Cortex M4 Devices Generic User Guide

Cortex-M4%20Devices%20Generic%20User%20Guide

Cortex_m4%20Generic%20User%20Guide

DUI0553A_cortex_m4_Generic_User_Guide

cortex_m4_User_Guide

CortexM4_Generic_Users_Guide_DUI0553A_cortex_m4_dgug

User Manual:

Open the PDF directly: View PDF .
Page Count: 276

Download
Open PDF In Browser	View PDF

Cortex -M4 Devices
™

Generic User Guide

Cortex-M4 Devices
Generic User Guide
Copyright © 2010 ARM. All rights reserved.
Release Information
The following changes have been made to this book.
Change history
Date

Issue

Confidentiality

Change

16 December 2010

Non-Confidential

First release

Proprietary Notice
Words and logos marked with ® or ™ are registered trademarks or trademarks of ARM® in the EU and other countries,
except as otherwise stated below in this proprietary notice. Other brands and names mentioned herein may be the
trademarks of their respective owners.
Neither the whole nor any part of the information contained in, or the product described in, this document may be
adapted or reproduced in any material form except with the prior written permission of the copyright holder.
The product described in this document is subject to continuous developments and improvements. All particulars of the
product and its use contained in this document are given by ARM in good faith. However, all warranties implied or
expressed, including but not limited to implied warranties of merchantability, or fitness for purpose, are excluded.
This document is intended only to assist the reader in the use of the product. ARM shall not be liable for any loss or
damage arising from the use of any information in this document, or any error or omission in such information, or any
incorrect use of the product.
Where the term ARM is used it means “ARM or any of its subsidiaries as appropriate”.
Confidentiality Status
This document is Non-Confidential. The right to use, copy and disclose this document may be subject to license
restrictions in accordance with the terms of the agreement entered into by ARM and the party that ARM delivered this
document to.
Product Status
The information in this document is final, that is for a developed product.
Web Address
http://www.arm.com

ARM DUI 0553A
ID121610

Contents
Cortex-M4 Devices Generic User Guide

Preface
About this book ........................................................................................................... vi
Feedback .................................................................................................................... ix

Chapter 1

Introduction
1.1

Chapter 2

About the Cortex-M4 processor and core peripherals ............................................. 1-2

The Cortex-M4 Processor
2.1
2.2
2.3
2.4
2.5

Chapter 3

Programmers model ................................................................................................ 2-2
Memory model ....................................................................................................... 2-12
Exception model .................................................................................................... 2-21
Fault handling ........................................................................................................ 2-29
Power management ............................................................................................... 2-32

The Cortex-M4 Instruction Set
3.1
3.2
3.3
3.4
3.5
3.6
3.7
3.8
3.9
3.10
3.11
3.12

ARM DUI 0553A
ID121610

Instruction set summary ........................................................................................... 3-2
CMSIS functions ...................................................................................................... 3-9
About the instruction descriptions .......................................................................... 3-11
Memory access instructions .................................................................................. 3-22
General data processing instructions .................................................................... 3-39
Multiply and divide instructions .............................................................................. 3-74
Saturating instructions ........................................................................................... 3-95
Packing and unpacking instructions .................................................................... 3-107
Bitfield instructions ............................................................................................... 3-114
Branch and control instructions ........................................................................... 3-118
Floating-point instructions .................................................................................... 3-126
Miscellaneous instructions ................................................................................... 3-157

iii

Chapter 4

Cortex-M4 Peripherals
4.1
4.2
4.3
4.4
4.5
4.6

Appendix A

About the Cortex-M4 peripherals ............................................................................. 4-2
Nested Vectored Interrupt Controller ....................................................................... 4-3
System control block .............................................................................................. 4-11
System timer, SysTick ........................................................................................... 4-33
Optional Memory Protection Unit ........................................................................... 4-37
Floating Point Unit (FPU) ....................................................................................... 4-48

Cortex-M4 Options
A.1

Cortex-M4 implementation options .......................................................................... A-2

Glossary

ARM DUI 0553A
ID121610

Preface

This preface introduces the Cortex-M4 Devices Generic User Guide. It contains the following
sections:
•
About this book on page vi
•
Feedback on page ix.

