()
statement, and it is specific for an
operation. State must be defined before it can be used in an operation definition.
version 31 2016-02-23 63/145
TTA-based Co-design Environment v1.14 User Manual
Access to control registers. Operations that can modify the program control flow can access the pro-
gram counter register and the return address of the target processor by means of the following expressions:
PROGRAM_COUNTER
RETURN_ADDRESS
Not all operations can access the control registers. Only operations implemented on a Global Control Unit
(see [CSJ04]) provide the necessary data. It is an error to use
PROGRAM_COUNTER
or
RETURN_ADDRESS
in
operations that are not implemented on a Global Control Unit.
Context Identification Each operation context can be identified with a single integer which can be
accessed with
CONTEXT_ID
.
Returning from operation behavior emulation functions. Normally, an emulation function returns
control to the caller when control flows out of the definition body. To return immediately, from an
arbitrary point in the definition body, the following normal C++ return statement can be used. The return
value is boolean indicating the success of the operation execution. In pratice, ’true’ is always returned:
return true;
.
Memory Interface. The memory interface of OSAL is very simplified allowing easy modeling of data
memory accessing operations. The following keywords are used to define the memory access behavior:
MEMORY.read(address, count, target) Reads count units of data from the address to the variable tar-
get.
MEMORY.write(address, count, data) Writes count units of data in variable data to the address.
4.4 OSAL search paths
Default paths, where OSAL operations are searched allows local overriding of global operation defini-
tions. The search paths are the following (the path where the operation is first found will be effective):
1. A custom path in the $TCE_OSAL_PATH/ environment variable.
2. Operations in current working directory:
$PWD/
3. $PWD/data/
where $PWD is your current working directory
4. TCE_SRC_ROOT/opset/base/
where TCE_SRC_ROOT/ is the TCE source code directory.
5. Users local custom operations:
$HOME/.tce/opset/custom/
where $HOME is users home directory.
6. System-wide shared custom operations: TCE_INSTALLATION_DIR/opset/base/
where TCE_INSTALLATION_DIR is the path where TCE accessories is installed (for example
/usr/local/share/tce).
7. Default predefined and standard operations:
TCE_INSTALLATION_DIR/opset/base/
where TCE_INSTALLATION_DIR is the path where TCE accessories is installed (for example
/usr/local/share/tce).
64/145 2016-02-23 version 31
![](asset-41.png)
TTA-BASED CO-DESIGN ENVIRONMENT
4.5 Processor Generator (ProGe)
Processor Generator (ProGe) produces a synthesizable hardware description of a TTA target proces-
sor specified by an architecture definition file (Section 2.2.1) and implementation definition file (Sec-
tion 2.2.4).
Input: HDB, ADF, IDF
Output: VHDL implementation of the processor
There is a command line client for executing this functionality, but the functionality can also be used in
the Processor Designer (Section 4.1). This section is a manual for the command line client.
Processor generation can be customized with plugin modules. The customizable parts are the generation
of control unit and the interconnection network.
The CLI-based version of the processor generator is invoked by means of a command line with the
following syntax:
generateprocessor <options> target
The sole, mandatory argument target gives the name of the file that contains the input data necessary
to generate the target processor. The file can be either a processor configuration file or an architecture
definition file. If the file specified is a PCF, then the names of ADF, BEM and IDF are defined in it.
The given PCF may be incomplete; its only mandatory entry is the reference to the ADF that defines the
architecture of the target processor. If the file specified is an ADF, or if PCF does not contain BEM and
IDF references, then a BEM is generated automatically and the default implementation of every building
block that has multiple implementations is used.
The processor architecture must be synthesizable, otherwise an error message is given.
Short
Name
Long Name Description
b
bem
BEM file or the processor. If not given, ProGe will generate it.
l
hdl
Specifies the HDL of the top-level file which wires together the
blocks taken from HDB with IC and GCU.1
i
idf
IDF file.
o
output
Name of the output directory. If not given, an output directory called
‘proge-output’ is created inside the current working directory.
u
plugin-parameter
Shows the parameters accepted by an IC/Decoder generator plug-in
and a short description of the plug-in. When this options are given,
any other type of option is ignored and ProGe returns immediately
without generating any output file or directory. Even though other
options are ignored they must be valid.
4.5.1 IC/Decoder Generators
IC/Decoder generators are implemented as external plug-in modules. This enables users to provide cus-
tomizable instruction decoder and IC implementations without recompiling or modifying the existing
code base. One can easily add different plug-ins to experiment with different implementation alterna-
tives, and as easily switch from one to another plug-in. To create a new plug-in, a software module that
implements a certain interface must be created, compiled to a shared object file and copied to appropriate
directory. Then it can be given as command line parameter to the generator.
Default IC Decoder Plugin The plugin that is selected by default ProDe’s Processor Implementation
dialog. Parameters that can be passed to default ic decoder plugin are listed below. Note that if bustrace
1In the initial version, the only keyword accepted is ‘vhdl’.
version 31 2016-02-23 65/145
![](asset-42.png)
TTA-based Co-design Environment v1.14 User Manual
or locktrace is enabled the processor can not be synthesized.
Plugin parameter Description
bustrace Value “yes” on the parameter adds code to IC implementation that
prints values of busses into a file called “bus.dump”. The values are
captured every clock period. By default the feature is disabled.
bustracestartingcycle The cycle when the bus tracing starts. Parameter is unset by default.
locktrace Value “yes” on the parameter adds code to decoder implementation
that prints signal value of global lock into a file called “lock.dump”.
The value is captured every clock period. By default the feature is
disabled.
locktracestartingcycle Parameter is unset by default. When numeric value is given it will
be the cycle when the global lock tracing starts. When value is “bus-
tracestartingcycle” then the value is inherited from bustracestarting-
cycle parameter.
4.6 Platform Integrator
Platform Integrator is a part of the Processor Generator tool. Platform Integrator supports automated
integration of TTA cores to different FPGA platforms. The main use case for platform integrator is to
create IP-components out of TTAs for SoC designs for vendor-specific system level design flows. In
addition, standalone processor integration to FPGA chips is supported. Platform Integrator is invoked
with switches to the generateprocessor command.
Platform Integrator specific command line parameters are the following:
Short
Name
Long Name Description
a
absolute-paths
Use absolute paths in generated platform integrator files. By default
integrator uses relative paths.
c
clock-frequency
Defines the target clock frequency. If not given integrator specific
default value will be used.
d
dmem
Data memory type. Available types depends on the platform inte-
grator. Types are ’vhdl_array’, ’onchip’, ’sram’, ’dram’ and ’none’.
Required.
e
entity-name
Name of the toplevel entity of the TTA core. Platform integrator
creates a toplevel entity called entity-name_toplevel. Required.
f
imem
Instruction memory type. Available types depends on the platform
integrator. Types are ’vhdl_array’, ’onchip’, ’sram’ and ’dram’. Re-
quired.
w
imem-width
Defines instruction memory width. This value overrides the instruc-
tion width from BEM.
g
integrator
Selects the platform integrator. Required.
n
list-integrators
List available integrators and information about them.
p
program
Name of tpef program. Required.
Here is an example how to use platform integrator:
generateprocessor -i arch.idf -o proge-output -g Stratix2DSP -d onchip -f vhdl_array
-e FFTmuncher -p fft_app.tpef arch.adf
66/145 2016-02-23 version 31
![](asset-43.png)
TTA-BASED CO-DESIGN ENVIRONMENT
This command would try to integrate processor arch.adf to Stratix2DSP FPGA board using vhdl rom array
for instruction memory and FPGA’s onchip memory for data memory. HDL files created by Platform
Integrator be stored to directory proge-output/platform and project files etc. are written to current working
directory.
4.7 Supported Platforms
This section introduces the Platform Integrators which are shipped with TCE.
4.7.1 Altera based devices
Notice that TCE does not contain the vendor specific IP components such as on-chip memory compo-
nents. In order to simulate and synthesize the hardware you need to have Altera Quartus II software and
libraries installed.
4.7.1.1 Stratix2DSP
Stratix2DSP integrator is targeted to integrate TTA core onto a Stratix II DSP Pro board with EP2S180F1020C3
FPGA device. By default, integrator maps the processor clock signal (clk) to FPGA pin AM17 which
provides 100 MHz clock signal from an oscillator. Reset signal (rstx), which is active low, is mapped to
FPGA pin AG18 which is connected SW8/CPU RESET pushbutton on the FPGA board. Pin AG18 will
be connected to ground when this button is pressed down thus reseting the processor.
Interfacing to FPGA board components can be done by using function units stored in stratixII.hdb. Exter-
nal ports of these function units will be automatically mapped to FPGA pins. It is also vital that the load
store unit of the processor is selected from this HDB, otherwise integrator will fail to create the memory
component mappings. Unidentified external ports will be connected to the new toplevel entity but they
are not mapped to any FPGA pin by default. In this case, the integrator will give a warning message:
“Warning: didn’t find mapping for signal name <signal>”.
Stratix2DSP integrator will also generate project files for Altera Quartus II software. These files will be
written to the current working directory and they are named as <entity_name>.qpf and <entity_name>.qsf.
Project files contain device settings, pin mappings and lists design HDL files. The integrator also creates a
quartus_synthesize.sh shell script which will execute the synthesis process and quartus_program_fpga.sh
which can be used to program the FPGA after successful synthesis.
Available instruction memory types are:
Mem Type Description
onchip Instruction memory is reserved from FPGAs internal memory blocks and the mem-
ory contents will be automatically instantiated during FPGA programming. De-
signer must ensure that instruction memory contents will fit on the FPGA. Otherwise
synthesis process will fail.
vhdl_array Instruction memory is written as a vhdl ROM array and it is synthesized into the
design. Increased synthesis time can be expected when this memory type is used.
Updating instruction memory requires resynthesis.
Available data memory types are:
Mem Type Description
none Indicates that the processor doesn’t have a load store unit or that the integrator
shouldn’t try to make memory connections.
version 31 2016-02-23 67/145
![](asset-44.png)
TTA-based Co-design Environment v1.14 User Manual
onchip Data memory is reserved from FPGA’s internal memory blocks. Memory contents
will be initialized during FPGA programming. When this option is used, load store
unit implementation must be selected as fu_lsu_with_bytemask_always_3 which is
stored in stratixII.hdb.
sram SRAM chip on the FPGA board will be used as data memory. Load store unit im-
plementation must be selected as fu_lsu_sram_static which is stored in stratixII.hdb.
Notice that data memory will be uninitialized when this option is used.
Usage examples:
1. There’s a processor with onchip memory compatible LSU and instruction memory will also use
onchip memory. Let the name of the new toplevel entity be example1. Command to execute
Stratix2DSP integrator is
generateprocessor -i proc.idf -o proge-out -g Stratix2DSP -d onchip -f onchip
-e example1 -p program.tpef proc.adf
2. Same case as before, but now we want to specify target clock frequency of 150 MHz. Quartus II
will then try to reach the specified clock frequency in synthesis process. Command is
generateprocessor -i proc.idf -o proge-out -c 150 -g Stratix2DSP -d onchip -f
onchip -e example1 -p program.tpef proc.adf
Notice that the clock signal will still be connected to a 100 MHz clock oscillator. You’ll have to
provide the 150 MHz clock signal yourself. This can be done for example by instantiating and
connecting a PLL component in proge-out/platform/example1.vhdl file, which is the new toplevel
HDL file.
3. There’s a processor with SRAM compatible LSU and instruction memory will be implemented as
VHDL array. In addition we wish to use absolute paths in project files. Command is then
generateprocessor -i proc.idf -o proge-out -a -g Stratix2DSP -d sram -f vhdl_array
-e example1 -p program.tpef proc.adf
4.7.1.2 Stratix3DevKit
Stratix3DevKit integrates a TTA core onto the Stratix III FPGA Development Kit board from Altera with
an EP3SL150F1152C2 FPGA device. Default clock signal is routed to 125 MHz oscillator connected to
pin B16. Active low reset is connected to CPU RESET button on pin AP5.
Board specific function units are stored in stratixIII.hdb. For successfull integration, you must select the
Load Store Unit from this HDB. This integrator will create project files for Altera Quartus II software in
the same manner as Stratix2DSP integrator (see subsection 4.7.1.1).
Available instruction memory types are:
Mem Type Description
onchip Instruction memory is reserved from FPGAs internal memory blocks and the mem-
ory contents will be automatically instantiated during FPGA programming. De-
signer must ensure that instruction memory contents will fit on the FPGA. Otherwise
synthesis process will fail.
vhdl_array Instruction memory is written as a vhdl ROM array and it is synthesized into the
design. Increased synthesis time can be expected when this memory type is used.
Updating instruction memory requires resynthesis.
Available data memory types are:
68/145 2016-02-23 version 31
![](asset-45.png)
TTA-BASED CO-DESIGN ENVIRONMENT
Mem Type Description
none Indicates that the processor doesn’t have a load store unit or that the integrator
shouldn’t try to make memory connections.
onchip Data memory is reserved from FPGA’s internal memory blocks. Memory contents
will be initialized during FPGA programming. When this option is used, load store
unit implementation must be selected as fu_onchip_ram_lsu_with_bytemask (FU
entry ID 1) which is stored in stratixIII.hdb.
4.7.1.3 AvalonIntegrator
AvalonIntegrator can be used to create an Altera SOPC Builder component from TTA processor which
includes a function unit that implements Avalon Memory Mapped Master interface.
Function units which implement the Avalon Memory Mapped Master interface are stored in avalon.hdb.
There are two ways of interfacing with the Avalon bus:
1. The load-store unit implements the Avalon MM Master interface. The load store unit imple-
mentation must be mapped to avalon_lsu which is stored in avalon.hdb. In this method the data
memory is selected in SOPC Builder and thus the data memory type parameter must be set to
’none’ when integrator is executed. Due to the lack of memory mapper in the current TCE version,
the data memory address space must be set to start from 0 in SOPC Builder. Also the size of data
memory should match the data memory address space size defined in ADF. It is also possible to
include Avalon SFU (see the next bullet) to a processor which has an Avalon LSU but the same
data memory restrictions still apply.
2. A special function unit implements the Avalon MM Master interface. In this case a special
function unit (SFU) is used to interface with Avalon and custom operation macros must be used
to communicate with other Avalon components. This SFU is called avalon_sfu and it’s stored in
avalon.hdb. It is also possible to include multiple Avalon SFUs to the processor but currently there
is no method to differentiate which SFU is used from C code.
In both cases the instruction memory of TTA is not visible to nor accessible via Avalon. Instruction
memory options are the same for Avalon Integrator as for Stratix2DSP Integrator. See section 4.7.1.1 for
more information.
Supported data memory configurations are:
Mem Type Description
none This option indicates that data memory type will be selected in SOPC Builder and
data memory is accessed through Avalon. LSU implementation must be mapped to
avalon_lsu which is stored in avalon.hdb.
onchip Data memory will be reserved from FPGA’s internal memory blocks and the
data memory is not visible to Avalon. LSU implementation must be mapped to
fu_lsu_with_bytemask_always_3 which is stored in stratixII.hdb (despite the name,
the LSU implementation is quite generic for Altera onchip memories).
Avalon Integrator creates a SOPC Builder component file <entity_name>_hw.tcl to current working di-
rectory. This file can be used to import the TTA processor to the SOPC Builder component library. It can
be done by either editing the IP search path from SOPC Builder’s options or adding a new component
using the created tcl-file.
If the processor has other function units with external ports, these external ports are connected to the
interface of the new toplevel entity. These ports are then added to a conduit interface in the SOPC Builder
component file which means that they will also be exported out of the SOPC Builder design.
version 31 2016-02-23 69/145
TTA-based Co-design Environment v1.14 User Manual
Usage examples:
1. A TTA with an Avalon LSU. Because of this, the data memory type must be set to ’none’. Instruc-
tion memory is reserved from onchip memory in this case. Let the name of the new toplevel entity
be tta_sopc. The command to execute Avalon Integrator is
generateprocessor -i proc.idf -o proge-out -a -g AvalonIntegrator -d none -f
onchip -e tta_sopc -p program.tpef proc.adf
The integrator will generate file tta_sopc_hw.tcl which can be used to import the component to
SOPC Builder.
2. A TTA with an Avalon SFU, the instruction memory is implemented as VHDL array and data
memory is reserved from the FPGA’s internal memory. Let the name of the new toplevel entity be
tta_sopc. Command is
generateprocessor -i proc.idf -o proge-out -a -g AvalonIntegrator -d onchip
-f vhdl_array -e tta_sopc -p program.tpef proc.adf
Integrator will generate file tta_sopc_hw.tcl which can be used to import the component to SOPC
Builder. Avalon SFU must be used with custom operation macros from C code. These macros are
named as _TCE_AVALON_<memory_operation>. For example _TCE_AVALON_STW(addr, data)
performs 32-bit store operation to address addr.
3. A TTA with an Avalon LSU and an Avalon SFU. Because Avalon LSU is used, data memory type
must be set to ’none’. Instruction memory is reserved from the onchip memory in this case. Let the
name of the new toplevel entity be dual_avalon_tta_sopc. Command is
generateprocessor -i proc.idf -o proge-out -a -g AvalonIntegrator -d none -f
onchip -e dual_avalon_tta_sopc -p program.tpef proc.adf
The integrator will generate file dual_avalon_tta_sopc_hw.tcl which can be used to import the com-
ponent to SOPC Builder. Component will now have 2 Avalon Memory Mapped Master interfaces.
4.7.1.4 KoskiIntegrator
Koski Integrator can be used to create a Koski/Kactus2 toolset compatible IP blocks from TTA processors.
The integrated TTA must have an function unit that interfaces with HiBi bus to be compatible with this
integrator. Currently this integrator only works with Altera FPGA devices.
HiBi compatible load store unit can be found from hibi_adapter.hdb. Currently it is the only function
unit shipped with TCE which implements the HiBi interface. This means that the only option for data
memory type is ’onchip’. In addition, the onchip memory is generated as a dual port ram which means
that the FPGA device must have support for dual port onchip memories.
Instruction memory options are the same as for Stratix2DSP integrator. See section 4.7.1.1 for more
information.
Koski Integrator will generate an IP-XACT (version 1.2) description file of the integrated component.
File is written to the current working directory and it is named as spirit_comp_def_<entity_name>.xml.
This file is used to import the created TTA IP-component to Koski tools.
Usage example:
There’s a processor with HiBi LSU and instruction memory is implemented using FPGA’s internal mem-
ory and toplevel entity will be named as ’koski_tta’. Command is then:
generateprocessor -i proc.idf -o proge-out -a -g KoskiIntegrator -d onchip -f onchip
-e koski_tta -p program.tpef proc.adf
Integrator creates IP-XACT file ’spirit_comp_def_koski_tta.xml’ to current working directory.
4.7.2 AlmaIFIntegrator
AlmaIF Integrator can be used to create an AlmaIF-compatible processor wrapper for a TTA processor.
AlmaIF is a simple control interface that can be used for control and debug access to devices in a system-
on-a-chip. Due to the constrants placed by the specification, the TTA processor must have data and
70/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
parameter memory spaces. The integrator recognizes these by the address space names, which should
be set to ’data’ and ’param’ respectively. These memory spaces and instruction memory can be accessed
through the AXI4 interface presented by the processor wrapper. Both data and parameter memory spaces
must have 32-bit data width; instruction memory does not have this constraint. All other memory spaces
are mapped as scratchpad memory, inaccessible from the AXI bus.
The LSU compatible with this integrator can be found in almaif.hdb. Currently it is the only function
unit shipped with TCE which implements this interface. This means that the only option for data and
instruction memory type is ’onchip’. The memories will be synthesized as Xilinx-compatible single port
block RAM.
Because the control interface for AlmaIF is implemented through the external debugger interface, this
must be enabled from the Implementation-dialog’s IC / Decoder Plugin tab by setting the bustrace plugin
parameter as ’external’.
Usage example:
There’s a processor with AlmaIF LSU and external debugger interface, and the toplevel entity will be
named ’almaif_tta’. Command is
generateprocessor -i proc.idf -o proge-out -a -g AlmaIFIntegrator -d onchip -f onchip
-e almaif_tta -p program.tpef proc.adf
4.8 Hardware Database Editor (HDB Editor)
HDB Editor (hdbeditor) is a graphical frontend for creating and modifying Hardware Databases i.e. HDB
files (see Section 2.2.3 for details). By default, all the example HDB files are stored in the directory hdb/
of the TCE installation directory.
4.8.1 Usage
This section is intended to familiarize the reader to basic usage of the HDB Editor.
HDB editor can be launched from command line by entering:
hdbeditor
You can also give a .hdb-file as parameter for the hdbeditor:
hdbeditor customHardware.hdb
4.8.1.1 Creating a new HDB file
Choose “File” | “Create HDB...”. From there, type a name for your .hdb file and save it in the default
HDB path (tce/hdb).
After that, you can start adding new TTA components such as function units, register files, buses and
sockets from “Edit” | “Add”.
4.8.1.2 Adding new components
A new function unit’s architecture can only be added through an existing ADF file unlike register files,
which can only be added by hand. The ADF files can be done in the ProDe tool. After adding a new
architecture, one can add an implementation for it by right-clicking on it and choosing “Add implemen-
tation”
The architecture implementation can be given either by hand or by a VHDL file.
After setting up the architecture, one can add new entries (function units, register files, buses, sockets) for
the architectures.
version 31 2016-02-23 71/145
![](asset-48.png)
TTA-based Co-design Environment v1.14 User Manual
Figure 4.8: HDB Editor Main window.
4.8.1.3 Adding FU/RF HDL source files
HDL files of Function Unit and Register File implementations must be added in right compilation order
i.e. the source file which needs to be compiled first is first in the list and so on.
4.8.1.4 Using SRAM based Register File Implementations
Register files of TTA machines may be implemented using synchronous SRAM memory blocks. To
support SRAM memory based register files the requirements are:
1. SRAM memory blocks have fixed latency of one for read operation.
2. SRAM memory blocks can handle as many writes and reads in a single clock cycle as specified in
the ADF.
3. TCE’s Default IC / Decoder Plugin is used to generate the decoder of the processor.
4. SRAM bypasses write data to reading port if the accesses occurs at same cycle and at same address.
Alternatively, the SRAM is wrapped in a module with the bypass logic.
To enable SRAM based register files in the implementation TCE must be notified if a implementation of
RF requires separate address cycle. This is done in HDBEditor by setting option “Separate address cycle”
to true in the Register File Implementation dialog. This option adjusts the timing of read accesses to be
a cycle earlier. The option is effective only if default IC / Decoder Plugin (see: 4.5.1) is used to generate
the machine since it does necessary modification to the instruction pipeline to tweak timings.
HDB includes a single SRAM register file implementation using dual-port block RAM inferred by Xilinx
XST synthesis tool.