ARM DUI 0553A
ID121610

Preface

About this book
This book is a generic user guide for devices that implement the ARM Cortex-M4 processor.
Implementers of Cortex-M4 designs make a number of implementation choices, that can affect
the functionality of the device. This means that, in this book:
•
some information is described as implementation-defined
•
some features are described as optional.
In this book, unless the context indicates otherwise:
Processor

Refers to the Cortex-M4 processor, as supplied by ARM.

Device

Refers to an implemented device, supplied by an ARM partner, that incorporates
a Cortex-M4 processor. In particular, your device refers to the particular
implementation of the Cortex-M4 that you are using. Some features of your
device depend on the implementation choices made by the ARM partner that
made the device.

Product revision status
The rnpn identifier indicates the revision status of the product described in this book, where:
rn

Identifies the major revision of the product.

Identifies the minor revision or modification status of the product.

Intended audience
This book is written for application and system-level software developers, familiar with
programming, who want to program a device that includes the Cortex-M4 processor.
Using this book
This book is organized into the following chapters:
Chapter 1 Introduction
Read this for an introduction to the Cortex-M4 processor and its features.
Chapter 2 The Cortex-M4 Processor
Read this for information about how to program the processor, the processor
memory model, exception and fault handling, and power management.
Chapter 3 The Cortex-M4 Instruction Set
Read this for information about the processor instruction set.
Chapter 4 Cortex-M4 Peripherals
Read this for information about Cortex-M4 peripherals.
Appendix A Cortex-M4 Options
Read this for information about the processor implementation and configuration
options.
Glossary

ARM DUI 0553A
ID121610

Read this for definitions of terms used in this book.

Preface

Typographical conventions
The typographical conventions used in this document are:
italic

Highlights important notes, introduces special terminology, denotes
internal cross-references, and citations.

bold

Used for terms in descriptive lists, where appropriate.

monospace

Denotes text that you can enter at the keyboard, such as commands, file
and program names, and source code.

monospace italic

Denotes arguments to monospace text where the argument is to be
replaced by a specific value.

< and >

Enclose replaceable terms for assembler syntax where they appear in code
or code fragments. For example:
CMP Rn,

ARM DUI 0553A
ID121610

vii

Preface

Additional reading
This section lists publications by ARM and by third parties.
See Infocenter, http://infocenter.arm.com, for access to ARM documentation.
See onARM, http://onarm.com, for embedded software development resources including the
Cortex Microcontroller Software Interface Standard (CMSIS).
ARM publications
This book contains information that is specific to this product. See the following documents for
other relevant information:
•
Cortex-M4 Technical Reference Manual (ARM DDI 0439)
•
ARMv7-M Architecture Reference Manual (ARM DDI 0403).
Other publications
This guide only provides generic information for devices that implement the ARM Cortex-M4
processor. For information about your device see the documentation published by the device
manufacturer.

ARM DUI 0553A
ID121610

viii

Preface

Feedback
ARM welcomes feedback on this product and its documentation.
Feedback on content
If you have comments on content then send an e-mail to errata@arm.com. Give:
•
the title
•
the number, ARM DUI 0553A
•
the page numbers to which your comments apply
•
a concise explanation of your comments.
ARM also welcomes general suggestions for additions and improvements.

ARM DUI 0553A
ID121610

Chapter 1
Introduction

This chapter introduces the Cortex-M4 processor and its features. It contains the following section:
•

ARM DUI 0553A
ID121610

About the Cortex-M4 processor and core peripherals on page 1-2.

1-1

Introduction

1.1

About the Cortex-M4 processor and core peripherals
The Cortex-M4 processor is a high performance 32-bit processor designed for the
microcontroller market. It offers significant benefits to developers, including:
•
outstanding processing performance combined with fast interrupt handling
•
enhanced system debug with extensive breakpoint and trace capabilities
•
efficient processor core, system and memories
•
ultra-low power consumption with integrated sleep mode and an optional deep sleep
mode
•
platform security robustness, with optional integrated Memory Protection Unit (MPU).
Cortex-M4
processor
Optional
WIC

Optional FPU

NVIC

Optional
Debug
Access Port

Optional
Embedded
Trace Macrocell

Processor
core

Optional
Serial Wire
viewer

Optional Memory
protection unit

Optional
Flash
patch

Optional
Data
watchpoints

Bus matrix
Code
interface

SRAM and
peripheral interface

Figure 1-1 Cortex-M4 implementation

The Cortex-M4 processor is built on a high-performance processor core, with a 3-stage pipeline
Harvard architecture, making it ideal for demanding embedded applications. The processor
delivers exceptional power efficiency through an efficient instruction set and extensively
optimized design, providing high-end processing hardware including optional
IEEE754-compliant single-precision floating-point computation, a range of single-cycle and
SIMD multiplication and multiply-with-accumulate capabilities, saturating arithmetic and
dedicated hardware division.
To facilitate the design of cost-sensitive devices, the Cortex-M4 processor implements
tightly-coupled system components that reduce processor area while significantly improving
interrupt handling and system debug capabilities. The Cortex-M4 processor implements a
version of the Thumb® instruction set based on Thumb-2 technology, ensuring high code density
and reduced program memory requirements. The Cortex-M4 instruction set provides the
exceptional performance expected of a modern 32-bit architecture, with the high code density
of 8-bit and 16-bit microcontrollers.
The Cortex-M4 processor closely integrates a configurable Nested Vectored Interrupt
Controller (NVIC), to deliver industry-leading interrupt performance. The NVIC includes a
Non Maskable Interrupt (NMI) that can provide up to 256 interrupt priority levels. The tight
integration of the processor core and NVIC provides fast execution of Interrupt Service

ARM DUI 0553A
ID121610

1-2

Introduction

Routines (ISRs), dramatically reducing the interrupt latency. This is achieved through the
hardware stacking of registers, and the ability to suspend load-multiple and store-multiple
operations. Interrupt handlers do not require wrapping in assembler code, removing any code
overhead from the ISRs. A tail-chain optimization also significantly reduces the overhead when
switching from one ISR to another.
To optimize low-power designs, the NVIC integrates with the sleep modes, that includes an
optional deep sleep function. This enables the entire device to be rapidly powered down while
still retaining program state.
1.1.1

System-level interface
The Cortex-M4 processor provides multiple interfaces using AMBA® technology to provide
high speed, low latency memory accesses. It supports unaligned data accesses and implements
atomic bit manipulation that enables faster peripheral controls, system spinlocks and thread-safe
Boolean data handling.
The Cortex-M4 processor has an optional Memory Protection Unit (MPU) that permits control
of individual regions in memory, enabling applications to utilize multiple privilege levels,
separating and protecting code, data and stack on a task-by-task basis. Such requirements are
becoming critical in many embedded applications such as automotive.

1.1.2

Optional integrated configurable debug
The Cortex-M4 processor can implement a complete hardware debug solution. This provides
high system visibility of the processor and memory through either a traditional JTAG port or a
2-pin Serial Wire Debug (SWD) port that is ideal for microcontrollers and other small package
devices.
For system trace the processor integrates an Instrumentation Trace Macrocell (ITM) alongside
data watchpoints and a profiling unit. To enable simple and cost-effective profiling of the system
events these generate, a Serial Wire Viewer (SWV) can export a stream of software-generated
messages, data trace, and profiling information through a single pin.
The optional Embedded Trace Macrocell™ (ETM) delivers unrivalled instruction trace capture
in an area far smaller than traditional trace units, enabling many low cost MCUs to implement
full instruction trace for the first time.
The optional Flash Patch and Breakpoint Unit (FPB) provides up to eight hardware breakpoint
comparators that debuggers can use. The comparators in the FPB also provide remap functions
of up to eight words in the program code in the CODE memory region. This enables applications
stored on a non-erasable, ROM-based microcontroller to be patched if a small programmable
memory, for example flash, is available in the device. During initialization, the application in
ROM detects, from the programmable memory, whether a patch is required. If a patch is
required, the application programs the FPB to remap a number of addresses. When those
addresses are accessed, the accesses are redirected to a remap table specified in the FPB
configuration, which means the program in the non-modifiable ROM can be patched.