4.9 Hardware Database Tester
HDB tester is an utility program which tests that a function unit or register file implementation in HDB
is equal to its simulation model. Tester can be used to test one or all implementations from HDB.
HDB tester uses simulation behaviour models to create reference output values from input stimulus. Then
it creates an RTL testbench with the unit under test and compares the output to the reference. HDB tester
requires an RTL simulator to be installed. Currently ’ghdl’ and ’modelsim’ are supported.
72/145 2016-02-23 version 31
![](asset-49.png)
TTA-BASED CO-DESIGN ENVIRONMENT
4.9.1 Usage
HDB tester executable is called testhdb and it is used as follows:
testhdb <options> hdb-file
HDB tester accepts following command line options:
Short
Name
Long Name Description
f
fuid
Entry id of FU component to be tested. If this is or RF ID are not
defined whole HDB will be tested.
d
leave-dirty
Don’t delete created files
r
rfid
Entry id of RF component to be tested. If this is or FU ID are not
defined whole HDB will be tested.
s
simulator
HDL simulator used to simulate testbench. Accepted values are
’ghdl’ and ’modelsim’. Default is ghdl. Simulator executable must
be found from PATH.
v
verbose
Enable verbose output. Prints all executed commands.
Example: test all FU and RF implementations in HDB:
testhdb asic_130nm_1.5V.hdb
Example: test FU implementation id 1 from HDB, keep testbench files and print commands:
testhdb -d -v -f 1 asic_130nm_1.5V.hdb
Example: test RF implementation id 125 from HDB and use modelsim as RTL simulator:
testhdb -r 125 -s modelsim asic_130nm_1.5V.hdb
If ’verbose’ option is not defined HDB tester won’t output anything when test passes.
4.9.2 Test conditions
Function units is not tested if it meets one of these conditions:
1. Function unit does not have an architecture in HDB
2. Function unit does not have an implementation in HDB
3. Function unit accesses memory
4. Function unit is not pipelined
5. Function unit has external ports
6. Function unit has only one port
Register file is not tested if it meets one of these conditions:
1. register file does not have an architecture in HDB
2. Register file does not have an implementation in HDB
3. Register file does not have a read port
4. Register file does not have a write port
5. Register file has bidirectional port(s)
6. Register file does not have latency of 1 cycle
version 31 2016-02-23 73/145
![](asset-4a.png)
TTA-based Co-design Environment v1.14 User Manual
4.10 Processor unit tester
Processor unit tester validates that the function units and register files of a processor are equal to their
simulation models. The basic operation of processor unit tester is similar to the HDB tester (section 4.9).
4.10.1 Usage
Processor unit tester is invoked as follows:
ttaunittester <options> idf-file
Accepted command line options are:
Short
Name
Long Name Description
a
adf
If ADF file is given IDF will be validated
d
leave-dirty
Don’t delete created files
s
simulator
HDL simulator used to simulate testbench. Accepted values are
’ghdl’ and ’modelsim’. Default is ghdl. Simulator executable must
be found from PATH.
v
verbose
Enable verbose output. Prints all executed commands.
Example: Test units defined in idf:
ttaunittester arch.idf
Example: Test units defined in idf and validate idf against adf:
ttaunittester -a arch.adf arch.idf
Example: Test units defined in idf, enable verbose output, don’t delete created files and use modelsim:
ttaunittester -v -d -s modelsim arch.idf
4.11 Function Unit Interface
Function unit interfaces follow a certain de facto standard in TCE. Here is an example of a such interface:
entity fu_add_sub_eq_gt_always_1 is
generic (
dataw : integer := 32;
busw : integer := 32);
port (
-- trigger port / operand2 input port
t1data : in std_logic_vector (dataw-1 downto 0);
t1opcode : in std_logic_vector (1 downto 0);
t1load : in std_logic;
-- operand1 input port
o1data : in std_logic_vector (dataw-1 downto 0);
o1load : in std_logic;
-- result output port
r1data : out std_logic_vector (busw-1 downto 0);
-- control signals
glock : in std_logic;
rstx : in std_logic;
clk : in std_logic);
end fu_add_sub_eq_gt_always_1;
74/145 2016-02-23 version 31
![](asset-4b.png)
TTA-BASED CO-DESIGN ENVIRONMENT
As you can see the actual implementation interface has more inputs/outputs than the architecture model
of a FU. Architecture input ports are called o1data and t1data and the output port is called r1data. In
addition there are load ports, operation code port and some control signals.
It is good practice to use (integer) generics for the port widths in the data I/O ports. This makes the
managing of Hardware Database (HDB) easier because you can use the name of the generic to define
port width formulas instead of using magic numbers. Whether the value of a generic is constant or
parametrizable depends on how you define it in the HDB. In the example entity declaration we have used
two different generics, data width (dataw) and bus width (busw). The input data ports depend on the
dataw and the output port is defined using busw. You can also use only one generic but with 2 generics
you can define different widths for input and output data if needed.
The names you set for the ports in VHDL code is up to you. When you define an FU implementation
in HDB you’ll have to do port mapping where you map the VHDL implementation names to the HDB
counterparts. I strongly recommend to use a consistent naming convention to avoid confusion. HDBEdi-
tor prefills the common control signal dialog fields with the defacto standard names. If you use your own
naming style for the ports I still recommend to use the standard names for control signals which are used
in the example above (glock = Global Lock, rstx = reset (x stands for active low), clk = clock).
The input ports have also load ports. The load signal is active high when the input value of an input port
is valid and should be read (most probably to a register inside the FU). If the input port is also a trigger
port the operation should be initiated.
If the FU has multiple operations there is also an operation code port in the interface. Opcode port is
usually bound to the trigger port because operations have at least one input. And if you wish to initiate an
operation with one input operand, the operand has to be written to the trigger port. Operation codes also
have to be declared in HDB.
FUs can also have ports connecting to the external of the TTA processor core. They are declared like any
other ports in the entity declaration but these ports must be defined as external in HDB.
See the TCE Tour-tutorial (Section 3.1) for an example how to add an FU implementation to hdb.
4.11.1 Operation code order
If there are multiple operations in a Function Unit the operation codes should be numbered according to
their alphabetical order. E.g. if there is an ALU with following operations: ‘Add’, ‘Sub’, ‘And’, ‘Xor’
and ‘Not’ the operation codes should be ordered as following:
Operation name Operation code
Add 0
And 1
Not 2
Sub 3
Xor 4
HDBEditor and Processor Generator will issue a warning message if this guideline is not used in operation
code numbering. Currently other numbering conventions can still be used but the support might be
dropped in future versions of TCE. In order to ensure compatibility it is recommended that you use this
new convention.
4.11.2 Summary of interface ports
4.11.2.1 Input/Output operand ports
VHDL-type: std_logic_vector
TCE naming convention is to use
t1data
for the triggering input port and
o1data, o2data ...
etc. for
the rest of input operands. Output port is usually called
r1data
.
Use generics when you define widths for these ports.
version 31 2016-02-23 75/145
TTA-based Co-design Environment v1.14 User Manual
4.11.2.2 Input load ports
VHDL-type: std_logic
The load ports are named after the input operand ports. For example load port of t1data is
t1load
and
o1data is
o1load
.
4.11.2.3 Operation code port
VHDL-type: std_logic_vector
Operation code port is usually called
t1opcode
and it is bound to the trigger port.
4.11.2.4 Control signals
VHDL-type: std_logic
There are four control signals available in FUs:
clk
is the most obvious, it is the clock signal
rstx
is active low reset (x stands for active low)
glock
is a Global Lock signal. For example if the global control unit (GCU) issues global lock the
ongoing operation should freeze and wait for the global lock signal to become inactive before resuming
execution.
glock_r
is global lock request. Rarely used in normal function units. It can be used to request global lock
state, for example, in case the result does not arrive in the time set by the static latency of the operation.
A common example using this signal is a load-store unit of which memory latency is sometimes larger
than the static latency informed to the compiler due to dynamic behavior of the memory system. In that
case, a global lock is requested after the static number of cycles has passed and the data has not arrived
to the load-store-unit from the memory system.
4.11.3 Reserved keywords in generics
Currently there is only one reserved keyword in generic definitions and it is
addrw
. This keyword is used
in load-store units to define trigger port’s width according to the data address space width. Processor
Generator identifies this keyword and defines the port width from the data address space assigned to the
LSU in the ADF.
76/145 2016-02-23 version 31
![](asset-4d.png)
TTA-BASED CO-DESIGN ENVIRONMENT
Chapter 5
CODE GENERATION TOOLS
5.1 TCE Compiler
TCE compiler compiles high level language (such as C/C++) source files provided by the toolset user and
produces a bitcode program or a parallel TTA program. Bitcode program is a non-architecture specific
sequential program. Parallel TTA program, however, is a architecture specific program that is optimized
for the target architecture.
The main idea between these two program formats is that you can compile your source code into bitcode
and then compile the bitcode into architecture dependent parallel code. This way you do not have to
compile the source code again every time you make changes in the architecture. Instead you only need to
compile the bitcode into parallel code.
The frontend compiler uses the LLVM C compiler which again is built on GCC version 4.
Input: program source file(s) in high-level language
Output: a fully linked TTA program
5.1.1 Usage of TCE compiler
The usage of the tcecc application is as follows:
tcecc <options> source-or-bc-file1 source-or-bc-file2 ...
The possible options of the application are the following:
Short
Name
Long Name Description
a adf-file Architectures for which the program is scheduled after the compila-
tion. This switch can be used once for each target architecture. Note:
there must be ’schedule’ installed.
s scheduler-config Configure file for scheduling command.
Ooptimization-level Optimization level. 0=no optimizations, 1=preserve program API,
2=do not respect original API, 3 = same that 2
k keep-symbols List of symbols whose optimization away is prevented. If you are
using this, remember to define at least the ’main’ symbol.
o output-name File name of the output binary.
d leave-dirty Does not delete files from each compilation phase.
c compile-only Compiles only. Does not link or optimize.
v verbose Prints out commands and outputs for each phase.
h help Prints out help info about program usage and parameters.
D preprocessor-define Preprocessor definition to be passed to gcc.
version 31 2016-02-23 77/145
TTA-based Co-design Environment v1.14 User Manual
I include-directory Include directory to be passed to gcc.
L library-directory Passed to gcc.
l library-link Passed to gcc.
W warning Ignored.
- scheduler-binary Scheduler binary to use instead of ’schedule’ in path.
- extra-llc-flags Options passed to llc.
- plugin-cache-dir Directory for cached llvm target plugins.
- no-plugin-cache Do not cache generated llvm target plugins.
- rebuild-plugin Rebuild plugin in the cache
- clear-plugin-cache Clear plugin cache completely.
5.1.1.1 Examples of usage
Usage of tcecc quite alike to gcc, excluding that warning options are ignored.
If you wish to compile your source code into optimized bitcode the usage is:
tcecc -O2 -o myProg myProg.c
On the other hand if you already have an architecture definition file of the target processor you can
compile the source code directly to parallel program:
tcecc -O2 -a myProcessor.adf -o myProg.tpef myProg.c
To compile the bitcode program into parallel program use:
tcecc -a myProcessor.adf -o myProg.tpef myProg.bc
Or if you want to a different scheduling configuration than the default:
tcecc -s /path/to/mySchedulerConfiguration.conf -a myProcessor.adf -o myProg.tpef
myProg.bc
Tcecc also has a “leave dirty” flag -d which preserves the intermediate files created by the compiler. After
compilation is complete tcecc will tell you where to find these files (usually it is /tmp/tcecc-xxxxxx/). For
example if you try to compile your C-code straight into a scheduled program and something goes wrong
in scheduling you can find the bitcode program from the temp directory.
tcecc -d -O2 -a myProcessor.adf -o myProg.tpef myProg.c
After compilation you should see this kind of message:
Intermediate files left in build dir /tmp/tcecc-xxxxxx
where xxxxxx is a random pattern of characters.
If you only want to compile the source code without linking (and optimization) use -c flag. Output file is
named after the source file with .o appendix if you do not define an output name with -o.
tcecc -c myProg.c
tcecc -c -o /another/path/myProg.o myProg.c
With tcecc you can explicitly define symbols you wish to preserve in the binary. This can be useful in
debugging and profiling if the compiler removes needed function labels. Symbols are given in a comma
seperated list.
tcecc -O2 -a myMach.adf -k main,foo,bar -o myProg.tpef myProg.c
Plugins
78/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
5.1.2 Custom operations
Tcecc compiler automatically defines macros for operations found from operation definition files in
OSAL search paths and includes them when compiling your program.
The macros use the following format:
_TCE_<opName>(input1, ... , inputN, output1, ... , outputN);
where <name> is the operation name defined in OSAL. Number of input and output operands depends on
the operation.
There are also versions of the custom operation macros which can be used to explicitly target an function
unit (FU).
Usage of the “explicit-FU“ macros is as follows:
_TCEFU_<opName>(<fuName>, input1, ... , inputN, output1, ... , outputN);
where <fuName> is the name of the target function unit in ProDe.
There are also versions of the custom operation macros which can be used to target specific address
space for operations that use memory. These may be need to be used for custom memory operations on
architectures with multiple memory address spaces, as the address space information of a pointer may be
lost with the ordinary custom operation call.
_TCEAS_<opName>("#<as-numerical-id>", input1, ... , inputN, output1, ... , outputN);
or
_TCEAS_<opName>("<as-name>", input1, ... , inputN, output1, ... , outputN);
where <as-numerical-id> is the numerical id of the target address space, and <as-name> is the name of
the target address space. This name is NOT necessaryly the same as the keyword in C code, this is the
name visible in ProDE. Numerical id 0 is always where where the stack is located.
Examples of usage:
_TCE_MAC(a, b, c);
_TCEFU_MAC("MAC1", a, b, c);
_TCEAS_LDW("#1", a, b)
_TCEAS_LDW("my_huge_global_addresspace", a, b)
5.1.3 Known issues
1. Currently it is not possible to simulate a bitcode format program. But the advantage of bitcode
simulation is quite non-existent because the bitcode does not even contain the final basic blocks
that the architecture dependent program has.
5.2 Binary Encoding Map Generator (BEMGenerator)
Binary Encoding Map Generator (BEMGenerator) creates a file that describes how to encode TTA
instructions for a given target processor into bit patterns that make up the executable bit image of the
program (before compression, if used).
Input: ADF
Output: BEM
version 31 2016-02-23 79/145
![](asset-50.png)
TTA-based Co-design Environment v1.14 User Manual
5.2.1 Usage
The usage of BEMgenerator is the following:
createbem -o outName.bem myProcessor.adf
5.3 Parallel Assembler and Disassembler
TCE Assembler compiles parallel TTA assembly programs to a TPEF binary. The Disassembler pro-
vides a textual disassembly of parallel TTA programs. Both tools can be executed only from the command
line.
Assembler Input: program source file in the TTA parallel assembler language and an ADF
Output: parallel TPEF
Disassembler Input: parallel TPEF
Output: textual disassembly of the program
Rest of this section describes the textual appearance of TTA programs, that is, how a TTA program should
be disassembled. The same textual format is accepted and assembled into a TTA program.
5.3.1 Usage of Disassembler
The usage of the tcedisasm application is as follows:
tcedisasm <options> adffile tpeffile
The adffile is the ADF file.
The tpeffile is the parallel TPEF file.
The possible options of the application are as follows:
Short
Name
Long Name Description
o outputfile The name of the output file.
h help Prints out help info about program usage and parameters.
The application disassembles given parallel TPEF file according to given ADF file. Program output is
directed to standard output stream if specific output file is not specified. The output is TTA parallel
assembler language.
The program output can then be used as an input for the assembler program tceasm.
The options can be given either using the short name or long name. If short name is used, a hyphen (-)
prefix must be used. For example -o followed by the name of the output file. If the long name is used, a
double hyphen (- -) prefix must be used, respectively.
5.3.1.1 An example of the usage
The following example generates a disassemble of a parallel TPEF in the file add4_schedule.tpef and
writes the output to a file named output_dis.asm.
tcedisasm -o output_dis.asm add4_supported.adf add4_schedule.tpef
80/145 2016-02-23 version 31
![](asset-51.png)
TTA-BASED CO-DESIGN ENVIRONMENT
5.3.2 Usage of Assembler
The usage of the tceasm application is as follows:
tceasm <options> adffile assemblerfile
The adffile is the ADF file.
The assemblerfile is the program source file in TTA parallel assembler language.
The possible options of the application are as follows:
Short
Name
Long Name Description
o outputfile The name of the output file.
q quiet Do Not print warnings.
h help Help info about program usage and parameters.
The application creates a TPEF binary file from given assembler file. Program output is written to a file
specified by outputfile parameter. If parameter is not given, the name of the output file will be the base
name of the given assembler file concatenated with .tpef.
The options can be given either using the short name or long name. If short name is used, a hyphen (-)
prefix must be used. For example -o followed by the name of the ouput file. If the long name is used, a
double hyphen (- -) prefix must be used, respectively.
5.3.2.1 An example of the usage
The following example generates a TPEF binary file named program.tpef.
tceasm add4_schedule.adf program.asm
5.3.3 Memory Areas
A TTA assembly file consists of several memory areas. Each area specifies the contents (instructions or
data) of part of an independently addressed memory (or address space). There are two kinds of memory
areas: code areas and data areas. Code areas begin a section of the file that defines TTA instructions. Data
areas begin a section that define groups of memory locations (each group is collectively termed “mem-
ory chunk” in this context) and reserve them to variables. By declaring data labels (see Section 5.3.7),
variables can be referred to using a name instead of their address.
Memory areas are introduced by a header, which defines the type of area and its properties. The header
is followed by several logical lines (described in Section 5.3.4), each declaring a TTA instruction or a
memory chunk. The end of an area in the assembly file is not marked. Simply, a memory area terminates
when the header of another memory area or the end of the assembly file is encountered.
The memory area header has one of the following formats:
\tt
CODE [\parm{start}] ;\\
DATA \parm{name} [\parm{start}] ;
A code area begins the declaration of TTA instructions that occupy a segment of the instruction memory.
A data area begins the declaration of memory chunks reserved to data structures and variables.
A TTA program can work with several independently addressed data memories. The locations of different
memories belong to different address spaces. The name parameter defines the address space a memory
area belongs to. The code area declaration does not have a name parameter, because TTA programs
support only one address space for instruction memory, so its name is redundant.
version 31 2016-02-23 81/145
TTA-based Co-design Environment v1.14 User Manual
The start parameter defines the starting address of the area being declared within its address space. The
start address can and usually is omitted. When omitted, the assembler will compute the start address and
arrange different memory area declarations that refer to the same address space. The way the start address
is computed is left to the assembler, which must follow only two rules:
1. If a memory area declaration for a given address space appears before another declaration for the
same address space, it is assigned a lower address.
2. The start address of a memory area declaration is at least equal to the size of the previous area
declared in the same address space plus its start address.
The second rule guarantees that the assembler reserves enough memory for an area to contain all the
(chunk or instruction) declarations in it.
5.3.4 General Line Format
The body of memory areas consists of logical lines. Each line can span one or more physical lines of
the text. Conversely, multiple logical lines can appear in a single physical lines. All logical lines are
terminated by a semicolon ‘;’.
The format of logical lines is free. Any number of whitespace characters (tabs, blanks and newlines) can
appear between any two tokens in a line. Whitespace is ignored and is only useful to improve readability.
See Section 5.3.14 for suggestions about formatting style and use of whitespaces.
Comments start with a hash character (‘#’) and end at the end of the physical line. Comments are ignored
by the syntax. A line that contains only a comment (and possibly whitespaces before the hash character)
is completely removed before interpreting the program.
5.3.5 Allowed characters
Names (labels, procedures, function units etc.) used in assembly code must obey the following format:
[a-zA-Z_][a-zA-z0-9_]*
Basically this means is that a name must begin with a letter from range a-z or A-Z or with an underscore.
After the first character numbers can also be used.
Upper case and lower case letters are treated as different characters. For example labels main: and Main:
are both unique.
5.3.6 Literals
Literals are expressions that represent constant values. There are two classes of literals: numeric literals
and strings.
Numeric literals. A numeric literal is a numeral in a positional system. The base of the system (or
radix) can be decimal, hexadecimal or binary. Hexadecimal numbers are prefixed with ‘0x’, binary
numbers are prefixed with ‘0b’. Numbers in base 10 do not have a prefix. Floating-point numbers can
only have decimal base.
Example: Numeric literals.
0x56F05A
7116083
0b11011001001010100110011
17.759
308e+55
82/145 2016-02-23 version 31
![](asset-53.png)
TTA-BASED CO-DESIGN ENVIRONMENT
The first three literals are interpreted as integer numbers expressed in base, respectively, 16, 10 and 2. An
all-digit literal string starting with ‘0’ digit is interpreted as a decimal number, not as an octal number,
as is customary in many high level languages.1The last two literals are interpreted as floating point
numbers. Unlike integer literals, floating-point literals can appear only in initialisation sequences of data
declarations (see Section 5.3.8 for details).
String literals. A string literal consists of a string of characters. The the numeric values stored in the
memory chunk initialised by a string literal depend on the character encoding of the host machine. The EXTENSION: charset di-
rective
use of string literals makes the assembly program less portable.
Literals are defined as sequences of characters enclosed in double (
"
) or single (
’
) quotes. A literal can
be split into multiple quoted strings of characters. All strings are concatenated to form a single sequence
of characters.
Double quotes can be used to escape single quotes and vice versa. To escape a string that contains both,
the declaration must be split into multiple strings.
Example: String literals. The following literals all declare the same string
Can’t open file "%1"
.
"Can’t open file" ’"%1"’
’Can’ "’" ’t open file "%1"’
"Can’t open" ’ file "%1"’
String literals can appear only in initialisation sequences of data declarations (see Section 5.3.8 for de-
tails).
Size, encoding and layout of string literals. By default, the size (number of MAU’s) of the value
defined by a string literal is equal to the number of characters. If one MAU is wider than the character
encoding, then the value stored in the MAU is padded with zeroes. The position of the padding bits
depends on the byte-order of the target architecture: most significant if “big endian”, least significant if
“little endian”.
If one character does not fit in a singe MAU, then each character is encoded in ⌈m/n⌉MAU’s, where nis
the MAU’s bit width and mis the number of bits taken by a character.
When necessary (for example, to avoid wasting bits), it is possible to specify how many characters are
packed in one MAU or, vice versa, how many MAU’s are taken to encode one character. The size specifier
for characters is prefixed to a quoted string and consists of a number followed by a semicolon.
If n>m, the prefixed number specifies the number of characters packed in a single MAU. For example,
if one MAU is 32 bits long and a character takes 8 bits, then the size specifier in
4:"My string"
means: pack 4 characters in one MAU. The size specifier cannot be greater than ⌈n/m⌉. The size ‘1’ is
equivalent to the default.