1.1.3

Cortex-M4 processor features and benefits summary
•
tight integration of system peripherals reduces area and development costs
•
Thumb instruction set combines high code density with 32-bit performance
•
optional IEEE754-compliant single-precision FPU
•
code-patch ability for ROM system updates
•
power control optimization of system components
•
integrated sleep modes for low power consumption

ARM DUI 0553A
ID121610

1-3

Introduction

•
•
•
•
•
•

1.1.4

fast code execution permits slower processor clock or increases sleep mode time
hardware division and fast digital-signal-processing orientated multiply accumulate
saturating arithmetic for signal processing
deterministic, high-performance interrupt handling for time-critical applications
optional Memory Protection Unit (MPU) for safety-critical applications
extensive implementation-defined debug and trace capabilities:
— Serial Wire Debug and Serial Wire Trace reduce the number of pins required for
debugging, tracing, and code profiling.

Cortex-M4 core peripherals
These are:
Nested Vectored Interrupt Controller
The NVIC is an embedded interrupt controller that supports low latency interrupt
processing.
System Control Block
The System Control Block (SCB) is the programmers model interface to the
processor. It provides system implementation information and system control,
including configuration, control, and reporting of system exceptions.
System timer
The system timer, SysTick, is a 24-bit count-down timer. Use this as a Real Time
Operating System (RTOS) tick timer or as a simple counter.
Memory Protection Unit
The Memory Protection Unit (MPU) improves system reliability by defining the
memory attributes for different memory regions. It provides up to eight different
regions, and an optional predefined background region.
Floating-point Unit
The Floating-Point Unit (FPU) provides IEEE754-compliant operations on
single-precision, 32-bit, floating-point values.

ARM DUI 0553A
ID121610

1-4

Chapter 2
The Cortex-M4 Processor

This chapter describes the Cortex-M4 processor. It contains the following sections:
•
Programmers model on page 2-2
•
Memory model on page 2-12
•
Exception model on page 2-21
•
Fault handling on page 2-29
•
Power management on page 2-32.

ARM DUI 0553A
ID121610

2-1

The Cortex-M4 Processor

2.1

Programmers model
This section describes the Cortex-M4 programmers model. In addition to the individual core
register descriptions, it contains information about the processor modes and privilege levels for
software execution and stacks.

2.1.1

Processor mode and privilege levels for software execution
The processor modes are:
Thread mode

Used to execute application software. The processor enters Thread mode
when it comes out of reset.

Handler mode

Used to handle exceptions. The processor returns to Thread mode when it
has finished all exception processing.

The privilege levels for software execution are:
Unprivileged

The software:
•

has limited access to the MSR and MRS instructions, and cannot use the
CPS instruction

•

cannot access the system timer, NVIC, or system control block

•

might have restricted access to memory or peripherals.

Unprivileged software executes at the unprivileged level.
Privileged

The software can use all the instructions and has access to all resources.
Privileged software executes at the privileged level.

In Thread mode, the CONTROL register controls whether software execution is privileged or
unprivileged, see CONTROL register on page 2-9. In Handler mode, software execution is
always privileged.
Only privileged software can write to the CONTROL register to change the privilege level for
software execution in Thread mode. Unprivileged software can use the SVC instruction to make
a supervisor call to transfer control to privileged software.
2.1.2

Stacks
The processor uses a full descending stack. This means the stack pointer holds the address of
the last stacked item in memory. When the processor pushes a new item onto the stack, it
decrements the stack pointer and then writes the item to the new memory location. The
processor implements two stacks, the main stack and the process stack, with a pointer for each
held in independent registers, see Stack Pointer on page 2-4.
In Thread mode, the CONTROL register controls whether the processor uses the main stack or
the process stack, see CONTROL register on page 2-9. In Handler mode, the processor always
uses the main stack. The options for processor operations are:
Table 2-1 Summary of processor mode, execution privilege level, and stack use options
Processor
mode

Used to execute

Privilege level for
software execution

Stack used

Thread

Applications

Privileged or unprivileged a

Main stack or process stack a

Handler

Exception handlers

Always privileged

Main stack

a. See CONTROL register on page 2-9.

ARM DUI 0553A
ID121610

2-2

The Cortex-M4 Processor

2.1.3

Core registers
The processor core registers are:
R0
R1
R2
Low registers

R3
R4
R5
General-purpose registers

R6
R7
R8
R9
High registers

R10
R11
R12

Stack Pointer

SP (R13)

Link Register

LR (R14)

Program Counter

PC (R15)
PSR

PSP‡

MSP‡

‡

Banked version of SP

Program status register

PRIMASK
FAULTMASK

Exception mask registers

Special registers

BASEPRI
CONTROL

CONTROL register

Table 2-2 Core register set summary
Name

Type a

Required privilege b

Reset value

Description

R0-R12

Either

Unknown

General-purpose registers on page 2-4

MSP

Privileged

See description

Stack Pointer on page 2-4

PSP

Either

Unknown

Stack Pointer on page 2-4

Either

0xFFFFFFFF

Link Register on page 2-4

Either

See description

Program Counter on page 2-4

PSR

Privileged

0x01000000

Program Status Register on page 2-4

ASPR

Either

Unknown

Application Program Status Register on page 2-5

IPSR

Privileged

0x00000000

Interrupt Program Status Register on page 2-6

EPSR

Privileged

0x01000000

Execution Program Status Register on page 2-6

PRIMASK

Privileged

0x00000000

Priority Mask Register on page 2-8

FAULTMASK

Privileged

0x00000000

Fault Mask Register on page 2-8

BASEPRI

Privileged

0x00000000

Base Priority Mask Register on page 2-9

CONTROL

Privileged

0x00000000

CONTROL register on page 2-9

a. Describes access type during program execution in thread mode and Handler mode. Debug access can differ.
b. An entry of Either means privileged and unprivileged software can access the register.

ARM DUI 0553A
ID121610

2-3

The Cortex-M4 Processor

General-purpose registers
R0-R12 are 32-bit general-purpose registers for data operations.
Stack Pointer
The Stack Pointer (SP) is register R13. In Thread mode, bit[1] of the CONTROL register
indicates the stack pointer to use:
•
0 = Main Stack Pointer (MSP). This is the reset value.
•
1 = Process Stack Pointer (PSP).
On reset, the processor loads the MSP with the value from address 0x00000000.
Link Register
The Link Register (LR) is register R14. It stores the return information for subroutines, function
calls, and exceptions. On reset, the processor sets the LR value to 0xFFFFFFFF.
Program Counter
The Program Counter (PC) is register R15. It contains the current program address. On reset,
the processor loads the PC with the value of the reset vector, which is at address 0x00000004.
Bit[0] of the value is loaded into the EPSR T-bit at reset and must be 1.
Program Status Register
The Program Status Register (PSR) combines:
•
Application Program Status Register (APSR)
•
Interrupt Program Status Register (IPSR)
•
Execution Program Status Register (EPSR).
These registers are mutually exclusive bitfields in the 32-bit PSR. The bit assignments are:
31 30 29 28 27 26 25 24 23

16 15

APSR N Z C V Q

Reserved

IPSR

EPSR

10 9 8

ICI/IT T

Reserved

ISR_NUMBER

ICI/IT

Reserved

Access these registers individually or as a combination of any two or all three registers, using
the register name as an argument to the MSR or MRS instructions. For example:
•
read all of the registers using PSR with the MRS instruction
•
write to the APSR N, Z, C, V, and Q bits using APSR_nzcvq with the MSR instruction.

ARM DUI 0553A
ID121610

2-4

The Cortex-M4 Processor

The PSR combinations and attributes are:
Table 2-3 PSR register combinations
Register

Type

Combination

PSR

RWa, b

APSR, EPSR, and IPSR

IEPSR

EPSR and IPSR

IAPSR

RWa

APSR and IPSR

EAPSR

RWb

APSR and EPSR

a. The processor ignores writes to the IPSR
bits.
b. Reads of the EPSR bits return zero, and the
processor ignores writes to the these bits

See the instruction descriptions MRS on page 3-163 and MSR on page 3-164 for more
information about how to access the program status registers.
Application Program Status Register

The APSR contains the current state of the condition flags from previous instruction executions.
See the register summary in Table 2-2 on page 2-3 for its attributes. The bit assignments are:
Table 2-4 APSR bit assignments

ARM DUI 0553A
ID121610

Bits

Name

Function

[31]

Negative flag

[30]

Zero flag

[29]

Carry or borrow flag

[28]

Overflow flag

[27]

DSP overflow and saturation flag

[26:20]

Reserved

[19:16]

GE[3:0]

Greater than or Equal flags. See SEL on
page 3-70 for more information.