If m>n, the prefixed number specifies the number of adjacent MAU’s used to encode one character. For
example, if MAU’s are 8-bit long and one character takes 16 bits, then the same size specifier means:
reserve 4 MAU’s to encode a single character. In this case, a 16-bit character is encoded in 32 bits, and
padding occurs as described above. The size of the specifier in this case cannot be smaller than ⌈n/m⌉,
which is the default value when the size is not specified explicitly.
5.3.7 Labels
A label is a name that can be used in lieu of a memory address. Labels “decorate” data or instruction
addresses and can be used to refer to, respectively, the address of a data structure or an instruction. The
address space of a label does not need to be specified explicitly, because it is implied by the memory area
declaration block the label belongs to.
A label declaration consists of a name string followed by a colon:
1This notation for octal literals has been deprecated.
version 31 2016-02-23 83/145
TTA-based Co-design Environment v1.14 User Manual
\tt
\parm{label-name}:
Only a restricted set of characters can appear in label names. See Section 5.3.5 for details.
A label must always appear at the beginning of a logical line and must be followed by a normal line
declaration (see Sections 5.3.8, 5.3.9 for details). Only whitespace or another label can precede a label.
Label declarations always refer to the address of the following memory location, which is the start location
of the element (data chunk or a TTA instruction) specified by the line.
Labels can be used instead of the address literal they represent in data definitions and instruction defini-
tions. They are referred to simply by their name (without the colon), as in the following examples:
# label reference inside a code area (as immediate)
aLabel -> r5 ;
# label reference inside a data area (as initialisation value)
DA 4 aLabel ;
5.3.8 Data Line
A data line consists of a directive that reserves a chunk of memory (expressed as an integer number of
minimum addressable units) for a data structure used by the TTA program:
\tt
DA \parm{size} [\parm{init-chunk-1} \parm{init-chunk-2} \ldots] ;
The keyword ‘DA’ (Data Area) introduces the declaration of a memory chunk. The parameter size gives
the size of the memory chunk in MAU’s of the address space of the memory area.
Memory chunks, by default, are initialised with zeroes. The memory chunk can also be initialised ex-
plicitly. In this case, size is followed by a number of literals (described in Section 5.3.6) or labels (Sec-
tion 5.3.7) that represent initialisation values. An initialisation value represents a constant integer number
and takes always an integer number of MAU’s.
Size of the initialisation values. The size of an initialisation value can be given by prepending the size
(in MAU’s) followed by a semicolon to the initialisation value. If not defined explicitly, the size of the
initialisation values is computed by means of a number of rules. If the declaration contains only one
initialisation value, then the numeric value is extended to size, otherwise, the rules are more complex and
depend on the type of initialisation value.
1. If the initialisation value is a numeric literal expressed in base 10, then it is extended to size MAU’s.
2. If the initialisation value is a numeric literal expressed in base 2 or 16, then its size is extended to
the minimum number of MAU’s necessary to represents all its digits, even if the most significant
digits are zeroes.
3. If the initialisation value is a label, then it is extended to size MAU’s.
Extension sign. Decimal literals are sign-extended. Binary, hexadecimal and string literal values are
zero-extended. Also the initialisation values represented by labels are always zero-extended.
Partial Initialisation. If the combined size of the initialisation values (computed or specified explicitly,
it does not matter) is smaller than the size declared by the ‘DA’ directive, then the remaining MAU’s are
initialised with zeroes.
Example: Padding of single initialisation elements. Given an 8-bit MAU, the following declarations:
84/145 2016-02-23 version 31
![](asset-55.png)
TTA-BASED CO-DESIGN ENVIRONMENT
DA 2 0xBB ; # equivalent to 2:0xBB
DA 2 0b110001 ; # 0x31 (padded with 2 zero bits)
DA 2 -13 ;
define 2-MAU initialisation values: 0x00BB, 0x0031, and 0xFFF3, respectively.
Example: Padding of of multi-element initialisation lists. The following declarations:
DA 4 0x00A8 0x11;
DA 4 0b0000000010100100 0x11 ;
are equivalent and force the size of the first initialisation value in each list to 16 bits (2 MAU’s) even if
the integer expressed by the declarations take less bits. The 4-MAU memory chunk is initialised, in both
declarations, with the number 0x00A81100. Another way to force the number of MAU’s taken by each
initialisation value is to specify it explicitly. The following declarations are equivalent to the declarations
above:
DA 4 2:0xA8 0x11;
DA 4 2:0b10100100 0x11;
Finally, the following declarations:
DA 2 1:0xA8 0x11;
DA 2 1:0b10100100 0x11;
define a memory chunk initialised with 0xA8110000. The initialisation value (in case of the binary literal,
after padding to MAU bit width) defines only the first MAU.
When labels appear in initialisation sequences consisting of multiple elements, the size of the label ad-
dress stored must be specified explicitly.
Example. Initialisation with Labels. The following declaration initialises a 6-MAU data structure where
the first 2 MAU’s contain characters ‘A’ and ‘C’, respectively, and the following 4 MAU’s contain two
addresses. The addresses, in this target architecture, take 2 MAU’s.
DA 6 0x41 0x43 2:nextPointer 2:prevPointer ;
5.3.9 Code Line
A code line defines a TTA instruction and consists of a comma-separated, fixed sequence of bus slots. A
bus slot in any given cycle can either program a data transport or encode part of a long immediate and
program the action of writing it to a destination (immediate) register for later use.2
A special case of code line that defines an empty TTA instruction. This line contains only three dots
separated by one or more white spaces:
. . . ; # completely empty TTA instruction
A special case of move slot is the empty move slot. An empty move slot does not program any data
transport nor encodes bits of a long immediate. A special token, consisting of three dots represents an
empty move slot. Thus, for a three-bus TTA processor, the following code line represents an empty
instruction:
... , ... , ... ; # completely empty TTA instruction
2The action of actually writing the long immediate to a destination register is encoded in a dedicated instruction field, and is
not repeated in each move slot that encodes part of the long immediate. This detail is irrelevant from the point of view of program
specification. Although the syntax is slightly redundant, because it repeats the destination register in every slot that encodes a piece
of a long immediate, it is chosen because it is simple and avoids any chance of ambiguity.
version 31 2016-02-23 85/145
TTA-based Co-design Environment v1.14 User Manual
5.3.10 Long Immediate Chunk
When a move slot encodes part of a long immediate, its declaration is surrounded by square brackets and
has the following format:
\tt
\parm{destination}=\parm{value}
where destination is a valid register specifier and value is a literal or a label that gives the value of
the immediate. The only valid register specifiers are those that represent a register that belongs to an
immediate unit. See section 5.3.12 for details on register specifiers.
When the bits of a long immediate occupy more than one move slot, the format of the immediate dec-
laration is slightly more complex. In this case, the value of the immediate (whether literal or label) is
declared in one and only one of the slots (it does not matter which one). The other slots contain only the
destination register specifier.
5.3.11 Data Transport
A data transport consists of one optional part (a guard expression) and two mandatory parts (a source and
a destination). All three can contain an port or register specifier, described in Section 5.3.12.
The guard expression consists of a single-character that represents the invert flag followed by a source
register specifier. The invert flag is expressed as follows:
1. Single-character token ‘!’: the result of the guard expression evaluates to zero if the source value is
nonzero, and evaluates to one if the source value is equal to zero.
2. Single-character token ‘?’: the result of the guard expression evaluates to zero if the source value
is zero, and evaluates to one if the source value is not zero.
The move source specifier can be either a register and port specifier or an in-line immediate. Register and
port specifiers can be GPR’s, FU output ports, long immediate registers, bridge registers. The format of
all these is specified in Section 5.3.12. The in-line immediate represents an integer constant and can be
defined as a literal or as a label. In the latter case, the in-line immediate can be followed by an equal sign
and a literal corresponding to the value of the label. The value of the labels is more likely to be shown
as a result of disassembling an existing program than in user input code, since users can demand data
allocation and address resolution to the assembler.
Example: Label with value. The following move copies the label ‘LAB’, which represents the address
0x051F0, to a GPR:
LAB=0x051F0 -> r.4
The move destination consists of a register and port specifier of two types: either GPR’s or FU input
ports.
5.3.12 Register Port Specifier
Any register or port of a TTA processor that can appear as a move or guard source, or as a move destination
is identified and referred to by means of a string. There are different types of register port specifiers:
1. General-purpose register.
2. Function unit port.
3. Immediate register.
4. Bridge register.
86/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
GPR’s are specified with a string of the following format:
\tt
\parm{reg-file}[.\parm{port}].\parm{index}
where reg-file is the name of the register file, port, which can be omitted, is the name of the port through DISCUSS: pending ??
which the register is accessed, and index is the address of the register within its register file.
Function unit input and output ports are specified with a string of the following format:
\tt
\parm{function-unit}.\parm{port}.[\parm{operation}]
where function-unit is the name of the function unit, port is the name of the port through which the reg-
ister is accessed, and operation, which is required only for opcode-setting ports, identifies the operation
performed as a side effect of the transport. It is not an error to specify operation also for ports that do
not set the opcode. Although it does not represent any real information encoded in the TTA program, this
could improve the readability of the program.
Immediate registers are specified with a string if the following format:
\tt
\parm{imm-unit}[.\parm{port}].\parm{index}
where imm-unit is the name of the immediate unit, port, which can be omitted, is the name of the port DISCUSS: pending ??
through which the register is accessed, and index is the address of the register within its unit.
Since any bus can be connected to at most two busses through bridges, it is not necessary to specify bridge
registers explicitly. Instead, the string that identifies a bridge register can only take one of two values:
‘{prev}’ or ‘{next}’. These strings identify the bus whose value in previous cycle is stored in the register
itself. A bus is identified by ‘{prev}’ if it is programmed by a bus slot that precedes the bus slot that reads
the bridge register. Conversely, if the bus is identified by ‘{next}’, then it is programmed by a bus slot
that follows the bus slots that reads the bridge register. In either case, the source bus slot must be adjacent
to the bus slot that contains the moves that reads the bridge register.
Example: possible register and port specifiers.
\begin{tabular}{lp{0.75\textwidth}}
\texttt{IA.0} & immediate unit ‘IA’, register with index 0\\
\texttt{RFA.5} & register file ‘RFA’, register with index 5\\
\texttt{U.s.add} & port ‘s’ of function unit ‘U’, opcode for operation
‘add’\\
\verb|{|\texttt{prev}\verb|}| & bridge register that contains the value on
the
bus programmed by the previous bus slot in previous
cycle\\
\end{tabular}
Alternative syntax of function unit sources and destinations. Most clients, especially user interfaces,
may find direct references to function unit ports inconvenient. For this reason, an alternative syntax is
supported for input and output ports of function units:
\tt
\parm{function-unit}.\parm{operation}.\parm{index}
where function-unit is the name of the function unit, operation identifies the operation performed as a side
effect of the transport and index is a number in the range [1,n], where nis the total number of inputs and
outputs of the operation. The operation input and output, indirectly, identifies also the FU input or output
and the port accessed. Contrary to the base syntax, which is requires the operation name only for opcode-
setting ports, this alternative syntax makes the operation name not optional. The main advantage of this
version 31 2016-02-23 87/145
TTA-based Co-design Environment v1.14 User Manual
syntax is that is makes the code easier to read, because it removes the need to know what is the operation
input or output bound to a port, because. The main drawback is an amount of (harmless) “fuzziness” and
inconsistency, because it forces the user to define an operation for ports that do not set the opcode, even
in cases where the operand is shared between two different operations. For example, suppose that the
operand ‘1’ of operations ‘add’ and ‘mul’ is bound to a port that does not set the opcode and its value is
shared between an ‘add’ and a ‘mul’:
r1 -> U1.add.1, r2 -> U1.add.2;
U1.add.3 -> r3, r4 -> U1.mul.2;
U1.mul.3 -> r5
it looks as if the shared move belonged only to ‘add’. One could have also written, correctly but less
clearly:
r1 -> U1.mul.1, r2 -> U1.add.2;
# same code follows
or even, assuming that operation ‘sub’ is also supported by the same unit and its operand ‘1’ is bound to
the same port:
r1 -> U1.sub.1, r2 -> U1.add.2;
# same code follows
This alternative syntax is the only one permitted for TTA moves where operations are not assigned to a
function unit of the target machine.
5.3.13 Assembler Command Directives
Command directives do not specify any code or data, but change the way the assembler treats (part of) the
code or data declared in the assembly program. A command directive is introduced by a colon followed
by the name string that identifies it, and must appear at the beginning of a new logical line (possibly with
whitespace before).
The assembler recognises the following directives.
procedure The ‘
:procedure
’ directive defines the starting point of a new procedure. This directive is
followed by one mandatory parameter: the name of the procedure. Procedure directives should appear
only in code areas. The procedure directive defines also, implicitly, the end of procedure declared by
the previous ‘:procedure’ directive. If the first code section of the assembly program contains any code
before a procedure directive, the code is assumed to be part of a nameless procedure. Code in following
code areas that precede any procedure directive is considered part of the last procedure declared in one of
the previous code areas.
Example: declaration of a procedure.
CODE ;
:procedure Foo ;
Foo:
r5 -> r6 , ... ;
...;
... , r7 -> add.1 ;
In this example, a procedure called ‘Foo’ is declared and a code label with the same name is declared at
the procedure start point. The code label could be given any name, or could be placed elsewhere in the
same procedure. In this case, the code label ‘Foo’ marks the first instruction of procedure ‘Foo’.
global The ‘
:global
directive declares that a given label is globally visible, that is, it could be linked
and resolved with external code. This directive is followed by one mandatory parameter: the name of the
label. The label must be defined in the same assembly file. The label may belong to the data or the code
section, indifferently.
88/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
extern The ‘
:extern
directive declares that a given label is globally visible and must be resolved an
external definition. This directive is followed by one mandatory parameter: the name of the label. The
label must not be defined in the assembly file.
There can be only one label with any given name that is declared global or external.
Example: declaration of undefined and defined global labels.
DATA dmem 0x540;
aVar:
DA 4 ;
:global aVar ;
:extern budVar ;
In this example, ‘aVar’ is declared to have global linkage scope (that is, it may be used to resolve refer-
ences from other object files, once the assembly is assembled). Also ‘budVar’ is declared to have global
linkage, but in this case the program does not define a data or code label with that name anywhere, and
the symbol must be resolved with a definition in an external file.
5.3.14 Assembly Format Style
This section describes a number of nonbinding guidelines that add to the assembly syntax specification
and are meant to improve programs’ readability.
Whitespaces. Although the format of the assembly is completely free-form, tabs, whitespaces and new
lines can be used to improve the assembly layout and the readability. The following rules are suggested:
1. Separate the following tokens with at least one whitespace character:
(a) Label declaration ‘name:’ and first move or ‘DA’ directive.
(b) Moves of an instruction and commas.
(c) Move source and destination and the ‘->’ or ‘<-’ token.
2. Do not separate the following tokens with whitespaces:
(a) Long immediate chunk declaration and the surrounding brackets.
(b) Label and, literal and the ‘=’ token in between.
(c) Any part of a register specifier (unit, port, operation, index) and the ‘.’ separator token.
(d) Register specifier, label or literal and the ‘=’ in between.
(e) Invert flag mark (‘!’ or ‘?’) and the register specifier of a guard expression.
(f) Initialisation chunk, the number of MAU’s it takes and the ‘:’ token in between.
(g) Colon ‘:’ and the nearby label or directive name.
End of Line. The length of physical lines accepted is only limited by client implementation. Lines
up to 1024 characters must be supported by any implementation that complies with these specifications.
However, it is a good rule, to improve readability, that physical line should not exceed the usual line length
of 80 or 120 characters. If a TTA instruction or a data declaration does not fit in the standard length, the
logical line should be split into multiple physical lines. The physical lines following the first piece of a
logical line can be indented at the same column or more to the right. In case of data declarations, the line
is split between two literals that form an initialisation data sequence. In case of TTA instructions, logical
lines should never be split after a token of type ‘X’ if it is recommended that no whitespace should follow
‘X’ tokens. To improve readability, TTA instructions should be split only past the comma that separates
two move slots:
version 31 2016-02-23 89/145
TTA-based Co-design Environment v1.14 User Manual
# good line breaking
r2 -> U.sub.1 , [i1=var] , r3 -> U.sub.2 , i1 -> L.ld.1 , [i1] ,
0 -> U.eq.2 ;
# bad line breaking
r2 -> U.sub.1 , [i1=var] , r3 -> U.sub.2 , i1 -> L.ld.1 , [i1] , 0 ->
U.eq.2 ;
# really bad line breaking
r2 -> U.sub.1 , [i1=var] , r3 -> U.sub.2 , i1 -> L.ld.1 , [i1] , 0 -> U.
eq.2 ;
Tabulation. The following rules can be taken as starting point for a rather orderly layout of the assembly
text, which resembles the layout of traditional assembly languages:
1. The first ncharacters of the assembly lines are reserved to labels. Instruction or data declarations
are always indented by ncharacters.
2. Labels appear in the same physical line of the instruction or data declaration they refer to. Labels
are no more than n−2 characters long.
This layout is particularly clean when the TTA instructions contain few bus slots and when multiple labels
for the same data chunk or instruction do not occur.
Example: Assembly layout style best suited target architectures with few busses.
DATA DMEM
var: DA 4;
CODE
lab_A: spr -> U.add.1 , 55 -> U.add.2 , spr -> r12 ;
[i0=0x7F] , U.add.3 -> spr , i0 -> r2 ;
loop_1: r2 -> U.sub.1 , r3 -> U.sub.2 , var -> L.ld.1 ;
r2 -> U.eq.1 , U.sub.3 -> r2 , 0 -> U.eq.2 ;
?U.eq.3 loop_1 -> C.jump.1 , L.ld.2 -> U.and.2 ;
0x1F -> U.and.1 , ... , ... ;
... , U.and.3 -> r8 , ... ;
An alternative layout of the assembly text is the following:
1. Instruction and data declarations are always indented by ncharacters.
2. Each label declaration appears in a separate physical line of the instruction or data declaration they
refer to, and starts from column 0.
This layout could be preferable when the TTA instructions contain so many bus slots that the logical line
is usually split into multiple physical lines, because it separates more clearly the code before and after a
label (which usually marks also a basic block entry point). In addition, this layout looks better when an
instruction or data declaration has multiple labels and when the label name is long.
Example: Assembly layout style best suited targets with many busses.
DATA DMEM
var:
DA 4;
CODE
a_long_label_name:
spr -> U.add.1 , 55 -> U.add.2 , spr -> r12 , [i0=0x7F], i0 -> r2,
90/145 2016-02-23 version 31
![](asset-5b.png)
TTA-BASED CO-DESIGN ENVIRONMENT
... ;
... , U.add.3 -> spr , ... , ... , ... , ... ;
loop_1:
r2 -> U.sub.1 , [i1=var] , r3 -> U.sub.2 , i1 -> L.ld.1 , [i1] ,
0 -> U.eq.2 ;
r2 -> U.eq.1 , U.sub.3 -> r2 , ... , ?U.eq.3 loop_1 -> C.jump.1 ,
0x1F -> U.and.1 , ... ;
L.ld.2 -> U.and.2 , ... , ... , ... , ... , ... ;
... , ... , ... , U.and.3 -> r8 , ... , ... ;
This example of assembly code is exactly equivalent to the code of previous example, except that the
address of ‘var’ data chunk (a 4-MAU word) is encoded in a long immediate and takes 2 move slots.
Layout of Memory Area Declarations. It is preferable to subdivide the contents of memories into
several memory area declarations and to group near each other area declarations of different address
spaces that are related to each other. This underlines the relation between data and code. The alternative,
a single area for each address space, mixes together all data and all procedures of a program.
Mixing Alternative Syntaxes. It is preferable to not mix alternative styles or even syntaxes, although
any client that works with the assembly language is expected to deal with syntax variants.
5.3.15 Error Conditions
This section describes all the possible logical errors that can occur while assembling a TTA program.
Address Width Overflow in Initialisation. A label is used as initialisation value of a data declaration,
and the label address computed by the assembler exceeds the number of MAU’s in the data declaration
that must be initialised.
Address Width Overflow in Long Immediate. A label is used as initialisation value of a long imme-
diate declaration, and the label address computed by the assembler exceeds total width of the move slots
that encode the immediate, when concatenated together.
Address Width Overflow in In-line Immediate. A label is used as initialisation value of an in-line
immediate, and the label address computed by the assembler exceeds width of source field that encodes
the immediate.
Unspecified Long Immediate Value. A long immediate is defined, but none of the move slots that
encode its bits defines the immediate value.
Multiply Defined Long Immediate Value. More than one of the move slots that contain the bits of a
long immediate defines the immediate value.3
Overlapping Memory Areas. The start address specified in the header of two memory area declara-
tions is such that, once computed the sizes of each memory area, there is an overlapping.
Multiple Global Symbols with Same Name. A ‘:global’ directive declares a symbol with given name
as globally visible, but multiple labels with given name are declared in the program.
3If the values are identical in every move slot, then the client could issue a warning rather than a critical error.
version 31 2016-02-23 91/145
TTA-based Co-design Environment v1.14 User Manual
Unknown Command Directive. A command directive has been found that is not one of the directives
supported by the assembler.
Misplaced Procedure Directive. A ‘:procedure’ directive appears inside a data area declaration block.
Procedure Directive Past End of Code. A ‘:procedure’ directive appears after the last code line in the
program.
Label Past End of Area. A label has been declared immediately before an area header or at the end of
the assembly program. Labels must be followed by a code line or a data line.
Character Size Specifier too Big. A size specified for the characters of a string literal is greater than
the maximum number of characters that can fit in one MAU.
Character Size Specifier too Small. A size specified for the characters of a string literal is smaller than
the minimum number of MAU’s necessary to encode one character.
Illegal Characters in Quoted String. A quoted string cannot contain non-printable characters (that is,
characters that cannot be printed in the host encoding) and end-of-line characters.
5.3.16 Warning Conditions
This section describes all conditions of target architecture or assembly syntax for which the client should
issue an optional warning to prepare users for potential errors or problematic conditions.
Equally Named Register File and Function Unit. A register file and a function unit of the target
architecture have the same name. This is one of the conditions for the activation of the disambiguation
rule.
Port with the Name of an Operation. A register file or a function unit port are identified by a string
name that is also a name of a valid operation supported by the target architecture. The first condition
(port of register file) is more serious, because it may require triggering a disambiguation rule. The second
condition (FU port) is not ambiguous, but is confusing and ugly. The second condition may be more or
less severe depending, respectively, whether the operation with the same name is supported by the same
FU or by another FU.