[15:0]

Reserved

2-5

The Cortex-M4 Processor

Interrupt Program Status Register

The IPSR contains the exception type number of the current Interrupt Service Routine (ISR).
See the register summary in Table 2-2 on page 2-3 for its attributes. The bit assignments are:
Table 2-5 IPSR bit assignments
Bits

Name

Function

[31:9]

Reserved

[8:0]

ISR_NUMBER

This is the number of the current exception:
0 = Thread mode
1 = Reserved
2 = NMI
3 = HardFault
4 = MemManage
5 = BusFault
6 = UsageFault
7-10 = Reserved
11 = SVCall
12 = Reserved for Debug
13 = Reserved
14 = PendSV
15 = SysTick
16 = IRQ0.
.
.
.
n+15 = IRQ(n-1)a
see Exception types on page 2-21 for more information.

a. The number of interrupts, n, is implementation-defined, in the range 1-240.

Execution Program Status Register

The EPSR contains the Thumb state bit, and the execution state bits for either the:
•

If-Then (IT) instruction

•

Interruptible-Continuable Instruction (ICI) field for an interrupted load multiple or store
multiple instruction.

See the register summary in Table 2-2 on page 2-3 for the EPSR attributes. The bit assignments
are:
Table 2-6 EPSR bit assignments

ARM DUI 0553A
ID121610

Bits

Name

Function

[31:27]

Reserved.

[26:25], [15:10]

ICI

Interruptible-continuable instruction bits, see Interruptible-continuable instructions
on page 2-7.

[26:25], [15:10]

Indicates the execution state bits of the IT instruction, see IT on page 3-122.

2-6

The Cortex-M4 Processor

Table 2-6 EPSR bit assignments (continued)
Bits

Name

Function

[24]

Thumb state bit, see Thumb state.

[23:16]

Reserved.

[9:0]

Reserved.

Attempts to read the EPSR directly through application software using the MSR instruction
always return zero. Attempts to write the EPSR using the MSR instruction in application software
are ignored.
Interruptible-continuable instructions

When an interrupt occurs during the execution of an LDM, STM, PUSH, or POP instruction, and when
an FPU is implemented an VLDM, VSTM, VPUSH, or VPOP instruction, the processor:
•
stops the load multiple or store multiple instruction operation temporarily
•
stores the next register operand in the multiple operation to EPSR bits[15:12].
After servicing the interrupt, the processor:
•
returns to the register pointed to by bits[15:12]
•
resumes execution of the multiple load or store instruction.
When the EPSR holds ICI execution state, bits[26:25,11:10] are zero.
If-Then block

The If-Then block contains up to four instructions following an IT instruction. Each instruction
in the block is conditional. The conditions for the instructions are either all the same, or some
can be the inverse of others. See IT on page 3-122 for more information.
Thumb state

The Cortex-M4 processor only supports execution of instructions in Thumb state. The following
can clear the T bit to 0:
•
instructions BLX, BX and POP{PC}
•
restoration from the stacked xPSR value on an exception return
•
bit[0] of the vector value on an exception entry or reset.
Attempting to execute instructions when the T bit is 0 results in a fault or lockup. See Lockup
on page 2-31 for more information.
Exception mask registers
The exception mask registers disable the handling of exceptions by the processor. Disable
exceptions where they might impact on timing critical tasks.
To access the exception mask registers use the MSR and MRS instructions, or the CPS instruction to
change the value of PRIMASK or FAULTMASK. See MRS on page 3-163, MSR on page 3-164,
and CPS on page 3-159 for more information.