Code without Procedure. The first code area of the program begins with code lines before the first
‘:procedure’ directive. A nameless procedure with local visibility is automatically created by the assem-
bler.
Procedure Spanning Multiple Code Areas. A code area contains code line before the first ‘:procedure’
directive, but it is not the first code area declared in the code. The code at the beginning of the area is
attached to the procedure declared by the last ‘:procedure’ directive.
Empty Quoted String. Empty quoted strings are ignored.
92/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
5.3.17 Disambiguation Rules
Certain syntactic structures may be assigned different and (in principle) equally valid interpretations.
In these cases, a disambiguation rule assigns priority to one of the two interpretation. Grammatically
ambiguous assembly code should be avoided. Clients that operate on TTA assembly syntax should issue
a warning whenever a disambiguation rule is employed.
Disambiguation of GPR and FU terms. When a GPR term includes also the RF port specifier, it can
be interpreted also as a function unit input or output.
Normally, the names of units, ports and operation rule out one of the two interpretations. Ambiguity can
only occur only if:
1. The target architecture contains a RF and a FU with identical name.
2. One of the RF ports has a name identical to one of the operations supported by the FU that has the
same name of the RF. PENDING: disambigua-
tion of unscheduled TTA
code ??
Ambiguity is resolved in favour of the GPR interpretation. No condition applies to the indices (register
index or operation input or output index). The first interpretation is chosen even when it results in a
semantic error (an index out of range) whereas the other interpretation would be valid.
Example. Disambiguation rule. The following move is interpreted as a move that writes the constant
55 to the register with index 2 of register file ‘xx’ through port ‘yy’. If there exists an FU called ‘xx’
that supports an operation ‘yy’ which has an input with index 2, this interpretation of the move is never
possible.
55 -> xx.yy.2
Even if the disambiguation rule is not triggered, clients should warn when the target architecture satisfies
one of the conditions above (or a similar condition). See Section 5.3.16 for a description of this and other
conditions for which a warning should be issued.
Disambiguation of variables and operation terms. In unscheduled code, operation terms cannot be
confused with variables. The special RF names ‘r’, ‘f’ and ‘b’ are reserved, respectively, to integer,
floating-point and Boolean register files of the universal machine. The assembler does not allow any
operation to have one of these names. PENDING: disam-
biguation of mixed TTA
code ??
Example. Unambiguous move term accessing a variable. The following move is interpreted as “copy
constant 55 to variable with index 2 of variable pool ‘r”’. There cannot exist an operation ‘r’, so the
interpretation of the move destination as operation term is impossible.
55 -> r.2
Disambiguation of Register File and Immediate Unit names Assembler syntax does not differentiate
unit names of immediate units from unit names of register files. The same register specifier of a move
source
x.2 -> alu.add.1
can represents a GPR or an immediate register depending on whether ‘x’ is an RF or a IU.
In this case the GPR interpretation is always preferred over the IU interpretation. However using the same
naming for IUs and GPRs restricts severely the programmability of target machine and is not encouraged.
5.4 Program Image Generator (PIG)
Program Image Generator (PIG) generates the bit image which can be uploaded to the target machine’s
memory for execution. Compression can be applied to the instruction memory image by means of in-
struction compression algorithm plugins.
version 31 2016-02-23 93/145
![](asset-5e.png)
TTA-based Co-design Environment v1.14 User Manual
Input: TPEF, BEM, ADF
Output: program bit image in alternative formats
5.4.1 Usage
The usage of the generatebits application is as follows:
generatebits <options> ADF
The possible options of the application are as follows:
Short
Name
Long Name Description
b bem The binary encoding map file.
c compressor Name of the code compressor plugin file.
u compressorparam Parameter to the code compressor in form ’name=value’.
d dataimages Creates data images.
g decompressor Generates a decompressor block.
o diformat The output format of data image(s) (’ascii’, ’array’, ’mif’ or ’binary’). Default
is ’ascii’.
w dmemwidthinmaus Width of data memory in MAUs. Default is 1.
x hdl-dir Directory root where are the ProGe generated HDL files generated by ProGe.
If given, PIG will write imem_mau_pkg and compressor in that directory. Oth-
erwise they are written to cwd.
f piformat Determines the output format of the program image and data images. Value
may be ’ascii’, ’binary’ or ’mif’.
s showcompressors Shows the compressor plugin descriptions.
p program The TPEF program file(s).
v verbose Display extra information during image creation: total immediates, different
immediate values, instr. addresses, data addresses, full instruction NOPs and
consecutive full instr. NOP groups.
The application prints the program image to the standard output stream. It can be easily forwarded to a
file, if wanted. The data images are generated to separate files, one for each address space. Names of the
files are determined by the name of the address space and the suffix is .img. The files are generated to the
directory in which the application is executed.
The binary encoding map input parameter may be omitted. If so, the BEM to be used is generated
automatically. The BEM is used in the intermediate format of the program image, before it is compressed
by the code compressor plugin. However, if no code compression is applied, the program image output
matches the format defined in the BEM.
The options can be given either using the short name or long name. If short name is used, a hyphen (-)
prefix must be used. For example -a followed by the name of the ADF file. If the long name is used, a
double hyphen (- -) prefix must be used, respectively.
5.4.1.1 An example of the usage
The following example generates a program image and data images of the address spaces in ASCII format
without code compression.
generatebits -b encodings.bem -p program.tpef -f ascii -d machine.adf
5.4.2 Dictionary Compressor
TCE toolset includes two different code compressors: ’InstructionDictionary’ compressor and ’MoveS-
lotDictionary’ compressor. It is also possible to create new code compressors.
94/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
How these compressors work is that they analyze program’s instruction memory and create a compressed
instruction memory image. In order to use the compressed image the compressor also creates a decom-
pressor module which replaces the default decompressor in the instruction decoder unit of a processor.
You can list the available compressor and their descriptions with
generatebits -s processor.adf
5.4.2.1 Instruction Dictionary compressor
Instruction dictionary compressor analyzes a program and creates a look up table of all the unique instruc-
tions. Compressed instruction image then consists of indices to the created look up table. Decompressor
module includes the look up table and it outputs the uncompressed instruction defined by the input index.
Here is an example how to use the instruction dictionary compressor:
generatebits -c InstructionDictionary.so -g -p program.tpef processor.adf
Command creates the compressed instruction memory image ’program.img’ and ’imem_mau_pkg.vhdl’
using InstructionDictionary.so plugin defined with the -c parameter. As the command also has -g param-
eter compressor creates the ’decompressor.vhdl’ file which defines the decompressor unit.
You can also give the existing proge-output directory (in this case processor_files) to generatebits:
generatebits -c InstructionDictionary.so -g -p program.tpef -x processor_files processor.adf
Now PIG will automatically write the ’imem_mau_pkg.vhdl’ and ’decompressor.vhdl’ to the right places
in processor_files directory.
5.4.2.2 Move Slot Dictionary compressor
Move slot dictionary compressor analyzes the program and creates a separate look up table for each of
the move slots in the processor. Every move slot in the compressed instruction then has an index to its
look up table. As with the instruction dictionary compressor the decompressor module holds the look up
tables and it decompresses the instruction on move slot basis.
Here is an example how to use the move slot dictionary compressor:
generatebits -c MoveSlotDictionary.so -g -p program.tpef processor.adf
Like earlier you can also use the proge-output folder parameter:
generatebits -c MoveSlotDictionary.so -g -p program.tpef -x processor_file processor.adf
5.4.2.3 Defining New Code Compressors
By default, PIG does not apply any code compression to the program image. However, user can create a
dynamic code compressor module which is loaded and used by PIG. To define a code compressor, a C++
class derived from CodeCompressorPlugin class must be created and compiled to a shared object file.
The class is exported as a plugin using a macro defined in CodeCompressor.hh file. The buildcompressor
script can be used to compile the plugin module.
5.4.2.4 Creating the Code Compressor Module
As mentioned, the code compressor class must be derived from CodeCompressorPlugin base class. The
code compressor class must implement the virtual compress() method of the CodeCompressorPlugin
class. An example of a simple dictionary compressor is defined in compressors/simple_dictionary.cc file
in the source code distribution.
Compress Method The compress method is the heart of the compressor. The compressor method
returns the complete program image as a bit vector. The task of the method is to call the addInstruction
method of the base class sequentially to add the instructions to the program image. The base class does
the rest of the job. You just have to provide the bits of each instruction in the parameter of addInstruction
version 31 2016-02-23 95/145
TTA-based Co-design Environment v1.14 User Manual
calls. Finally, when all the instructions are added, the program image can be returned by calling the
programBits method of the base class.
The instruction bits are provided as InstructionBitVector instances. That class provides capability to mark
which bits of the instruction refer to address of another instruction and thus are likely to be changed when
the real address of the referred instruction is known. It is the responsibility of code compressor to mark
that kind of bits to the instruction bit vectors given in addInstruction calls. The base class will change the
bits when the referred instruction is added (when its address is known).
The code compressor should tell the base class which instructions must be placed in the beginning of
a MAU block. For example, the jump targets should be in the beginning of MAU. Otherwise instruc-
tion fetching will get difficult. This can be done by calling the setInstructionToStartAtBeginningOfMAU
method of the base class. If all the instructions are meant to start at the beginning of MAU, setAllInstruc-
tionsToStartAtBeginningOfMAU method is handy. By default, all the instructions are concatenated with-
out any pad bits in between them, whatever the MAU of the instruction memory is. Note that setInstruc-
tionToStartAtBeginningOfMAU /setAllInstructionsToStartAtBeginningOfMAU method(s) must be called
before starting to add instructions with addInstruction method.
Helper Methods Provided by the Base Class The CodeCompressorPlugin base class provides some
handy methods that might be useful for the code compressor implementation. The following lists the
most important ones:
•program(): Returns the POM of the program.
•tpefProgram(): Returns the TPEF of the program.
•binaryEncoding(): Returns the BEM used to encode the instructions.
•machine(): Returns the machine.
•bemBits(): Returns the program image encoded with the rules of BEM.
•bemInstructionBits(): Returns the bits of the given instruction encoded with the rules of BEM.
5.4.2.5 Building the Shared Object
When the code compressor module is coded, it must be compiled to a shared object. It can be done with
the buildcompressor script. The script takes the name of the object file to be created and the source file(s)
as command line parameters. The output of the script, assuming that everything goes fine, is the dynamic
module that can be given to the generatebits application.
5.5 TPEF Dumper (dumptpef)
TPEF Dumper is a program used for displaying information from the given TPEF.
Input: TPEF
Output: dumped data from TPEF (printed to the standard output)
5.5.1 Usage
The usage of the dumptpef application is as follows:
dumptpef <options>
The possible options of the application are as follows:
96/145 2016-02-23 version 31
![](asset-61.png)
TTA-BASED CO-DESIGN ENVIRONMENT
Short
Name
Long Name Description
f file-headers Prints the file headers.
l logical Prints only logical information. Can be used for checking if two files contain
the same program and data and connections even if it is in different order.
m mem Print information about memory usage of reserved sections.
r reloc Prints the relocation tables.
j section Prints the elements of section by section index.
s section-headers Prints the section headers.
t syms Prints the symbol tables.
version 31 2016-02-23 97/145
TTA-based Co-design Environment v1.14 User Manual
Chapter 6
CO-DESIGN TOOLS
The tools that deal with both the program and the architecture are introduced here.
6.1 Architecture Simulation and Debugging
TTA Processor Simulator simulates the process of running a TTA program on its target TTA processor.
Provides profiling, utilization, and tracing data for Explorer, Estimator and Compiler Backend. Addition-
ally, it offers debugging capabilities.
Input: TPEF, [ADF]
Output: TraceDB
There are two user interfaces to the simulating and debugging functionalities. One for the command
line more suitable for scripting, and another with more user-friendly graphical interface more suitable
for program debugging sessions. Both interfaces provide a console which supports the Tcl scripting
language.
6.1.1 Processor Simulator CLI (ttasim)
The command line user interface of TTA Simulator is called ’ttasim’. The command line user interface is
also visible in the graphical user interface in form of a console window. This manual covers the simulator
control language used to control the command line simulator and gives examples of its usage.
6.1.1.1 Usage
The usage of the command line user interface of the simulator is as follows:
ttasim <options>
In case of a parallel simulation, a machine description file can be given before giving the simulated
program file. Neither machine file or the program file are mandatory; they can also be given by means of
the simulator control language.
The possible options for the application are as follows:
98/145 2016-02-23 version 31
![](asset-63.png)
TTA-BASED CO-DESIGN ENVIRONMENT
Short
Name
Long Name Description
a adf Sets the architecture definition file (ADF).
d debugmode Start simulator in interactive "debugging mode". This is enabled by default.
Use –no-debugmode to disable.
e execute-script Executes the given string as a simulator control language script. For an exam-
ples of usage, see later in this section.
p program Sets the program to be simulated. Program must be given as a TTA program
exchange format file (.TPEF)
q quick Simulates the program as fast as possible using the compiled simulation engine.
Example: Simulating a Parallel Program Without Entering Interactive Mode The following com-
mand simulates a parallel program until the program ends, without entering the debugging mode after
simulation.
ttasim --no-debugmode -a machine.adf -p program.tpef
Example: Simulating a Program Until Main Function Without Entering Interactive Mode The
following command simulates a program until its main function and prints consumed clock cycles so far.
This is achieved by utilizing the simulator control language and the ’-e’ option, which allows entering
scripts from the command line.
ttasim --no-debugmode -e "until main; puts [info proc cycles];" -a machine.adf -p program.tpef
Using the Interactive Debugging Mode Simulator is started in debugging mode by default. In inter-
active mode, simulator prints a prompt "(ttasim)" and waits for simulator control language commands.
This example uses simulator control language to load a machine and a program, run the simulation, print
the consumed clock cycles, and quit simulation.
ttasim
(ttasim) mach machine.adf
(ttasim) prog program.tpf
(ttasim) run
(ttasim) info proc cycles
54454
(ttasim) quit
6.1.2 Fast Compiled Simulation Engine
The command line version of the Simulator, ’ttasim’, supports two different simulation engines. The de-
fault simulation engine interprets each instruction and then simulates the processor behavior accordingly.
While this is good for many cases, it can be relatively slow when compared to the computer it is being
simulated on. Therefore, the Simulator also has a highly optimized mode that uses compiled simulation
techniques for achieving faster simulation execution. In this simulation, the TTA program and machine
are compiled into a single binary plug-in file which contains functions for simulating basic blocks directly
in native machine code, allowing as fast execution as possible.
6.1.2.1 Usage
Example: Simulating a Parallel Program Using The Compiled Simulation Engine The following
command simulates a parallel program using the compiled simulation engine. (“-q”)
ttasim -a machine.adf -p program.tpef -q
version 31 2016-02-23 99/145
![](asset-64.png)
TTA-based Co-design Environment v1.14 User Manual
Currently, the behaviour of the compiled simulation can only be controlled with a limited set of Simulator
commands (such as ’stepi’, ’run’, ’until’, ’kill’). Also, the simulation runs only at an accuracy of basic
blocks so there is no way to investigate processor components between single cycles.
The following environment variables can be used to control the compiled simulation behavior:
Environment variable Description Default value
TTASIM_COMPILER Specifies the used compiler. “gcc”
TTASIM_COMPILER_FLAGS Compile flags given to the compiler. “-O0”
TTASIM_COMPILER_THREADS Number of threads used to compile. “3”
6.1.2.2 ccache
http://ccache.samba.org/
The compiled simulator can benefit quite a bit from different third party software. The first one we
describe here is a compiler cache software called ccache. Ccache works by saving compiled binary files
into a cache. When ccache notices that a file about to be compiled is the same as file found in the cache,
it simply reloads file from the cache, thus eliminating recompilation of unmodified files and saving time.
This can be very useful when running the same simulation program again, due to drastically reduced
compilation times.
6.1.2.3 distcc
http://distcc.samba.org/
Another useful tool to use together with the compiled simulator is a distributed compiler called distcc.
Distcc works by distributing the compilation of simulation engine to multiple computers and compiling
the generated source files in parallel.
After installing distcc, you can set ttasim to use the distcc compiler using the following environment
variable:
export TTASIM_COMPILER="distcc"
export TTASIM_COMPILER="ccache distcc"
Also, remember to set the amount of used threads high enough. A good number of threads to use would
be approximately the amount of CPU cores available. For example, setting 6 compiler threads can be
done like following:
export TTASIM_COMPILER_THREADS=6
6.1.3 Remote Debugger
When a TTA has been implemented to FPGA (or ASIC), ttasim can be used as a remote debug interface
to the processor. ’ttasim’ can connect to the TCE built-in debugger (under construction) with
ttasim -a machine.adf -p program.tpef -r
A convenience stub implementation for user-implemented debugging support in the TTA is given in
’tce/src/applibs/Simulator/CustomDBGController.cc’
ttasim -a machine.adf -p program.tpef -c
Integration with ’proxim’ is currently missing.
100/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
6.1.4 Simulator Control Language
This section describes all the Simulator commands that can be entered when the Simulator runs in debug
mode. The Simulator displays a new line with the prompt string only when it is ready to accept new
commands (the simulation is not running). The running simulation can be interrupted at any time by the
key combination CTRL-c. The simulator stops simulation and prompts the user for new commands as if
it had been stopped by a breakpoint.
The Simulator control language is based on the Toolset Control Language. It extends the predefined set
of Tcl commands with a set of commands that allow to perform the functions listed above. In addition
to predefined commands, all basic properties of Tcl (expression evaluation, parameter substitution rules,
operators, loop constructs, functions, and so on) are supported.
6.1.4.1 Initialization
When the Simulator is run in debug mode, it automatically reads and executes the initialization command
file ‘.ttasim-init’ if found in the user home directory. The ‘.ttasim-init’ file allows user to define specific
simulator settings (described in section 6.1.4.2) which are enabled everytime ttasim is executed.
After the initialization command sequence is completed, the Simulator processes the command line op-
tions, and then reads the initialization command file with the same name in current working directory.
After it has processed the initialization files and the command line options, the Simulator is ready to
accept new commands, and prompts the user for input. The prompt line contains the string ‘(ttasim)’.
6.1.4.2 Simulation Settings
Simulation settings are inspected and modified with the following commands.
setting variable value Sets a new value of environment variable variable.
setting variable Prints the current value contained by environment variable variable.
setting Prints all settings and their current values.
Currently, the following settings are supported.
bus_trace boolean Enables writing of the bus trace. Bus trace stores values written to each bus in each
simulated clock cycle.
execution_trace boolean Enables writing of the basic execution trace. Basic execution trace stores the
address of the executed instruction in each simulated clock cycle.
history_filename string The name of the file to store the command history, if command history saving
is enabled.
history_save boolean Enables saving command history to a file.
history_size integer Maximum count of last commands stored in memory. This does not affect writing
of the command history log, all commands are written to the log if logging is enabled.
next_instruction_printing boolean Print the next executed instruction when simulation stops, for ex-
ample, after single-stepping or at a breakpoint.
procedure_transfer_tracking boolean Enables procedure transfer tracking. This trace can be used to
easily observe which procedures were called and in which order. The trace is saved in ’proce-
dure_transfer’ table of Trace DB. This information could be derived from ’execution_trace’, but
simulation is very slow when it is enabled, this type of tracking should be faster.
profile_data_saving boolean Save program profile data to trace database after simulation.
version 31 2016-02-23 101/145
TTA-based Co-design Environment v1.14 User Manual
rf_tracking boolean Enables concurrent register file access tracking. This type of tracking makes the
simulation speed much worse, so it is not enabled by default. The produced statistics can be
browsed after simulation by using the command ’info proc stats’.
simulation_time_statistics boolean Prints time statistics for the last command ran (run, until, nexti,
stepi).
simulation_timeout integer Stops the simulation after specified timeout. Value of zero means no time-
out.
static_compilation boolean Switch between static and dynamic compilation when running compiled
simulation.
utilization_data_saving boolean Save processor utilization data to trace database after simulation.
6.1.4.3 Control of How the Simulation Runs
The commands described in this section allow to control the simulation process.
Before simulation can start, a program must be loaded into the Simulator. If no program is loaded, the
command run causes the following message:
Simulation not initialized.
run Starts simulation of the program currently loaded into the Simulator. The program can be loaded by
prog command (see Section 6.1.4.6) or may be given directly as argument, on the command line.
Simulation runs until either a breakpoint is encountered or the program terminates.
resume [count]Resume simulation of the program until the simulation is finished or a breakpoint is
reached. The count argument gives the number of times the continue command is repeated, that is,
the number of times breakpoints should be ignored.
stepi [count]Advances simulation to the next machine instructions, stepping into the first instruction a
new procedure if a function call is simulated. The count argument gives the number of machine
instruction to simulate.
nexti [count]Advances simulation to the next machine instructions in current procedure. If the instruc-
tion contains a function call, simulation proceeds until control returns from it, to the instruction
past the function call. The count argument gives the number of machine instruction to simulate.
until [arg]Continue running until the program location specified by arg is reached. Any valid argument
that applies to command break (see Section 6.1.4.5) is also a valid argument for until. If the
argument is omitted, the implied program location is the next instruction. In practice, this command
is useful when simulation control is inside a loop and the given location is outside it: simulation
will continue for as many iterations as required in order to exit the loop (and reach the designated
program location).
kill Terminate the simulation. The program being simulated remains loaded and the simulation can be
restarted from the beginning by means of command run. The Simulator will prompt the user for
confirmation before terminating the simulation.
quit This command is used to terminate simulation and exit the Simulator.
6.1.4.4 Examining and modifying Program Code and Data
The Simulator allows to examine and modify the program being simulated and the data it uses
x [/anfu][addr]This low-level command prints the data in memory starting at specified addresses addr.
The optional parameters nand uspecify how much memory to display and how to format it.
102/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
aParameter [/a address_space] can be used to select the address space if there are multiple address
spaces in the target machine.
nRepeat count: how many data words (counting by units u) to display. If omitted, it defaults to 1.
fTarget filename. Setting this causes the memory contents to be printed as binary data to the given
file.
uUnit size: ‘b’ (MAU, a byte in byte-addressed memories), ‘h’ (double MAU), ‘w’ (quadruple
word, a ‘word’ in byte-addressed 32-bit architectures). The unit size is ignored for formats
‘s’ and ‘i’.
If addr is omitted, then the first address past the last address displayed by the previous xcommand
is implied. If the value of nor uis not specified, the value given in the most recent xcommand is
maintained.
The values printed by command xare not entered in the value history (see Section 6.1.4.9).
load_data [/a address space]address file [size]Reads binary data from filename to the specified address
in memory. Optional parameter /a address_space can be used to select the address space if there
are multiple address spaces in the target machine. Optional parameter size specifies read size in
bytes.
symbol_address datasym Returns the address of the given data symbol (usually a global variable).
disassemble [addr1 [addr2]] Prints a range of memory addresses as machine instructions. When two
arguments addr1,addr2 are given, addr1 specifies the first address of the range to display, and
addr2 specifies the last address (not displayed). If only one argument, addr1, is given, then the
function that contains addr1 is disassembled. If no argument is given, the default memory range is
the function surrounding the program counter of the selected frame.
6.1.4.5 Control Where and When to Stop Program Simulation
A breakpoint stops the simulation whenever the Simulator reaches a certain point in the program. It
is possible to add a condition to a breakpoint, to control when the Simulator must stop with increased
precision. There are two kinds of breakpoints: breakpoints (proper) and watchpoints. A watchpoint is a
special breakpoint that stops simulation as soon as the value of an expression changes.
where num is a unique number that identifies the breakpoint or watchpoint and description describes the
properties of the breakpoint. The properties include: whether the breakpoint must be deleted or disabled
after it is reached; whether the breakpoint is currently disabled; the program address of the breakpoint, in
case of a program breakpoint; the expression that, when modified by the program, causes the Simulator
to stop, in case of a watchpoint.
bp address Sets a breakpoint at address address. Argument can also be a code label such as global
procedure name (e.g. ’main’).
bp args if Sets a conditional breakpoint. The arguments args are the same as for unconditional break-
points. After entering this command, Simulator prompts for the condition expression. Condition is
evaluated each time the breakpoint is reached, and the simulation only when the condition evaluates
as true.
tbp args Sets a temporary breakpoint, which is automatically deleted after the first time it stops the
simulation. The arguments args are the same as for the bp command. Conditional temporary
breakpoints are also possible (see command condition below).
watch Sets a watchpoint for the expression expr. The Simulator will stop when the value of given
expression is modified by the program. Conditional watchpoints are also possible (see command
condition below).
version 31 2016-02-23 103/145
TTA-based Co-design Environment v1.14 User Manual
condition [num] [expr]Specifies a condition under which breakpoint num stops simulation. The Simu-
lator evaluates the expression expr whenever the breakpoint is reached, and stops simulation only
if the expression evaluates as true (nonzero). The Simulator checks expr for syntactic correctness
as the expression is entered.
When condition is given without expression argument, it removes any condition attached to the
breakpoint, which becomes an ordinary unconditional breakpoint.
ignore [num] [count]Sets the number of times the breakpoint num must be ignored when reached. A
count value zero means that the breakpoint will stop simulation next time it is reached.
enablebp [delete|once] [num ...] Enables the breakpoint specified by num. If once flag is specified, the
breakpoint will be automatically disabled after it is reached once. If delete flag is specified, the
breakpoint will be automatically deleted after it is reached once.
disablebp [num ...] Disables the breakpoint specified by num. A disabled breakpoint has no effect, but
all its options (ignore-counts, conditions and commands) are remembered in case the breakpoint is
enabled again.
deletebp [num ...] Deletes the breakpoint specified by num. If no arguments are given, deletes all
breakpoints currently set, asking first for confirmation.
info breakpoints [num]Prints a table of all breakpoints and watchpoints. Each breakpoint is printed in
a separate line. The two commands are synonymous.
6.1.4.6 Specifying Files and Directories
The Simulator needs to know the file name of the program to simulate/debug and, usually, the Architecture
Definition File (ADF) that describes the architecture of the target processor on which the program is going
to run.
prog [filename]Load the program to be simulated from file filename. If no directory is specified with
set directory, the Simulator will search in the current directory.
If no argument is specified, the Simulator discards any information it has on the program.
mach [filename]Load the machine to be simulated from file filename. If no directory is specified with
set directory, the Simulator will search in the current directory.
In case a parallel program is tried to be simulated without machine, an error message is printed and
simulation is terminated immediately. In some cases the machine file can be stored in the TPEF
file.
conf [filename]Load the processor configuration to be simulated from file filename. If no directory is
specified with set directory, the Simulator will search in the current directory.
Simulator expects to find the simulated machine from the processor configuration file. Other set-
tings are ignored. This can be used as replacement for the mach command.
6.1.4.7 Examining State of Target Processor and Simulation
The current contents of any programmer visible state, which includes any programmable register, bus,
or the last data word read from or written to a port, can be displayed. The value is displayed in base
10 to allow using it easily in Tcl expressions or conditions. This makes it possible, for example, to set
a conditional breakpoint which stops simulation only if the value of some register is greater than some
constant.
info proc cycles Displays the total execution cycle count and the total stall cycles count.
info proc mapping Displays the address spaces and the address ranges occupied by the program: address
space, start and end address occupied, size.
104/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
info proc stats In case of parallel simulation, displays current processor utilization statistics. In case
’rf_tracking’ setting is enabled and running parallel simulation, also lists the detailed register file
access information.
info regfiles Prints the name of all the register files of the target processor.
info registers regfile [regname]Prints the value of register regname in register file regfile, where regfile
is the name of a register file of the target processor, and regname is the name of a register that
belongs to the specified register file.
If regname is omitted, the value of all registers of the specified register file is displayed.
info funits Prints the name of all function units of the target processor.
info iunits Prints the name of all immediate units of the target processor.
info immediates iunit [regname]Prints the value of immediate register regname in immediate unit iunit,
where iunit is the name of an immediate unit of the target processor, and regname is the name of a
register that belongs to the specified unit.
If regname is omitted, the value of all registers of the specified immediate unit is displayed.
info ports unit [portname]Prints the last data word read from or written to port portname of unit unit,
where unit may be any function unit, register file or immediate unit of the target processor. The
value of the data word is relative to the selected stack frame.
If portname is omitted, the last value on every port of the specified unit is displayed.
info busses [busname]Displays the name of all bus segments of transport bus busname. If the argument
is omitted, displays the name of the segments of all busses of the target processor.
info segments bus [segmentname]] Prints the value currently transported by bus segment segmentname
of the transport bus busname.
If no segment name is given, the Simulator displays the contents of all segments of transport bus
bus.
info program Displays information about the status of the program: whether it is loaded or running,
why it stopped.
info program is_instruction_reference ins_addr move_index Returns 1 if the source of the given move
refers to an instruction address, 0 otherwise.
info stats executed_operations Prints the total count of executed operations.
info stats register_reads Prints the total count of register reads.
info stats register_writes Prints the total count of register writes.
6.1.4.8 Miscellaneous Support Commands and Features
The following commands are facilities for finer control on the behaviour of the simulation control lan-
guage.
help [command ] Prints a help message briefly describing command command. If no argument is given,
prints a general help message and a listing of supported commands.
version 31 2016-02-23 105/145
TTA-based Co-design Environment v1.14 User Manual
6.1.4.9 Command and Value History Logs
All commands given during a simulation/debugging session are saved in a command history log. This
forms a complete log of the session, and can be stored or reloaded at any moment. By loading and running
a complete session log, it is possible to resume the same state in which the session was saved.
It is possible to run a sequence of commands stored in a command file at any time during simulation in
debug mode using the source command. The lines in a command file are executed sequentially and are
not printed as they are executed. An error in any command terminates execution of the command file.
commands [num ]Displays the last num commands in the command history log. If the argument is
omitted, the num value defaults to 10.
source filename Executes the command file filename.
6.1.5 Traces
Simulation traces are stored in a SQLite 3 binary file and multiple pure ascii files. The SQLite file
is named after the program file by appending ’.trace’ to its end. The additional trace files append yet
another extension to this, such as ’.calls’ for the call profile and ’.profile’ for the instruction execution
profile. The SQLite file can be browsed by executing SQL queries using the sqlite client and the pure text
files can be browsed using any text viewer/editor.
By default simulation traces are dumped in the same directory as loaded program file. It is possible to
override that directory by setting an environment variable TTASIM_TRACE_DIR pointing to desired
location.
6.1.5.1 Profile Data
The simulator is able to produce enough data to provide an inclusive call profile. In order to visualize this
data in a call graph, the kcachegrind GUI tool can be used.
First, produce a trace by running the following commands before initializing the simulation by loading a
machine and a program:
setting profile_data_saving 1
setting procedure_transfer_tracking 1
It’s recommend to produce an assembly file from the program to make the profile data contain information
about the program and the kcachegrind able to show the assembly lines for the cost data:
tcedisasm -F mymachine.adf myprogram.tpef
After this, the simulation should collect the information necessary to build a kcachegrind compatible
trace file. The file can be produced with an helper script shipped with TCE as follows:
generate_cachegrind myprogram.tpef.trace
This command generates a file myprogram.tpef.trace.cachegrind which can be loaded to the kcachegrind
for visualized inclusive call profile:
kcachegrind myprogram.tpef.trace.cachegrind
Alternatively, the call profile can be dumped to the command line using the ’callgrind_annotate’ tool from
the ’valgrind’ package:
callgrind_annotate myprogram.tpef.trace.cachegrind --inclusive=yes
In case –inclusive=yes is not given, exclusive call profile is printed. Exclusive profile shows the cycles
for each function itself and does not include the cost from the called functions.
106/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
6.1.6 Processor Simulator GUI (Proxim)
Processor Simulator GUI (Proxim) is a graphical frontend for the TTA Processor Simulator.
6.1.6.1 Usage
This section is intended to familiarize the reader to basic usage of Proxim. This chapter includes instruc-
tions to accomplish only the most common tasks to get the user started in using the Simulator GUI.
The following windows are available:
•Machine State window Displays the state of the simulated processor.
•Disassembly window for displaying machine level source code of the simulated application.
•Simulator console for controlling the simulator using the simulator control language.
•Simulation Control Window: Floating tool window with shortcut buttons for items in the Program
menu.
Console Window Textual output from the simulator and all commands sent to the simulator engine are
displayed in the Simulator Console window, as well as the input and output from the simulated program.
Using the window, the simulator can be controlled directly with Simulator Control Language accepted
also by the command line interface of the simulator. For list of available commands, enter ’help’ in the
console.
Most of the commands can be executed using graphical dialogs and menus, but the console allows faster
access to simulator functionality for users familiar with the Simulator Control Language. Additionally,
all commands performed using the GUI are echoed to the console, and appended to the console command
history.
The console keeps track of performed commands in command history. Commands in the command
history can be previewed and reused either by selecting the command using up and down arrow keys in
the console window, or by selecting the command from the Command History.
The Command menu in the main window menubar contains all GUI functionality related to the console
window.
Simulation Control Window Running simulation can be controlled using the Simulation Control
window.
Consequences of the window buttons are as follows:
•Run/Stop: If simulation is not running, the button is labeled ’Run’, and it starts simulation of the
program loaded in the simulator. If simulation is running, the button is labled ’Stop’, and it will
stop the simulation.
•Stepi: Advances simulation to the next machine instructions.
•Nexti: Advances simulation to the next machine instructions in current procedure.
•Continue: Resumes simulation of the program until the simulation is finished or a breakpoint is
reached.
•Kill: Terminates the simulation. The program being simulated remains loaded and the simulation
can be restarted from the begining.
version 31 2016-02-23 107/145
TTA-based Co-design Environment v1.14 User Manual
Disassembly Window The disassembly window displays the machine code of the simulated program.
The machine code is displayed one instruction per line. Instruction address and instruction moves are
displayed for each line. Clicking right mouse button on an instruction displays a context menu with the
following items:
•Toggle breakpoint: Sets a breakpoint or deletes existing breakpoint at the selected instruction.
•Edit breakpoint...: Opens selected breakpoint in Breakpoint Properties dialog.
Machine State Window The Machine State Window displays the state of the processor running the
simulated program. The window is split horizontally to two subwindows. The window on the left is
called Status Window, and it displays general information about the state of the processor, simulation and
the selected processor block. The subwindow on the right, called Machine Window, displays the machine
running the simulation.
The blocks used by the current instruction are drawn in red color. The block utilization is updated every
time the simulation stops.
Blocks can be selected by clicking them with LMB. When a block is selected, the bottom of the status
window will show the status of the selected block.
6.1.6.2 Profiling with Proxim
Proxim offers simple methods for profiling your program. After you have executed your program you
can select “Source” -> “Profile data” -> “Highlight top execution count” from the top menu. This opens
a dialog which shows execution counts of various instruction address ranges. The list is arranged in
descending order by the execution count.
If you click a line on the list the disassembly window will focus on the address range specified on that
line. You can trace in which function the specific address range belongs to by scrolling the disassembly
window up until you find a label which identifies the function. You must understand at least a little about
assembly coding to find the actual spot in C code that produces the assembly code.
6.2 System Level Simulation with SystemC
TCE provides easy “hooks” to attach cycle-count accurate TTA simulation models to SystemC system
level simulations. The hooks allow instantiating TTA cores running a fixed program as SystemC modules.
In order to simulate I/O from TTA cores, or just to model the functionality of an function unit (FU) in
more detail, the fast (but not necessarily bit accurate) default pipeline simulation models of the FUs can
be overriden with as accurate SystemC models as the system level simulation requires.
The TCE SystemC integration layer is implemented in “tce_systemc.hh” which should be included in
your simulation code to get access to the TTA simulation hooks. In addition, to link the simulation binary
successfully, the TCE libraries should be linked in via “-ltce” or a similar switch.
For a full example of a system level simulation with multiple TTA cores, see Appendix B.
6.2.1 Instantiating TTACores
New TTA cores are added to the simulation by instantiating objects of class TTACore. The constructor
takes three parameters: the name for the TTA core SystemC module, the file name for the architecture
description, and the program to be loaded to the TTA.
For example:
...
#include <tce_systemc.hh>
...
int sc_main(int argc, char* argv[]) {
108/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
...
TTACore tta("tta_core_name", "processor.adf", "program.tpef");
tta.clock(clk.signal());
tta.global_lock(glock);
...
}
As you can see, the TTA core model presents only two external ports that must be connected in your
simulation model: the clock and the global lock. The global lock freezes the whole core when it’s up and
can be used in case of dynamic latencies, for example.
6.2.2 Describing Detailed Operation Pipeline Simulation Models
The default TTA simulation model is a model that produces cycle-count accuracy but does not simulate
the details not visible to the programmer. It’s an architecture simulator which is optimized for simulation
speed. However, in case the core is to be connected to other hardware blocks in the system, the input and
output behavior must be modeled accurately.
The TCE SystemC API provides a way to override functional unit simulation models with more detailed
SystemC modules. This is done by describing one or more “operation simulation models” with the macro
TCE_SC_OPERATION_SIMULATOR and defining more accurate simulation behavior for operation
pipelines.
In the following, a simulation model to replace the default load-store simulation model is described. This
model simulates memory mapped I/O by redirecting accesses outside the data memory address space of
the internal TTA memory to I/O registers.
TCE_SC_OPERATION_SIMULATOR(LSUModel) {
sc_in<int> reg_value_in;
sc_out<int> reg_value_out;
sc_out<bool> reg_value_update;
TCE_SC_OPERATION_SIMULATOR_CTOR(LSUModel) {}
TCE_SC_SIMULATE_CYCLE_START {
reg_value_update = 0;
}
TCE_SC_SIMULATE_STAGE {
unsigned address = TCE_SC_UINT(1);
// overwrite only the stage 0 simulation behavior of loads and
// stores to out of data memory addresses
if (address <= LAST_DMEM_ADDR || TCE_SC_OPSTAGE > 0) {
return false;
}
// do not check for the address, assume all out of data memory
// addresses update the shared register value
if (TCE_SC_OPERATION.writesMemory()) {
int value = TCE_SC_INT(2);
reg_value_out.write(value);
reg_value_update.write(1);
} else { // a load, the operand 2 is the data output
int value = reg_value_in.read();
TCE_SC_OUTPUT(2) = value;
}
return true;
}
version 31 2016-02-23 109/145
TTA-based Co-design Environment v1.14 User Manual
};
In the above example, TCE_SC_SIMULATE_CYCLE_START is used to describe behavior that is
produced once per each simulated TTA cycle, before any of the operation stages are simulated. In this
case, the update signal of the I/O register is initialized to 0 to avoid garbage to be written to the register
in case write operation is not triggered at that cycle.
TCE_SC_SIMULATE_STAGE is used to define the parts of the operation pipeline to override. The code
overrides the default LSU operation stage 0 in case the address is greater than LAST_DMEM_ADDR
which in this simulation stores the last address in the TTA’s local data memory. Otherwise, it falls back
to the default simulation model by returning false from the function. The default simulation behavior
accesses the TTA local memory simulated with the TTACore simulation model like in a standalone TTA
simulation. Returning true signals that the simulation behavior was overridden and the default behavior
should not be produced by TTACore.
The actual simulation behavior code checks whether the executed operation is a memory write. In that
case it stores the written value to the shared register and enables its update signal. In case it’s a read (we
know it’s a read in case it’s not a write as this is a load-store unit with only memory accessing operations),
it reads the shared register value and places it in the output queue of the functional unit.
In more detail: TCE_SC_OUTPUT(2) = value instructs the simulator to write the given value to the
functional unit port bound to the operand 2 of the executed operation (in this case operand 2 of a load
operation is the first result operand. This follows the convention of OSAL operation behavior models (see
Section 4.3.5 for further details). Similarly, TCE_SC_UINT(2) and TCE_SC_INT(1) are used to read
the value written to the operand 2 and 1 of the operation as unsigned and signed integers, respectively. In
case of the basic load/store operations, operand 1 is the memory address and in case of stores, operand 2
is the value to write.
Finally, the simulation model is instantiated and the original LSU simulation model of TTACore is re-
placed with the newly defined one:
...
LSUModel lsu("LSU");
tta.setOperationSimulator("LSU", lsu);
...
6.3 Processor Cost/Performance Estimator (estimate)
Processor Cost/Performance Estimator provides estimates of energy consumption, die area, and maxi-
mum clock rate of TTA designs (program and processor combination).
Input: ADF, IDF, [TPEF and TraceDB]
Output: Estimate (printed to the standard output)
6.3.1 Command Line Options
The usage of the estimate application is as follows:
estimate {-p [TPEF] -t [TraceDB]} ADF IDF
The possible options of the application are as follows:
110/145 2016-02-23 version 31
![](asset-6f.png)
TTA-BASED CO-DESIGN ENVIRONMENT
Short
Name
Long Name Description
p program Sets the TTA program exchange format file (TPEF) from which to load the
estimated program (required for energy estimation only).
t trace sets the simulation trace database (TraceDB) from which to load the simulation
data of the estimated program (required for energy estimation only).
a total-area Runs total area estimation.
l longest-path Runs longest path estimation.
e total-energy Runs total energy consumption estimation.
If
tpef
and
tracedb
are not given, the energy estimation will NOT be performed since the Estimator
requires utilization information about the resources. However, the area and timing estimation will be
done. If only one of
tracedb
and
tpef
is given, it is ignored.
6.4 Automatic Design Space Explorer (explore)
Automatic Design Space Explorer automates the process of searching for target processor configurations
with favourable cost/performance characteristics for a given set of applications by evaluating hundreds or
even thousands of processor configurations.
Input: ADF (a starting point architecture), TPEF, HDB
Output: ExpResDB (Section 2.2.8)
6.4.1 Explorer Application format
Applications are given as directories that contain the application specific files to the Explorer. Below is a
description of all the possible files inside the application directory.
file name Description
program.bc The program byte code (produced using “tcecc –emit-llvm”).
description.txt The application description.
simulate.ttasim TTASIM simulation script piped to the TTASIM to produce
ttasim.out file. If no such file is given the simulation is started
with "until 0" command.
correct_simulation_output Correct standard output of the simulation used in verifying. If
you use this verification method, you need to add verification
printouts to your code or the simulation script and produce this
file from a successful run.
max_runtime The applications maximum runtime (as an ascii number in
nanoseconds).
setup.sh Simulation setup script, if something needs to be done before
simulation.
verify.sh Simulation verify script for additional verifying of the simula-
tion, returns 0 if OK. If missing only correct_simulation_output
is used in verifying.
Below is an example of the file structure of HelloWorld application. As The maximum runtime file is
missing application is expected not to have a maximum runtime requirement.
HelloWorld/program.bc
HelloWorld/correct_simulation_output
HelloWorld/description.txt
6.4.2 Command Line Options
The exploration result database <output_dsdb> is required always. The database can be queried applica-
tions can be added into and removed from the database and the explored configurations in the database
version 31 2016-02-23 111/145
TTA-based Co-design Environment v1.14 User Manual
can be written as files for further examination.
Please refer to explroe -h for a full listing of possible options.
Depending on the exploration plugin, the exploring results machine configurations in to the exploration
result database dsdb. The best results from the previous exploration run are given at the end of the
exploration:
explore -e RemoveUnconnectedComponents -a data/FFTTest -hdb=data/initial.hdb data/test.dsdb
Best result configurations:
1
Exploration plugins may also estimate the costs of configurations with the available applications. If there
are estimation results for the configuratios those can be queried with option –conf_summary by giving
the ordering of the results.
The Explorer plugins explained in chapters below can be listed with a command:
explore -g
And their parameters with a command:
explore -p <plugin name>
These commands can help if, for some reason, this documentation is not up-to-date.
6.4.3 Explorer Plugin: ConnectionSweeper
ConnectionSweeper reduces the interconnection network gradually until a given cycle count worsening
threshold is reached. The algorithm tries to reduce register file connections first as they tend to be more
expensive.
Example:
explore -v -e ConnectionSweeper -u cc_worsening_threshold=10 -s 1 database.dsdb
This reduces the connections in the IC network starting from the configuration number 1 until the cycle
count drops over 10%. This algorithm might take quite a while to finish, thus the verbose switch is
recommended to print the progress and possibly to pick up interesting configurations during the execution
of the algorithm.
The explore tool has a pareto set finder which can output the interesting configurations from the DSDB
after the IC exploration. The pareto set can be printed with:
explore -pareto_set C database.dsdb
This prints out the pareto efficient configurations using the number of connections in the architecture
and the execution cycle count as the quality measures. For visualizing the pareto set you can use the
pareto_vis script (installed with TCE) to plot the configurations:
explore -pareto_set C database.dsdb | pareto_vis
6.4.4 Explorer Plugin: SimpleICOptimizer
SimpleICOptimizer is an explorer plugin that optimizes the interconnection network of the given config-
uration by removing the connections that are not used in the parallel program.
This is so useful functionality especially when generating ASIPs for FPGAs that there’s a shortcut script
for invoking the plugin.
Usage:
minimize-ic unoptimized.adf program.tpef target-ic-optimized.adf
However, if you want more customized execution, you should read on.
Parameters that can be passed to the SimpleICOptimizer are:
112/145 2016-02-23 version 31
![](asset-71.png)
TTA-BASED CO-DESIGN ENVIRONMENT
Param Name Default Value Description
tpef no default value name of the scheduled program file
add_only false Boolean value. If set true the connections of the given configu-
ration won’t be emptied, only new ones may be added
evaluate true Boolean value. True evaluates the result config.
If you pass a scheduled tpef to the plugin, it tries to optimize the configuration for running the given pro-
gram. If multiple tpefs are given, the first one will be used and others discarded. Plugin tries to schedule
sequential program(s) from the application path(s) defined in the dsdb and use them in optimization if
tpef is not given.
Using the plugin requires user to define the configuration he wishes optimize. This is done by giving -s
<configuration_ID> option to the explorer.
Let there be 2 configurations in database.dsdb and application directory path app/. You can optimize the
first configuration with:
explore -e SimpleICOptimizer -s 1 database.dsdb
If the optimization was succesfull, explorer should output:
Best result configuration:
3
Add_only option can be used for example if you have an application which isn’t included in application
paths defined in database.dsdb but you still want to run it with the same processor configuration. First
export the optimized configuration (which is id 3 in this case):
explore -w 3 database.dsdb
Next schedule the program:
schedule -t 3.adf -o app_dir2/app2.scheduled.tpef app_dir2/app2.seq
And then run explorer:
explore -e SimpleICOptimizer -s 3 -u add_only=true -u tpef=app_dir2/app2.scheduled.tpef
database.dsdb
The plugin now uses the optimized configuration created earlier and adds connections needed to run the
other program. If the plugin finds a new configuration it will be added to the database, otherwise the
existing configuration was already optimal. Because the plugin won’t remove existing connections the
new machine configuration is able to run both programs.
6.4.5 Explorer Plugin: RemoveUnconnectedComponents
Explorer plugin that removes unconnected ports from units or creates connections to these ports if they
are FUs, but removes FUs that have no connections. Also removes unconnected buses. If all ports from a
unit are removed, also the unit is removed.
You can pass a parameter to the plugin:
Param Name Default Value Description
allow_remove false Allows the removal of unconnected ports and FUs
When using the plugin you must define the configuration you wish the plugin to remove unconnected
components. This is done by passing -s <configuration_ID> to explorer.
If you do not allow removal the plugin will connect unconnected ports to some sockets. It can be done
with:
explore -e RemoveUnconnectedComponents -s 3 database.dsdb
or
explore -e RemoveUnconnectedComponents -s 3 -u allow_remove=false database.dsdb
if you wish to emphasise you do not want to remove components. This will reconnect the unconnected
ports from the configuration 3 in database.dsdb.
version 31 2016-02-23 113/145
![](asset-72.png)
TTA-based Co-design Environment v1.14 User Manual
And if you want to remove the unconnected components:
explore -e RemoveUnconnectedComponents -s 3 -u allow_remove=true database.dsdb
6.4.6 Explorer Plugin: GrowMachine
GrowMachine is an Explorer plugin that adds resources to the machine until cycle count doesn’t go down
anymore.
Parameters that can be passed to the GrowMachine are:
Param Name Default Value Description
superiority 2 Percentage value of how much faster schedules are wanted until
cycle count optimization is stopped
Using the plugin requires user to define the configuration he wishes optimize. This is done by giving -s
<configuration_ID> option to the explorer.
Example of usage:
explore -e GrowMachine -s 1 database.dsdb
6.4.7 Explorer Plugin: ImmediateGenerator
ImmediateGenerator is an Explorer plugin that creates or modifies machine instruction templates. Typical
usage is to split an instruction template slot among buses.
Parameters that can be passed to the ImmediateGenerator are:
Param Name Default Value Description
print false Print information about machines instruction templates.
remove_it_name no default value Remove instruction template with a given name
add_it_name no default value Add empty instruction template with a given name.
modify_it_name no default value Modify instruction template with a given name.
width 32 Instruction template supported width.
width_part 8 Minimum size of width per slot.
split false Split immediate among slots.
dst_imm_unit no default value Destination immediate unit.
Example of adding a new 32 width immediate template named newTemplate that is splitted among busses:
explore -e ImmediateGenerator -s 1 -u add_it_name="newTemplate" -u width=32 -u split=true
database.dsdb
6.4.8 Explorer Plugin: ImplementationSelector
ImplementationSelector is an Explorer plugin that selects implementations for units in a given configura-
tion ADF. It creates a new configuration with a IDF.
Parameters that can be passed to the ImmediateGenerator are:
Param Name Default Value Description
ic_dec DefaultICDecoder Name of the ic decoder plugin.
ic_hdb asic_130nm_1.5V.hdb name of the HDB where the implementations are
selected.
adf no default value An ADF for the implementations are selected if no
database is used.
Example of creating implementation for configuration ID 1, in the database:
explore -e ImplementationSelector -s 1 database.dsdb
114/145 2016-02-23 version 31
![](asset-73.png)
TTA-BASED CO-DESIGN ENVIRONMENT
6.4.9 Explorer Plugin: MinimizeMachine
MinimizeMachine is an Explorer plugin that removes resources from a machine until the real time re-
quirements of the applications are not reached anymore.
Parameters that can be passed to the ImmediateGenerator are:
Param Name Default Value Description
min_bus true Minimize buses.
min_fu true Minimize function units.
min_rf true Minimize register files.
frequency no default value Running frequency for the applications.
Example of minimizing configuration ID 1, in the database, with a frequency of 50 MHz:
explore -e MinimizeMachine -s 1 -u frequency=50 database.dsdb
6.4.10 Explorer Plugin: ADFCombiner
ADFCombiner is an Explorer plugin that helps creating clustered architectures. From two input architec-
tures, one (described as extra) is copied, and other one (described as node) is replicated defined number
of times.
The connections are create between register files in neighboring clusters and the extra.
In case the build_idf is set to true, hardware databases to be used for search for component implementa-
tions can be specified with -b parameter.
Parameters that can be passed to the ADFCombiner are:
Param Name Default Value Description
node_count 4 Number of times the node is replicated.
node node.adf The architecture of node that will be replicated.
extra extra.adf The architecture which will be added just once.
build_idf false If defined, ADFCombiner will try to create .idf definition file.
vector_lsu false If defined, the VectorLSGenerator plugin will be called to cre-
ate wide load store unit.
address_spaces data The semicolon separated list of address space names to be used
by generated wide load store units (one unit per address space).
Example of creating architecture with eight clusters, eight wide load and store, without creating imple-
mentation definition file:
explore -e ADFCombiner -u node=myNode.adf -u extra=myExtra.adf -u node_count=8 -u
vector_lsu=true -u address_spaces="local;global" test.dsdb
If successful, explorer will print out configuration number of architecture created in test.dsdb. The created
architecture can be written to architecture file with:
explore -w 2 test.dsdb
Example of creating architecture with four clusters with implementation definition file:
explore -e ADFCombiner -u node=myNode.adf -u extra=myExtra.adf -u node_count=4 -u
build_idf=true -b default.hdb -b stream.hdb test.dsdb
If successful, explorer will print out configuration number of architecture created in test.dsdb. The created
architecture and implementation definition file can be written to architecture file with:
explore -w 2 test.dsdb
version 31 2016-02-23 115/145
TTA-based Co-design Environment v1.14 User Manual
Chapter 7
PROCESSOR TEMPLATE
This chapter details features of the processor template of TCE. It describes the architectural overview of
the template, giving an overview of the processor paradigm used, and discusses its specifics. The chapter
also describes the programmer interface, especially from the point of view of the high-level language
compilation.
7.1 Architecture Template
7.1.1 Transport Triggered Architecture
The processor template from which the application specific processors designed with TCE are defined is
called Transport Triggered Architecture (TTA). For a detailed description behind the TTA idea, refer to
[Cor97]. A short introduction is presented in [Bou01].
TTA is based on VLIW (Very Long Instruction Word) processor paradigm but solves its major bot-
tlenecks: the complexity of the register file (RF) and register file bypass network. TTA is statically
scheduled at compile time and supports instruction-level parallelism (ILP) like VLIW. TTA instructions
are commonly hundreds of bits wide. TTAs can have multiple independent function units (FUs) and a
customized interconnection network, as illustrated in Fig. 7.1.
The term transport-triggered means that instruction words control the operand transfers on the intercon-
nection network and the operation execution happens as a side effect of these transfers. An operation is
triggered to execute when a data operand is transferred to a specific trigger input port of an FU. In the
instructions there is one slot for each transport bus. For example, “FU0.out0 -> LSU.trig.stw” moves
data from the output port 0 of function unit 0 to the trigger input port of load-store unit (LSU). After that
the LSU starts executing the triggered operation, in this case store word to memory.
In a basic case, all FU input and output ports are registers which relieves the pressure on the register
file. Thanks to the programming model, operands can be bypassed directly from one FU to another.
This is called software bypassing. Additionally, if the bypassed operand is not needed by anyone else
there is no need to write the operand to a register file. This optimization technique is called dead result
elimination. Combining software bypassing with dead result elimination helps to reduce register file
traffic. Moreover, the TCE TTA template gives freedom for partitioning register files. For example, there
can be several small and simple register files instead one centralized multiported RF.
The transport programming is also beneficial because it allows easy scalability of the architecture and
compiler, as well as supports varying pipeline depths at FUs. At the same time, the number of FU inputs
and outputs is not restricted, unlike in most processor templates which support only instructions with 2
inputs and 1 output value. User can also create instruction set extensions with a special function unit
(SFU) which can have arbitrary number of I/O operands. Instruction set extension is a powerful way of
enhancing the performance of certain applications.
116/145 2016-02-23 version 31
![](asset-75.png)
TTA-BASED CO-DESIGN ENVIRONMENT
Figure 7.1: TTA basic structure
Table 7.1: Configurable aspects of the TCE’s TTA template
Property Values Example
Functional unit Type, count 3x ALU, 2x LSU, 1x MUL, 1x ctrl...
Register file (RF) # registers, #RFs, #ports, width 16x 32b RF 2x rd + 2x wr ports, 16x 1b boolean RF
Interconnection network #buses, #sockets 5 buses, total of 43 write and 44 read sockets
Memory interfaces Count, type 2x LSU for SRAM w/ 32b data & 32b addr
Special FU User-defined functionality dct, semaphor_lock, FIFO I/O
version 31 2016-02-23 117/145
TTA-based Co-design Environment v1.14 User Manual
7.1.2 Immediates/Constants
The TTA template supports two ways of transporting program constants in instructions, such as “add
value+5”.Short immediates are encoded in the move slot’s source field, and thus consume a part of a
single move slot. The constants transported in the source field should usually be relatively small in size,
otherwise the width of a move slot is dominated by the immediate field.
Wider constants can be transported by means of so called long immediates. Long immediates can be
defined using an ADF parameter called instruction template. The instruction template defines which slots
are used for pieces of the instruction template or for defining the transports (moves). The slots cannot be
used for regular data transports when they are used for transporting pieces of a long immediate.
An instruction template defining a long immediate also provides a target to which the long immediate
must be transported. The target register resides in a so called immediate unit which is written directly
from the control unit, not through the transport buses. The immediate unit is like a register file expect
that it contains only read ports and is written only by the instruction decoder in the control unit when it
detects an instruction with a long immediate (see Fig. 7.1).
Thus, in order to support the special long immediate encoding, one has to add a) an instruction template
that supports transporting the pieces of the immediate using full move slots b) at least one long immediate
unit (a read-only register file) to which the instruction writes the immediates and of which registers the
immediates can be read to the datapath.
7.1.3 Operations, Function Units, and Operand Bindings
Due to the way TCE abstracts operations and function units, an additional concept of operand binding is
needed to connect the two in processor designs.
Operations in TCE are defined in a separate database (OSAL, Sections 2.2.6 and 4.3 ) in order to allow
defining a reusable database of “operation semantics”. The operations are used in processor designs
by adding function units (FU) that implement the wanted operations. Operands of the operations can be
mapped to different ports of the implementing FU, which affects programming of the processor. Mapping
of operation operands to the FU ports must be therefore described by the processor designer explicitly.
Example. Designer adds an FU called ’ALU’ which implements operations ’ADD’, ’SUB’, and ’NOT’.
ALU has two input ports called ’in1’ and ’in2t’ (triggering), and an output port called ’out’. A logical
binding of the ’ADD’ and ’SUB’ operands to ALU ports is the following:
ADD.1 (the first input operand) bound to ALU.in1
ADD.2 (the second input operand) bound to ALU.in2t
ADD.3 (the output operand) bound to ALU.out
SUB.1 (the first input operand) bound to ALU.in1
SUB.2 (the second input operand) bound to ALU.in2t
SUB.3 (the output operand) bound to ALU.out
However, operation ’NOT’, that is, the bitwise negation has only one input thus it must be bound to port
’FU.in2t’ so it can be triggered:
NOT.1 bound to ALU.in2t
NOT.2 (the output operand) bound to ALU.out
Because we have a choice in how we bind the ’ADD’ and ’SUB’ input operands, the binding has to
be explicit in the architecture definition. The operand binding described above defines architecturally
different TTA function unit from the following:
SUB.2 bound to ALU.in1
SUB.1 bound to ALU.in2t
SUB.3 bound to ALU.out
118/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
With the rest of the operands bound similarly as in the first example.
Due to the differing ’SUB’ input bindings, one cannot run code scheduled for the previous processor on a
machine with an ALU with the latter operand bindings. This small detail is important to understand when
designing more complex FUs, with multiple operations with different number of operands of varying size,
but is usually transparent to the basic user of TCE.
Reasons for wanting to fine tune the operand bindings might include using input ports of a smaller width
for some operation operands. For example, the width of the address operands in memory accessing
operations of a load store unit is often smaller than the data width. Similarly, the second operand of a
shift operation that defines the number of bits to shift requires less bits than the shifted data operand.
7.1.4 Datapath Connectivity Levels
The datapath interconnection network of TTAs is visible to the programmer (i.e. the compiler in practice).
This enables full customization of the connectivity, making it possible to remove connections that are
rarely, if at all, used by the programs the processor at hand is designed to run. This offers notable saving
in the HW area. However, the less connections the machine has, the more challenging it becomes to
automatically produce efficient code for it. This section describes the different TTA “connectivity levels”
and their support in the TCE design flow.
The currently identified and supported connectivity levels are, in the order of descending level of connec-
tivity, as follows:
1. Fully connected. Completely connected interconnection network “matrix”. All bus-socket and
socket-port connections are there. There is a shortcut for creating this type of connectivity in the
ProDe tool.
The easy target for the high-level language compiler tcecc. However, not a realistic design usually
due to its high implementation costs.
2. Directly reachable. The connectivity has been reduced. However, there is still at least one direct
connection from each function unit (FU) and register file (RF) output to all inputs.
An easy target for tcecc.
3. Fully RF connected. All FUs are connected to all RFs. That is, you can read and write any general
purpose register (GPR) from any FU with a single move. However, some or all bypass connections
between FUs might be missing.
An easy target for tcecc. However, reduction of bypass connections means that less software by-
passing can be done.
4. Reachable. All FUs are connected to at least one RF and all RFs (and thus other FUs) can be
reached via one or more additional register copy moves.
Compilation is fully supported by tcecc. The number of copies is not limited by tcecc. However,
this style of connectivity results in suboptimal code due to the additional register copies which
introduce additional moves, consume registers, and produce dependencies to the code which hinder
parallelism.
5. RF disconnected. Some FUs are not connected to any RF or there are “separated islands” without
connectivity to other “islands”.
Not supported by tcecc. However, any connectivity type is supported by the TCE assembler. Thus,
one can resort to manual TTA assembly programming in this case.
7.2 Programmer Interface
This section describes the programmer interface, which is the application binary interface followed by
tcecc.
version 31 2016-02-23 119/145
TTA-based Co-design Environment v1.14 User Manual
The programs produced by tcecc are fully linked and not expected to be relinked afterwards to other
programs. Therefore, details such as the function calling convention can be customized per program by
the compiler. For these parts, this section describes the current default behavior for reference.
7.2.1 Default Data Address Space Layout
When compiling from a higher-level language using tcecc, there has to be at least one byte-addressible
data address space in the machine. This is the “default address space”, marked with the numerical id 0 in
case there are multiple data address spaces.
The C/C++ compilation lays out the global variables starting from the first data memory location to the
default address space. The first location after the global variables is the start location for heap, in case the
program uses dynamic memory allocation. It grows upwards. The stack, used to store the function local
variables and to pass parameters in the function calls, grows from the largest memory address downwards.
The start address of the stack can be changed with the tcecc switch –init-sp.
As both heap and stack grow dynamically towards each other, to way to increase the available space for
heap/stack is to increase the size of your data memory address space in ADF.
7.2.2 Instruction Address Space
The default GCU assumes that instruction memory is intruction addressable. In other words the instruc-
tion word must fit in the MAU of the instruction memory. This way the next instruction can be referenced
by incrementing the current program counter by one.
How to interface the instruction memory with an actual memory chip is out of scope in TCE because
there are too many different platforms and chips and possibilities. But as an advantage this gives the user
free hands to implement almost any kind of memory hierarchy the user wants. Most probably you must
implement a memory adapter to bind the memory chip and the TTA processor interfaces together.
7.2.3 Alignment of Words in Memory
tcecc aligns the addresses of data word to the data type size in bytes. For example, if a floating-point
word takes 4 MAU’s, its address must be a multiple of 8.
The memory acessing operations in the base operation set (ldq, ldh, ldw, ldd, stq, sth, stw, and ldd) are
aligned with their size. Operations stq/ldq are for accessing single bytes, thus their alignment is 1 byte,
for sth/ldh it is 2 bytes, and for stw/ldw it is 4 bytes. Thus, one cannot assume to be able to access, for
example, a 4 byte word at address 3 using stw/ldw.
Double precision floating point word operations std/ldd which access 64-bit words are aligned at 8-byte
addresses.
The effect of misaligned word accesses on processor implementations are undefined. Typically: (1) the
processor accesses the nearest (lower) aligned address instead (because of zeroing the lower address
bits); (2) the processor halts or rises an exception (unlikely). More likely is that, a processor could enter
a slower operation mode and perform a mis-aligned memory access. However, these behaviors are not
required by the above mentioned aligned base memory operations shipped with TCE.
7.2.4 Stack Frame Layout
The register assigned to act as the stack pointer register (referred to as SPR) is the register number 0
of the first 32b register file in the ADF. There is optional frame pointer(FPR) which is enabled on only
functions with variable arrays or alloca calls. The FPR is the the third 32-bit register available in the ADF,
calculated in round-robin fashion between register files.
The default mode without FPR is called NoFP mode and the mode with FPR is called HasFP mode. On
NoFP mode the FPR is a freely usable GPR with callee-save semantics. The two stack frame modes are
function-specific, same program can contain functions which use either mode and the parameter passing
120/145 2016-02-23 version 31
![](asset-79.png)
TTA-BASED CO-DESIGN ENVIRONMENT
Figure 7.2: Stack layout(NoFP mode on left, HasFP mode on right) for a situation with two call frames,
A() calling B(). Functions like B that do not call any other functions do not store their return address to
the stack.
of the two modes is compatible in such way that code in either mode can call any function which uses
either mode.
The stack of TTA’s supported by TCE grows downwards.
The stack of a program is divided up into blocks of contiguous memory called frames. Each frame is
activated upon entering a function and is destroyed (its storage is made available for new frames) when
that function invocation returns.
In NoFP mode, the stack pointer register (SPR) always points to the place of last outgoing variable that the
function can call, and this pointer is only adjusted at the beginning and end of a function. If the function
contains calls to another functions, the first word after(below) the old stack pointer contains the old return
address. After(below) that is the area for the local variables. The local variables and spilled registers
are the same area, there is no distinction between them. After(below) the local variables is the are for
outgoing parameters. When a function calls multiple other functions, the space allocated for outgoing
parameters is calculated by the size of parameters of the call which needs most parameter area.
In HasFP mode, space for the outgoing variables is not allocated in the beginning of the function. In the
beginning of a function, Old Return Address(RA) and FPR are pushed to stack, and SPR is then copied
to FPR. Then space for local variables is allocated by decreasing the SPR. The stack pointer(SPR) always
points to the topmost item of the stack. Usually this is the end of the local variable area. When a function
is called, the function parameters are pushed into the stack, and SPR adjusted accordingly, allocating the
area for the outgoing parameters from the stack
grows downwards, therefore the incoming arguments and the local variables on the local stack frame have
positive offsets when accessed from the current function, whereas the outgoing arguments have negative
stack offsets.
The value of the return address register is pushed to stack before the function’s local variables in case the
function is not a leaf function (a function without calls to other functions). If the called function returns
an object that does not fit to the return value register, the function gets an implicit argument that points to
the location in the callee’s stack where the return value object will be written by the called function.
7.2.5 Word Byte Order
TCE-TTA processors follow the big endian byte order.
version 31 2016-02-23 121/145
TTA-based Co-design Environment v1.14 User Manual
7.2.6 Function Calling Conventions
The current default calling convention generated by tcecc is as follows.
The second 32-bit GPR (first register of second 32-bit register file, or second register of only 32-bit
register file) of ADF is a return value register which is used to return 32b values such as ints or f loats
from functions. The same register is reused for the first passed parameter value in function calls, if
suitable.
In a function call, the arguments and return values that do not fit in the 32b return value / parameter
register are pushed to stack.
The return location for the last executed CALL operation is stored in the RA register in the control unit.
This is pushed to stack in case of multiple function call layers. Refer to the stack frame layout for more
information.
7.2.7 Register Context Saving Conventions
All registers except SPR may have to be saved and restored as a result of a function call. When a register
carries a variable in the caller code that remains live across a function call, and uses a register overwritten
in the callee, it needs to be saved and restored. Registers are subdivided in two groups: those that are
saved and restored around a function call site, and those that are saved at the beginning of the called
function and are restored before it returns. The first group is termed caller-saved registers, the second
group callee-saved registers.
The compiler has the freedom to optimize the context saving convention. Depending on the properties of
the program and the capacity of the register allocator, a varying fraction of the total GPR’s in a register
file may be assigned to either above mentioned groups. The register allocator may even dedicate one
register file completely to caller-saved registers and another to callee-saved registers.
The current default register saving convention followed by tcecc is to have treat all registers except
FPR(when not used as FPR) as caller saved.
122/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
Chapter 8
PRODUCING EFFICIENT TTA
DESIGNS WITH TCE
This chapter gives some tips on how to avoid the most common bottlenecks that often arise in the TTA
designs produced using TCE and how to get good performance for your architecture.
8.1 Registers and Register Files
Make sure you have enough general purpose registers and register file read ports. Using more than two
read ports in one register file may make it big, slow and power-hungry, so consider splitting the RF to
multiple register files.
However, it should be noted that the current compiler does not perform intelligent distribution of variables
to the register files, so if the machine is not equally connected to all the function units that consume
each variable, you might end up with “connectivity register copies” (see below). The current register
distribution method is round robin which balances the RF load somewhat, but not perfectly, as it doesn’t
take in account the computation that uses each variable.
To produce better register distribution, the compiler should balance the RF load at the granularity of
“computation trees”, not at the granularity of a single register to produce more parallelizable code and to
reduce register copies. Improving this part of the code generation is being worked on with high priority.
8.2 Interconnection Network
See Section 7.1.4 for the definitions of the different connectivity classes of the TTAs designed with TCE.
Some additional points worth considering:
•’Fully connected’ on architectures that are not very small usually lead to low maximum clock
frequencies.
•Having only the ’Reachable’ connectivity leads to the need to add extra register-to-register transfers
created by the compiler and may cause considerable slowdowns due to the difficulties to parallelize
such code.
The best compromize between instructions-per-cycle (IPC) and clock speed is usually ’Fully RF con-
nected’ with additional FU to FU connections for commonly used bypasses.
This, of course, doesn’t always hold. For example, if the clock speed is limited by some other component
than the interconnect, Directly Reachable might give slightly better performance. Sometimes some ar-
chitecture with only ’Reachable’ connectivity might give so big clocks speed improvement that it might
outweight the IPC penalty. So, it needs some experimenting from the designer as it depends on the
parallelism of the program among other things.
version 31 2016-02-23 123/145
TTA-based Co-design Environment v1.14 User Manual
With some future improvements to the compiler the performance of ’Reachable’ architectures may get a
definite improvement, at which point we’ll update these hints.
There is a tool called ConnectionSweeper which can be used to optimize the interconnection network
automatically. ConnectionSweeper is documented in Section 6.4.3.
8.2.1 Negative short immediates
Make sure you have one or more buses (actually “move slots“) with signed short immediate fields. Small
negative numbers such as -1, and -4 are very common in the generated code, and long immediates often
cause bottlenecks on the machine code.
8.3 Operation Set
Optimize your operation set for your code. Make sure you support the performance-critical instructions
with hardware operations. Software-emulation of some operations is usually 50-100 times slower than
hardware execution and might prevent parallelization of multiple iterations of the core kernel loop.
For example, if you do a lot of floating point calculations, use hardware floating point units. If your inner
loop contains integer multiplications, put a multiplier in your architecture.
On the other hand, if you do not need a floating point unit, integer multiplier or a divider, do not put
them to your achitecture “just in case”, as they increase the size of the processor, power consumption,
and also the size of the instruction word. It also makes the interconnection network more complex which
often also reduces the clock speed and therefore performance. Unless, of course, you need the “general
purposity“ and you predict that some of the future programs that are going to run in your processor might
be able to exploit such hardware units.
After your selection of basic operations is good for your program, and you still need more performance
or lower power consumption, then consider using custom operations.
124/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
Chapter 9
TROUBLESHOOTING
This chapter gives solutions to common problems encountered while using TCE.
9.1 Simulation
Problems with simulation, both with command line (ttasim) and graphical user interfaces (proxim) are
listed here.
9.1.1 Failing to Load Operation Behavior Definitions
It might be possible that you have defined some custom operations to
~/.tce/opset/custom
which
conflict with your new definitions, or the simulation behaviors are compiled with an older compiler and
not compatible with your new compiler. Workaround for this is to either delete the old definitions or
rebuild them.
9.2 Limitations of the Current Toolset Version
This section lists the most user-visible limitations placed by the current toolset version.
9.2.1 Integer Width
The simulator supports only integer computations with maximum word width of 32 bits.
9.2.2 Instruction Addressing During Simulation
The details of encoding and compression of the instruction memory are not taken into account before
the actual generation of the bit image of the instruction memory. This decision was taken to allow sim-
plification in the other parts of the toolset, and to allow easy "exploration" with different encodings and
compression algorithms in the bit generation phase.
This implies that every time you see an instruction address in architectural simulation, you are actu-
ally seeing an instruction index. That is, instruction addressing (one instruction per instruction memory
address) is assumed.
We might change this in the future toolset versions to allow seeing exact instruction memory addresses
during simulation, if that is seen as a necessity. Currently it does not seem to be a very important feature.
version 31 2016-02-23 125/145
TTA-based Co-design Environment v1.14 User Manual
9.2.3 Data Memory Addressing
There is no tool to map data memory accesses in the source code to the actual target’s memories. There-
fore, you need to have a data memory which provides byte-addressing with 32-bit words. The data will be
accessed using operations LDQ, LDH, LDW, STQ, STH, STW, which access the memory in 1 (q), 2 (h),
and 4 (w) byte chunks. This should not be a problem, as it is rather easy to implement byte-addressing
in case the actual memory is of width of 2’s exponent multiple of the byte. The parallel assembler allows
any kind of minimum addressable units (MAU) in the load/store units. In that case, LDQ/STQ just access
a single MAU, etc. One should keep in mind the 32-bit integer word limitation of simulation. Thus, if the
MAU is 32-bits, one cannot use LDH or LDW because they would require 64 and 128 bits, respectively.
9.2.4 Ideal Memory Model in Simulation
The simulator assumes ideal memory model which generates no stalls and returns data for the next in-
struction. This so called ’Ideal SRAM’ model allows modeling all types of memories in the point of view
of the programmer. It is up to the load/store unit implementation to generate the lock signals in case the
memory model does not match the ideal model.
There are hooks for adding more memory models which generate lock signals in the simulation, but
for the v1.0 the simulator does not provide other memory models, and thus does not support lock cycle
simulation.
9.2.5 Guards
The guard support as specified in the ADF specification [CSJ04] is only partially supported in TCE.
‘Operators other than logical negation are not supported. That is, supported guards always “watch” a
single register (FU output or a GPR). In addition, the shipped default scheduling algorithm in compiler
backend requires a register guard. Thus, if more exotic guarded execution is required, one has to write
the programs in parallel assembly (Section 5.3).
9.2.6 Operation Pipeline Description Limitations
Even though supported by the ADF and ProDe, writing of operands after triggering an operation is not
supported neither by the compiler nor the simulator. However, setting different latencies for outputs of
multi-result operations is supported. For example, using this feature one can have an iterative operation
pipeline which computes several results which are ready after the count of stages in an iteration.
9.2.7 Encoding of XML Files
TCE uses XML to store data of the architectures and implementation locations (see Section 2.2.1 and
Section 2.2.4). The encoding of the XML files must be in 7-bit ascii. If other encodings are used, the
result is undefined.
9.2.8 Floating Point Support
The simulator supports the half (16 bits), single (32 bits) and double (64 bits) precision floating point
types. However, at this time only single precision float operations (ADDF, MULF, etc.) are selected
automatically when starting from C/C++ code. Currently, the compiler converts doubles to floats to allow
compiling and running code with doubles with reduced precision.
126/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
Appendix A
FREQUENTLY ASKED QUESTIONS
A.1 Memory Related
Questions related to memory accessing.
A.1.1 Load Store Unit
In the LSU implementations shipped with TCE the two LSB bits are used by the LSU to control a so called
write mask that handles writing of bytes. This means that the memory address outside the processor is 2
bits narrower than inside the processor. When you set the data address space width in ProDe, the width
is the address width inside the processor.
A.2 Processor Generator
A.2.0.1 Warning: Processor Generator failed to generate a test bench
The automatic testbench generator currently does not support any special function units that connect sig-
nals out from toplevel.vhdl. This warning can be ignored if you are not planning to use the automatically
generated test bench.
A.2.0.2 Warning: Opcode defined in HDB for operation ...
Processor Generator gives a warning message if the operation codes in FU are not numbered according to
the alphabetical order of the operations. The VHDL implementation and the HDB entry of the FU should
be fixed to use this kind of opcode numbering. See section 4.11.1 for more details.
A.3 tcecc
A.3.0.3 Disappearing code
Tcecc is using the efficient LLVM compiler framework for its optimizations. Sometimes LLVM is too
aggressive and removes code that you might not intend to get removed. This happens in case you write
your output to a global variable which you might dump only in the simulator and never read it in your
program nor e.g. print it with printf(). In this case LLVM sees writes to a global variable which is never
read by the program, thus thinks the computation producing the written value is useless and removes it as
dead code.
A solution to the problem is to always mark the global “output/verification” variables as ’volatile’ which
means that the content of the variable should be always written to the memory. Thus, LLVM has to assume
that there might be other readers of the same variable that are just not visible in the current program.
version 31 2016-02-23 127/145
TTA-based Co-design Environment v1.14 User Manual
A.4 Hardware characteristics
A.4.1 Interrupt support
Currently TCE does not have support for TTA interrupts. This is mostly due to the fact that context saving
is an expensive operation on TTA because the processor context can be huge. In general, TTA might not
be the best architecture choice for the only processor in a control-oriented or reactive system.
A.5 Misc
A.5.1 File Search Paths
The tools in TCE search for files such as HDBs and OSAL operation definitions, in addition to the TCE
install locations, in the current working directory and its ’data’ subdirectory.
128/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
Appendix B
SystemC Simulation Example
This SystemC simulation example simulates a system with two TTA cores communicating through two
shared registers with memory mapped access. One of the registers (busyReg) is used to denote that the
“receiver TTA” is busy processing the previously written data which the “sender TTA” writes (to dataReg).
The receiver uses iprintf to print out the received data.
The C source codes for the programs running in the two TTAs are shown below:. mem_map.h defines
constants for the memory mapped I/O addresses. This file is included by both TTA programs and the
SystemC simulation code presented later.
mmio_recv.c: The C program running in the receiver TTA:
#include <stdio.h>
#include <stdlib.h>
#include "mem_map.h"
/**
* mmio_recv.c:
*
* Polls an I/O register and prints out its value whenever it changes.
*/
int main() {
int last_value = 0;
char values_received = 0;
do {
int new_value;
*BUSY_ADDR = values_received;
// should place a barrier here to ensure the compiler doesn’t
// move the while loop above the write (both access constant
// addresses thus are trivial to alias analyze)
while ((new_value = *DATA_ADDR) == last_value);
++values_received;
iprintf("mmio_recv got %d\n", new_value);
last_value = new_value;
} while(1);
return EXIT_SUCCESS;
}
mmio_send.c: The C program running in the sender TTA.
#include <stdio.h>
#include <stdlib.h>
version 31 2016-02-23 129/145
TTA-based Co-design Environment v1.14 User Manual
#include "mem_map.h"
/**
* mmio_send.c:
*
* Write data to a memory mapped shared register. Another TTA (see
* mmio_recv.c) is polling this register and printing it to a console
* whenever its value changes.
*/
int main() {
/* These should get printed to the console through the another TTA. */
char old_values_received = 0, values_received = 0;
for (int i = 1; i <= 10; ++i) {
int new_value = 1234 * i;
*DATA_ADDR = new_value;
/* Wait until the other TTA has processed the value. This is
signalled by writing the count of values received so far
to the BUSY_ADDR. */
while ((values_received = *BUSY_ADDR) == old_values_received);
old_values_received = values_received;
}
return EXIT_SUCCESS;
}
mem_map.h: The memory mapped I/O addresses as constants:
#define LAST_DMEM_ADDR (32*1024 - 1)
#define DATA_ADDR ((volatile int*)(LAST_DMEM_ADDR + 1))
#define BUSY_ADDR ((volatile char*)(LAST_DMEM_ADDR + 1 + 4))
The simple register is defined in register.hh and the load-store unit simulation model that overrides the
default TTA simulator one in lsu_model.hh.
register.hh: SystemC model for an integer register:
#ifndef SC_REGISTER_HH
#define SC_REGISTER_HH
#include <systemc>
SC_MODULE(Register) {
sc_in<int> input;
sc_out<int> output;
sc_in<bool> updateValue;
sc_in<bool> clock;
int value;
void run() {
if (updateValue) {
value = input;
}
output = value;
}
SC_CTOR(Register) {
130/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
SC_METHOD(run);
sensitive << clock.pos();
sensitive << input;
sensitive << updateValue;
value = 0;
}
};
#endif
lsu_model.hh: The load-store unit model for the TTAs:
#ifndef SC_LSU_MODEL_HH
#define SC_LSU_MODEL_HH
#include <systemc>
#include <tce_systemc.hh>
#include "mem_map.h"
TCE_SC_OPERATION_SIMULATOR(LSUModel) {
/* The same LSU simulation model is used for the sender and
the receiver TTAs. The former writes to the data reg and
reads the busy reg, latter vice-versa. */
sc_in<int> reg_value_in;
sc_out<int> reg_value_out;
sc_out<bool> reg_value_update;
TCE_SC_OPERATION_SIMULATOR_CTOR(LSUModel) {}
TCE_SC_SIMULATE_CYCLE_START {
// initialize the update signal to 0 so we won’t update any
// garbage to the register unless a write operation writes
// to it
reg_value_update = 0;
}
TCE_SC_SIMULATE_STAGE {
unsigned address = TCE_SC_UINT(1);
// overwrite only the stage 0 simulation behavior of loads and
// stores to out of data memory addresses
if (address <= LAST_DMEM_ADDR || TCE_SC_OPSTAGE > 0) {
return false;
}
// do not check for the address, assume all out of data memory
// addresses update the shared register value
if (TCE_SC_OPERATION.writesMemory()) {
int value = TCE_SC_INT(2);
reg_value_out.write(value);
reg_value_update.write(1);
} else { // a load, the operand 2 is the data output
int value = reg_value_in.read();
TCE_SC_OUTPUT(2) = value;
}
return true;
}
};
version 31 2016-02-23 131/145
TTA-based Co-design Environment v1.14 User Manual
#endif
Finally, the actual main SystemC simulation code is defined as follows. As can be noted, both of the TTA
cores use the same architecture loaded from mmio.adf of which contents are not presented here. In order
to make this example work, the TCE-included minimal_with_io.adf architecture can be used instead.
simulator.cc: The main simulation code:
#include <iostream>
#include "systemc.h"
#include "register.hh"
#include "lsu_model.hh"
int sc_main(int argc, char* argv[]) {
// 100MHz clock frequency (1 us clock period)
sc_clock clk("clock", 1, SC_US);
sc_signal<bool> glock;
sc_signal<int> busyRegDataIn;
sc_signal<int> dataRegDataIn;
sc_signal<int> busyRegDataOut;
sc_signal<int> dataRegDataOut;
sc_signal<bool> busyRegUpdate;
sc_signal<bool> dataRegUpdate;
Register dataReg("data_reg");
dataReg.input(dataRegDataIn);
dataReg.output(dataRegDataOut);
dataReg.updateValue(dataRegUpdate);
dataReg.clock(clk.signal());
Register busyReg("busy_reg");
busyReg.input(busyRegDataIn);
busyReg.output(busyRegDataOut);
busyReg.updateValue(busyRegUpdate);
busyReg.clock(clk.signal());
// the sender TTA:
TTACore sender_tta("sender_tta", "mmio.adf", "mmio_send.tpef");
sender_tta.clock(clk.signal());
sender_tta.global_lock(glock);
// the LSU writes to the data register and reads from the
// busy reg to synchronize
LSUModel lsu1("LSU1");
sender_tta.setOperationSimulator("LSU", lsu1);
lsu1.reg_value_in(busyRegDataOut);
lsu1.reg_value_out(dataRegDataIn);
lsu1.reg_value_update(dataRegUpdate);
// the receiver TTA:
132/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
TTACore recv_tta("recv_tta", "mmio.adf", "mmio_recv.tpef");
recv_tta.clock(clk.signal());
recv_tta.global_lock(glock);
// the LSU writes to the busy reg to synchronize the execution
// and reads from the data reg
LSUModel lsu2("LSU2");
recv_tta.setOperationSimulator("LSU", lsu2);
lsu2.reg_value_in(dataRegDataOut);
lsu2.reg_value_out(busyRegDataIn);
lsu2.reg_value_update(busyRegUpdate);
// simulate for 0.2 sec = 200K cycles
sc_time runtime(0.2, SC_SEC);
sc_start(runtime);
return EXIT_SUCCESS;
}
The simulator can be compiled with the following command (assuming gcc used):
g++ ‘tce-config --includes --libs‘ simulator.cc -lsystemc -O3 -o simulator
The simulation should produce output similar to the following:
./simulator
SystemC 2.2.0 --- Aug 30 2010 13:05:02
Copyright (c) 1996-2006 by all Contributors
ALL RIGHTS RESERVED
mmio_recv got 1234
mmio_recv got 3702
mmio_recv got 4936
mmio_recv got 6170
mmio_recv got 7404
mmio_recv got 9872
mmio_recv got 12340
version 31 2016-02-23 133/145
TTA-based Co-design Environment v1.14 User Manual
Appendix C
Copyright notices
Here are the copyright notices of the libraries used in the software.
C.1 Xerces
Xerces-C++ is released under the terms of the Apache License, version 2.0. The complete text is presented
here.
Apache License Version 2.0, January 2004 http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction, and distribution as defined by Sec-
tions 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by the copyright owner that is granting
the License.
"Legal Entity" shall mean the union of the acting entity and all other entities that control, are controlled
by, or are under common control with that entity. For the 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 (50outstanding shares, or (iii) beneficial ownership of such
entity.
"You" (or "Your") shall mean an individual or Legal Entity exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications, including but not limited to
software source code, documentation source, and configuration files.
"Object" form shall mean any form resulting from mechanical transformation or translation of a Source
form, including but not limited to compiled object code, generated documentation, and conversions to
other media types.
"Work" shall mean the work of authorship, whether in Source or Object form, made available under the
License, as indicated by a copyright notice that is included in or attached to the work (an example is
provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object form, that is based on (or derived
from) the Work and for which the editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes of this License, Derivative Works
shall not include works that remain separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including the original version of the Work and any
modifications or additions to that Work or Derivative Works thereof, that is intentionally submitted to
Licensor for inclusion in the Work by the copyright owner or by an individual or Legal Entity authorized
to submit on behalf of the copyright owner. For the purposes of this definition, "submitted" means any
form of electronic, verbal, or written communication sent to the Licensor or its representatives, including
134/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
but not limited to communication on electronic mailing lists, source code control systems, and issue
tracking systems that are managed by, or on behalf of, the Licensor for the purpose of discussing and
improving the Work, but excluding communication that is conspicuously marked or otherwise designated
in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity on behalf of whom a Contribution
has been received by Licensor and subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of this License, each Contributor
hereby grants to You a perpetual, worldwide, non-exclusive, no-charge, royalty-free, irrevocable copy-
right license to reproduce, prepare Derivative Works of, publicly display, publicly perform, sublicense,
and distribute the Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of this License, each Contributor hereby
grants to You a perpetual, worldwide, non-exclusive, no-charge, royalty-free, irrevocable (except as stated
in this section) patent license to make, have made, use, offer to sell, sell, import, and otherwise transfer
the Work, where such license applies only to those patent claims licensable by such Contributor that are
necessarily infringed by their Contribution(s) alone or by combination of their Contribution(s) with the
Work to which such Contribution(s) was submitted. If You institute patent litigation against any entity
(including a cross-claim or counterclaim in a lawsuit) alleging that the Work or a Contribution incor-
porated within the Work constitutes direct or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the Work or Derivative Works thereof in
any medium, with or without modifications, and in Source or Object form, provided that You meet the
following conditions:
(a) You must give any other recipients of the Work or Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works that You distribute, all copyright, patent,
trademark, and attribution notices from the Source form of the Work, excluding those notices that do not
pertain to any part of the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its distribution, then any Derivative Works that
You distribute must include a readable copy of the attribution notices contained within such NOTICE
file, excluding those notices that do not pertain to any part of the Derivative Works, in at least one of the
following places: within a NOTICE text file distributed as part of the Derivative Works; within the Source
form or documentation, if provided along with the Derivative Works; or, within a display generated by the
Derivative Works, if and wherever such third-party notices normally appear. The contents of the NOTICE
file are for informational purposes only and do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside or as an addendum to the NOTICE text
from the Work, provided that such additional attribution notices cannot be construed as modifying the
License.
You may add Your own copyright statement to Your modifications and may provide additional or different
license terms and conditions for use, reproduction, or distribution of Your modifications, or for any such
Derivative Works as a whole, provided Your use, reproduction, and distribution of the Work otherwise
complies with the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise, any Contribution intentionally
submitted for inclusion in the Work by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions. Notwithstanding the above, nothing herein shall
supersede or modify the terms of any separate license agreement you may have executed with Licensor
regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade names, trademarks, service marks,
or product names of the Licensor, except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or agreed to in writing, Licensor provides
the Work (and each Contributor provides its Contributions) on an "AS IS" BASIS, WITHOUT WAR-
RANTIES OR CONDITIONS OF ANY KIND, either express or implied, including, without limitation,
any warranties or conditions of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS
version 31 2016-02-23 135/145
TTA-based Co-design Environment v1.14 User Manual
FOR A PARTICULAR PURPOSE. You are solely responsible for determining the appropriateness of
using or redistributing the Work and assume any risks associated with Your exercise of permissions under
this License.
8. Limitation of Liability. In no event and under no legal theory, whether in tort (including negligence),
contract, or otherwise, unless required by applicable law (such as deliberate and grossly negligent acts)
or agreed to in writing, shall any Contributor be liable to You for damages, including any direct, indirect,
special, incidental, or consequential damages of any character arising as a result of this License or out
of the use or inability to use the Work (including but not limited to damages for loss of goodwill, work
stoppage, computer failure or malfunction, or any and all other commercial damages or losses), even if
such Contributor has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing the Work or Derivative Works
thereof, You may choose to offer, and charge a fee for, acceptance of support, warranty, indemnity, or
other liability obligations and/or rights consistent with this License. However, in accepting such obliga-
tions, You may act only on Your own behalf and on Your sole responsibility, not on behalf of any other
Contributor, and only if You agree to indemnify, defend, and hold each Contributor harmless for any
liability incurred by, or claims asserted against, such Contributor by reason of your accepting any such
warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following boilerplate notice, with the fields enclosed
by brackets "[]" replaced with your own identifying information. (Do Not include the brackets!) The text
should be enclosed in the appropriate comment syntax for the file format. We also recommend that a file
or class name and description of purpose be included on the same "printed page" as the copyright notice
for easier identification within third-party archives.
Copyright [yyyy] [name of copyright owner]
Licensed under the Apache License, Version 2.0 (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.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software distributed under the License is dis-
tributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either
express or implied. See the License for the specific language governing permissions and limitations under
the License.
C.2 wxWidgets
Copyright (c) 1998 Julian Smart, Robert Roebling [, ...]
Everyone is permitted to copy and distribute verbatim copies of this licence document, but changing it is
not allowed.
WXWINDOWS LIBRARY LICENCE
TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
This library is free software; you can redistribute it and/or modify it under the terms of the GNU Library
General Public Licence as published by the Free Software Foundation; either version 2 of the Licence, or
(at your option) any later version.
This library is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without
even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU Library General Public Licence for more details.
You should have received a copy of the GNU Library General Public Licence along with this software,
usually in a file named COPYING.LIB. If not, write to the Free Software Foundation, Inc., 59 Temple
Place, Suite 330, Boston, MA 02111-1307 USA.
EXCEPTION NOTICE
136/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
1. As a special exception, the copyright holders of this library give permission for additional uses of the
text contained in this release of the library as licenced under the wxWindows Library Licence, applying
either version 3 of the Licence, or (at your option) any later version of the Licence as published by the
copyright holders of version 3 of the Licence document.
2. The exception is that you may use, copy, link, modify and distribute under the user’s own terms, binary
object code versions of works based on the Library.
3. If you copy code from files distributed under the terms of the GNU General Public Licence or the GNU
Library General Public Licence into a copy of this library, as this licence permits, the exception does not
apply to the code that you add in this way. To avoid misleading anyone as to the status of such modified
files, you must delete this exception notice from such code and/or adjust the licensing conditions notice
accordingly.
4. If you write modifications of your own for this library, it is your choice whether to permit this exception
to apply to your modifications. If you do not wish that, you must delete the exception notice from such
code and/or adjust the licensing conditions notice accordingly.
C.3 L-GPL
GNU LIBRARY GENERAL PUBLIC LICENSE
==================================
Version 2, June 1991
Copyright (C) 1991 Free Software Foundation, Inc. 675 Mass Ave, Cambridge, MA 02139, USA
Everyone is permitted to copy and distribute verbatim copies of this license document, but changing it is
not allowed.
[This is the first released version of the library GPL. It is numbered 2 because it goes with version 2 of
the ordinary GPL.]
Preamble
The licenses for most software are designed to take away your freedom to share and change it. By
contrast, the GNU General Public Licenses are intended to guarantee your freedom to share and change
free software–to make sure the software is free for all its users.
This license, the Library General Public License, applies to some specially designated Free Software
Foundation software, and to any other libraries whose authors decide to use it. You can use it for your
libraries, too.
When we speak of free software, we are referring to freedom, not price. Our General Public Licenses are
designed to make sure that you have the freedom to distribute copies of free software (and charge for this
service if you wish), that you receive source code or can get it if you want it, that you can change the
software or use pieces of it in new free programs; and that you know you can do these things.
To protect your rights, we need to make restrictions that forbid anyone to deny you these rights or to ask
you to surrender the rights. These restrictions translate to certain responsibilities for you if you distribute
copies of the library, or if you modify it.
For example, if you distribute copies of the library, whether gratis or for a fee, you must give the recipients
all the rights that we gave you. You must make sure that they, too, receive or can get the source code. If
you link a program with the library, you must provide complete object files to the recipients so that they
can relink them with the library, after making changes to the library and recompiling it. And you must
show them these terms so they know their rights.
Our method of protecting your rights has two steps: (1) copyright the library, and (2) offer you this license
which gives you legal permission to copy, distribute and/or modify the library.
Also, for each distributor’s protection, we want to make certain that everyone understands that there is
no warranty for this free library. If the library is modified by someone else and passed on, we want its
recipients to know that what they have is not the original version, so that any problems introduced by
others will not reflect on the original authors’ reputations.
version 31 2016-02-23 137/145
TTA-based Co-design Environment v1.14 User Manual
Finally, any free program is threatened constantly by software patents. We wish to avoid the danger that
companies distributing free software will individually obtain patent licenses, thus in effect transforming
the program into proprietary software. To prevent this, we have made it clear that any patent must be
licensed for everyone’s free use or not licensed at all.
Most GNU software, including some libraries, is covered by the ordinary GNU General Public License,
which was designed for utility programs. This license, the GNU Library General Public License, applies
to certain designated libraries. This license is quite different from the ordinary one; be sure to read it in
full, and do not assume that anything in it is the same as in the ordinary license.
The reason we have a separate public license for some libraries is that they blur the distinction we usually
make between modifying or adding to a program and simply using it. Linking a program with a library,
without changing the library, is in some sense simply using the library, and is analogous to running a
utility program or application program. However, in a textual and legal sense, the linked executable is a
combined work, a derivative of the original library, and the ordinary General Public License treats it as
such.
Because of this blurred distinction, using the ordinary General Public License for libraries did not effec-
tively promote software sharing, because most developers did not use the libraries. We concluded that
weaker conditions might promote sharing better.
However, unrestricted linking of non-free programs would deprive the users of those programs of all
benefit from the free status of the libraries themselves. This Library General Public License is intended
to permit developers of non-free programs to use free libraries, while preserving your freedom as a user
of such programs to change the free libraries that are incorporated in them. (We have not seen how to
achieve this as regards changes in header files, but we have achieved it as regards changes in the actual
functions of the Library.) The hope is that this will lead to faster development of free libraries.
The precise terms and conditions for copying, distribution and modification follow. Pay close attention
to the difference between a "work based on the library" and a "work that uses the library". The former
contains code derived from the library, while the latter only works together with the library.
Note that it is possible for a library to be covered by the ordinary General Public License rather than by
this special one.
GNU LIBRARY GENERAL PUBLIC LICENSE
TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION
0. This License Agreement applies to any software library which contains a notice placed by the copyright
holder or other authorized party saying it may be distributed under the terms of this Library General Public
License (also called "this License"). Each licensee is addressed as "you".
A "library" means a collection of software functions and/or data prepared so as to be conveniently linked
with application programs (which use some of those functions and data) to form executables.
The "Library", below, refers to any such software library or work which has been distributed under these
terms. A "work based on the Library" means either the Library or any derivative work under copyright
law: that is to say, a work containing the Library or a portion of it, either verbatim or with modifications
and/or translated straightforwardly into another language. (Hereinafter, translation is included without
limitation in the term "modification".)
"Source code" for a work means the preferred form of the work for making modifications to it. For a
library, complete source code means all the source code for all modules it contains, plus any associated
interface definition files, plus the scripts used to control compilation and installation of the library.
Activities other than copying, distribution and modification are not covered by this License; they are
outside its scope. The act of running a program using the Library is not restricted, and output from such
a program is covered only if its contents constitute a work based on the Library (independent of the use
of the Library in a tool for writing it). Whether that is true depends on what the Library does and what
the program that uses the Library does.
1. You may copy and distribute verbatim copies of the Library’s complete source code as you receive it,
in any medium, provided that you conspicuously and appropriately publish on each copy an appropriate
copyright notice and disclaimer of warranty; keep intact all the notices that refer to this License and to
the absence of any warranty; and distribute a copy of this License along with the Library.
138/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
You may charge a fee for the physical act of transferring a copy, and you may at your option offer warranty
protection in exchange for a fee.
2. You may modify your copy or copies of the Library or any portion of it, thus forming a work based
on the Library, and copy and distribute such modifications or work under the terms of Section 1 above,
provided that you also meet all of these conditions:
a) The modified work must itself be a software library.
b) You must cause the files modified to carry prominent notices stating that you changed the files and the
date of any change.
c) You must cause the whole of the work to be licensed at no charge to all third parties under the terms of
this License.
d) If a facility in the modified Library refers to a function or a table of data to be supplied by an application
program that uses the facility, other than as an argument passed when the facility is invoked, then you
must make a good faith effort to ensure that, in the event an application does not supply such function or
table, the facility still operates, and performs whatever part of its purpose remains meaningful.
(For example, a function in a library to compute square roots has a purpose that is entirely well-defined
independent of the application. Therefore, Subsection 2d requires that any application-supplied function
or table used by this function must be optional: if the application does not supply it, the square root
function must still compute square roots.)
These requirements apply to the modified work as a whole. If identifiable sections of that work are not de-
rived from the Library, and can be reasonably considered independent and separate works in themselves,
then this License, and its terms, do not apply to those sections when you distribute them as separate works.
But when you distribute the same sections as part of a whole which is a work based on the Library, the
distribution of the whole must be on the terms of this License, whose permissions for other licensees
extend to the entire whole, and thus to each and every part regardless of who wrote it.
Thus, it is not the intent of this section to claim rights or contest your rights to work written entirely by
you; rather, the intent is to exercise the right to control the distribution of derivative or collective works
based on the Library.
In addition, mere aggregation of another work not based on the Library with the Library (or with a work
based on the Library) on a volume of a storage or distribution medium does not bring the other work
under the scope of this License.
3. You may opt to apply the terms of the ordinary GNU General Public License instead of this License
to a given copy of the Library. To do this, you must alter all the notices that refer to this License, so that
they refer to the ordinary GNU General Public License, version 2, instead of to this License. (If a newer
version than version 2 of the ordinary GNU General Public License has appeared, then you can specify
that version instead if you wish.) Do not make any other change in these notices.
Once this change is made in a given copy, it is irreversible for that copy, so the ordinary GNU General
Public License applies to all subsequent copies and derivative works made from that copy.
This option is useful when you wish to copy part of the code of the Library into a program that is not a
library.
4. You may copy and distribute the Library (or a portion or derivative of it, under Section 2) in object
code or executable form under the terms of Sections 1 and 2 above provided that you accompany it with
the complete corresponding machine-readable source code, which must be distributed under the terms of
Sections 1 and 2 above on a medium customarily used for software interchange.
If distribution of object code is made by offering access to copy from a designated place, then offering
equivalent access to copy the source code from the same place satisfies the requirement to distribute the
source code, even though third parties are not compelled to copy the source along with the object code.
5. A program that contains no derivative of any portion of the Library, but is designed to work with the
Library by being compiled or linked with it, is called a "work that uses the Library". Such a work, in
isolation, is not a derivative work of the Library, and therefore falls outside the scope of this License.
However, linking a "work that uses the Library" with the Library creates an executable that is a derivative
of the Library (because it contains portions of the Library), rather than a "work that uses the library". The
executable is therefore covered by this License. Section 6 states terms for distribution of such executables.
version 31 2016-02-23 139/145
TTA-based Co-design Environment v1.14 User Manual
When a "work that uses the Library" uses material from a header file that is part of the Library, the object
code for the work may be a derivative work of the Library even though the source code is not. Whether
this is true is especially significant if the work can be linked without the Library, or if the work is itself a
library. The threshold for this to be true is not precisely defined by law.
If such an object file uses only numerical parameters, data structure layouts and accessors, and small
macros and small inline functions (ten lines or less in length), then the use of the object file is unrestricted,
regardless of whether it is legally a derivative work. (Executables containing this object code plus portions
of the Library will still fall under Section 6.)
Otherwise, if the work is a derivative of the Library, you may distribute the object code for the work under
the terms of Section 6. Any executables containing that work also fall under Section 6, whether or not
they are linked directly with the Library itself.
6. As an exception to the Sections above, you may also compile or link a "work that uses the Library"
with the Library to produce a work containing portions of the Library, and distribute that work under
terms of your choice, provided that the terms permit modification of the work for the customer’s own use
and reverse engineering for debugging such modifications.
You must give prominent notice with each copy of the work that the Library is used in it and that the
Library and its use are covered by this License. You must supply a copy of this License. If the work
during execution displays copyright notices, you must include the copyright notice for the Library among
them, as well as a reference directing the user to the copy of this License. Also, you must do one of these
things:
a) Accompany the work with the complete corresponding machine-readable source code for the Library
including whatever changes were used in the work (which must be distributed under Sections 1 and 2
above); and, if the work is an executable linked with the Library, with the complete machine-readable
"work that uses the Library", as object code and/or source code, so that the user can modify the Library
and then relink to produce a modified executable containing the modified Library. (It is understood
that the user who changes the contents of definitions files in the Library will not necessarily be able to
recompile the application to use the modified definitions.)
b) Accompany the work with a written offer, valid for at least three years, to give the same user the mate-
rials specified in Subsection 6a, above, for a charge no more than the cost of performing this distribution.
c) If distribution of the work is made by offering access to copy from a designated place, offer equivalent
access to copy the above specified materials from the same place.
d) Verify that the user has already received a copy of these materials or that you have already sent this
user a copy.
For an executable, the required form of the "work that uses the Library" must include any data and utility
programs needed for reproducing the executable from it. However, as a special exception, the source
code distributed need not include anything that is normally distributed (in either source or binary form)
with the major components (compiler, kernel, and so on) of the operating system on which the executable
runs, unless that component itself accompanies the executable.
It may happen that this requirement contradicts the license restrictions of other proprietary libraries that
do not normally accompany the operating system. Such a contradiction means you cannot use both them
and the Library together in an executable that you distribute.
7. You may place library facilities that are a work based on the Library side-by-side in a single library
together with other library facilities not covered by this License, and distribute such a combined library,
provided that the separate distribution of the work based on the Library and of the other library facilities
is otherwise permitted, and provided that you do these two things:
a) Accompany the combined library with a copy of the same work based on the Library, uncombined with
any other library facilities. This must be distributed under the terms of the Sections above.
b) Give prominent notice with the combined library of the fact that part of it is a work based on the
Library, and explaining where to find the accompanying uncombined form of the same work.
8. You may not copy, modify, sublicense, link with, or distribute the Library except as expressly provided
under this License. Any attempt otherwise to copy, modify, sublicense, link with, or distribute the Library
is void, and will automatically terminate your rights under this License. However, parties who have
received copies, or rights, from you under this License will not have their licenses terminated so long as
140/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
such parties remain in full compliance.
9. You are not required to accept this License, since you have not signed it. However, nothing else grants
you permission to modify or distribute the Library or its derivative works. These actions are prohibited
by law if you do not accept this License. Therefore, by modifying or distributing the Library (or any
work based on the Library), you indicate your acceptance of this License to do so, and all its terms and
conditions for copying, distributing or modifying the Library or works based on it.
10. Each time you redistribute the Library (or any work based on the Library), the recipient automatically
receives a license from the original licensor to copy, distribute, link with or modify the Library subject to
these terms and conditions. You may not impose any further restrictions on the recipients’ exercise of the
rights granted herein. You are not responsible for enforcing compliance by third parties to this License.
11. If, as a consequence of a court judgment or allegation of patent infringement or for any other reason
(not limited to patent issues), conditions are imposed on you (whether by court order, agreement or
otherwise) that contradict the conditions of this License, they do not excuse you from the conditions of
this License. If you cannot distribute so as to satisfy simultaneously your obligations under this License
and any other pertinent obligations, then as a consequence you may not distribute the Library at all. For
example, if a patent license would not permit royalty-free redistribution of the Library by all those who
receive copies directly or indirectly through you, then the only way you could satisfy both it and this
License would be to refrain entirely from distribution of the Library.
If any portion of this section is held invalid or unenforceable under any particular circumstance, the
balance of the section is intended to apply, and the section as a whole is intended to apply in other
circumstances.
It is not the purpose of this section to induce you to infringe any patents or other property right claims or
to contest validity of any such claims; this section has the sole purpose of protecting the integrity of the
free software distribution system which is implemented by public license practices. Many people have
made generous contributions to the wide range of software distributed through that system in reliance
on consistent application of that system; it is up to the author/donor to decide if he or she is willing to
distribute software through any other system and a licensee cannot impose that choice.
This section is intended to make thoroughly clear what is believed to be a consequence of the rest of this
License.
12. If the distribution and/or use of the Library is restricted in certain countries either by patents or by
copyrighted interfaces, the original copyright holder who places the Library under this License may add
an explicit geographical distribution limitation excluding those countries, so that distribution is permitted
only in or among countries not thus excluded. In such case, this License incorporates the limitation as if
written in the body of this License.
13. The Free Software Foundation may publish revised and/or new versions of the Library General Public
License from time to time. Such new versions will be similar in spirit to the present version, but may
differ in detail to address new problems or concerns.
Each version is given a distinguishing version number. If the Library specifies a version number of
this License which applies to it and "any later version", you have the option of following the terms and
conditions either of that version or of any later version published by the Free Software Foundation. If
the Library does not specify a license version number, you may choose any version ever published by the
Free Software Foundation.
14. If you wish to incorporate parts of the Library into other free programs whose distribution conditions
are incompatible with these, write to the author to ask for permission. For software which is copyrighted
by the Free Software Foundation, write to the Free Software Foundation; we sometimes make exceptions
for this. Our decision will be guided by the two goals of preserving the free status of all derivatives of our
free software and of promoting the sharing and reuse of software generally.
NO WARRANTY
15. BECAUSE THE LIBRARY IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY FOR
THE LIBRARY, TO THE EXTENT PERMITTED BY APPLICABLE LAW. EXCEPT WHEN OTH-
ERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES PRO-
VIDE THE LIBRARY "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED
version 31 2016-02-23 141/145
TTA-based Co-design Environment v1.14 User Manual
OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MER-
CHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE. THE ENTIRE RISK AS TO THE
QUALITY AND PERFORMANCE OF THE LIBRARY IS WITH YOU. SHOULD THE LIBRARY
PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING, REPAIR OR
CORRECTION.
16. IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR RE-
DISTRIBUTE THE LIBRARY AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES,
INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARIS-
ING OUT OF THE USE OR INABILITY TO USE THE LIBRARY (INCLUDING BUT NOT LIMITED
TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY
YOU OR THIRD PARTIES OR A FAILURE OF THE LIBRARY TO OPERATE WITH ANY OTHER
SOFTWARE), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POS-
SIBILITY OF SUCH DAMAGES.
END OF TERMS AND CONDITIONS
Appendix: How to Apply These Terms to Your New Libraries
If you develop a new library, and you want it to be of the greatest possible use to the public, we rec-
ommend making it free software that everyone can redistribute and change. You can do so by permitting
redistribution under these terms (or, alternatively, under the terms of the ordinary General Public License).
To apply these terms, attach the following notices to the library. It is safest to attach them to the start of
each source file to most effectively convey the exclusion of warranty; and each file should have at least
the "copyright" line and a pointer to where the full notice is found.
<one line to give the library’s name and a brief idea of what it does.> Copyright (C) <year> <name of
author>
This library is free software; you can redistribute it and/or modify it under the terms of the GNU Library
General Public License as published by the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
This library is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without
even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
See the GNU Library General Public License for more details.
You should have received a copy of the GNU Library General Public License along with this library; if
not, write to the Free Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
Also add information on how to contact you by electronic and paper mail.
You should also get your employer (if you work as a programmer) or your school, if any, to sign a
"copyright disclaimer" for the library, if necessary. Here is a sample; alter the names:
Yoyodyne, Inc., hereby disclaims all copyright interest in the library ‘Frob’ (a library for tweaking knobs)
written by James Random Hacker.
<signature of Ty Coon>, 1 April 1990 Ty Coon, President of Vice
That’s all there is to it!
C.4 TCL
Tcl/Tk License Terms
This software is copyrighted by the Regents of the University of California, Sun Microsystems, Inc.,
Scriptics Corporation, and other parties. The following terms apply to all files associated with the software
unless explicitly disclaimed in individual files.
The authors hereby grant permission to use, copy, modify, distribute, and license this software and its
documentation for any purpose, provided that existing copyright notices are retained in all copies and
that this notice is included verbatim in any distributions. No written agreement, license, or royalty fee
142/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
is required for any of the authorized uses. Modifications to this software may be copyrighted by their
authors and need not follow the licensing terms described here, provided that the new terms are clearly
indicated on the first page of each file where they apply.
IN NO EVENT SHALL THE AUTHORS OR DISTRIBUTORS BE LIABLE TO ANY PARTY FOR
DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES ARISING OUT
OF THE USE OF THIS SOFTWARE, ITS DOCUMENTATION, OR ANY DERIVATIVES THEREOF,
EVEN IF THE AUTHORS HAVE BEEN ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
THE AUTHORS AND DISTRIBUTORS SPECIFICALLY DISCLAIM ANY WARRANTIES, INCLUD-
ING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY, FITNESS
FOR A PARTICULAR PURPOSE, AND NON-INFRINGEMENT. THIS SOFTWARE IS PROVIDED
ON AN "AS IS" BASIS, AND THE AUTHORS AND DISTRIBUTORS HAVE NO OBLIGATION TO
PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS.
GOVERNMENT USE: If you are acquiring this software on behalf of the U.S. government, the Gov-
ernment shall have only "Restricted Rights" in the software and related documentation as defined in the
Federal Acquisition Regulations (FARs) in Clause 52.227.19 (c) (2). If you are acquiring the software
on behalf of the Department of Defense, the software shall be classified as "Commercial Computer Soft-
ware" and the Government shall have only "Restricted Rights" as defined in Clause 252.227-7013 (c) (1)
of DFARs. Notwithstanding the foregoing, the authors grant the U.S. Government and others acting in its
behalf permission to use and distribute the software in accordance with the terms specified in this license.
C.5 SQLite
SQLite is public domain. For more information, see http://www.sqlite.org/copyright.html
C.6 Editline
-
Copyright (c) 1997 The NetBSD Foundation, Inc.
All rights reserved.
This code is derived from software contributed to The NetBSD Foundation by Jaromir Dolecek.
Redistribution and use in source and binary forms, with or without modification, are permitted provided
that the following conditions are met:
1. Redistributions of source code must retain the above copyright notice, this list of conditions and the
following disclaimer.
2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and
the following disclaimer in the documentation and/or other materials provided with the distribution.
3. All advertising materials mentioning features or use of this software must display the following ac-
knowledgement: This product includes software developed by the NetBSD Foundation, Inc. and its
contributors.
4. Neither the name of The NetBSD Foundation nor the names of its contributors may be used to endorse
or promote products derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
“AS IS” AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICU-
LAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIB-
UTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUB-
STITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUP-
TION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY
version 31 2016-02-23 143/145
TTA-based Co-design Environment v1.14 User Manual
WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
SUCH DAMAGE.
144/145 2016-02-23 version 31
TTA-BASED CO-DESIGN ENVIRONMENT
Bibliography
[Bou01] Jani Boutellier. Transport-triggered processors. Computer Science and Engineering Laboratory,
University of Oulu, [Online]. Available: http://tce.cs.tut.fi/slides/tta_uo.ppt, Jan 2001.
[Cil04] Andrea Cilio. TTA Program Exchange Format. S-004, [Online] Available:
http://tce.cs.tut.fi/specs/TPEF.pdf, 2004.
[Cor97] Henk Corporaal. Microprocessor Architectures; from VLIW to TTA. John Wiley, 1997.
[CSJ04] Andrea Cilio, Henjo Schot, and Johan Janssen. Processor Architecture Defini-
tion File Format for a New TTA Design Framework. S-003, [Online] Available:
http://tce.cs.tut.fi/specs/ADF.pdf, 2004.
[llv08] The LLVM project home page. Website, 2008.
http://www.llvm.org
.
[lpt] TTA-based Co-design Environment (TCE). [Online]. Available: https://launchpad.net/tce.
[TCE] TTA-based Co-design Environment (TCE). Department of Computer Systems, Tampere Uni-
versity of Technology, [Online]. Available: http://tce.cs.tut.fi/index.html.
version 31 2016-02-23 145/145