ARM DUI 0553A
ID121610

2-7

The Cortex-M4 Processor

Priority Mask Register

The PRIMASK register prevents activation of all exceptions with configurable priority. See the
register summary in Table 2-2 on page 2-3 for its attributes. The bit assignments are:
1 0

31
Reserved

PRIMASK

Table 2-7 PRIMASK register bit assignments
Bits

Name

Function

[31:1]

Reserved

[0]

PRIMASK

0 = no effect
1 = prevents the activation of all exceptions with configurable priority.

Fault Mask Register

The FAULTMASK register prevents activation of all exceptions except for Non-Maskable
Interrupt (NMI). See the register summary in Table 2-2 on page 2-3 for its attributes. The bit
assignments are:
31

1 0
Reserved
FAULTMASK

Table 2-8 FAULTMASK register bit assignments
Bits

Name

Function

[31:1]

Reserved

[0]

FAULTMASK

0 = no effect
1 = prevents the activation of all exceptions except for NMI.

The processor clears the FAULTMASK bit to 0 on exit from any exception handler except the
NMI handler.

ARM DUI 0553A
ID121610

2-8

The Cortex-M4 Processor

Base Priority Mask Register

The BASEPRI register defines the minimum priority for exception processing. When BASEPRI
is set to a nonzero value, it prevents the activation of all exceptions with the same or lower
priority level as the BASEPRI value. See the register summary in Table 2-2 on page 2-3 for its
attributes. The bit assignments are:
31

8 7
Reserved

0
BASEPRI

Table 2-9 BASEPRI register bit assignments
Bits

Name

Function

[31:8]

Reserved

[7:0]

BASEPRI a

Priority mask bits:
0x00 = no effect
Nonzero = defines the base priority for exception processing.
The processor does not process any exception with a priority value greater than or equal to BASEPRI.

a. This field is similar to the priority fields in the interrupt priority registers. Register priority value fields are eight bits wide, and
non-implemented low-order bits read as zero and ignore writes. See Interrupt Priority Registers on page 4-7 for more information.
Remember that higher priority field values correspond to lower exception priorities.

CONTROL register
The CONTROL register controls the stack used and the privilege level for software execution
when the processor is in Thread mode and, if implemented, indicates whether the FPU state is
active. See the register summary in Table 2-2 on page 2-3 for its attributes. The bit assignments
are:
Table 2-10 CONTROL register bit assignments
Bits

Name

Function

[31:3]

Reserved.

[2]

FPCA

When floating-point is implemented this bit indicates whether context floating-point is currently active:
0 = no floating-point context active
1 = floating-point context active.
The Cortex-M4 uses this bit to determine whether to preserve floating-point state when processing an exception.

[1]

SPSEL

Defines the currently active stack pointer: In Handler mode this bit reads as zero and ignores writes. The
Cortex-M4 updates this bit automatically on exception return:
0 = MSP is the current stack pointer
1 = PSP is the current stack pointer.

[0]

nPRIV

Defines the Thread mode privilege level:
0 = privileged
1 = unprivileged.

Handler mode always uses the MSP, so the processor ignores explicit writes to the active stack
pointer bit of the CONTROL register when in Handler mode. The exception entry and return
mechanisms automatically update the CONTROL register based on the EXC_RETURN value,
see Table 2-17 on page 2-28.

ARM DUI 0553A
ID121610

2-9

The Cortex-M4 Processor

In an OS environment, ARM recommends that threads running in Thread mode use the process
stack and the kernel and exception handlers use the main stack.
By default, Thread mode uses the MSP. To switch the stack pointer used in Thread mode to the
PSP, either:
•

use the MSR instruction to set the Active stack pointer bit to 1, see MSR on page 3-164.

•

perform an exception return to Thread mode with the appropriate EXC_RETURN value,
see Table 2-17 on page 2-28.

Note
When changing the stack pointer, software must use an ISB instruction immediately after the MSR
instruction. This ensures that instructions after the ISB instruction execute using the new stack
pointer. See ISB on page 3-162

2.1.4

Exceptions and interrupts
The Cortex-M4 processor supports interrupts and system exceptions. The processor and the
NVIC prioritize and handle all exceptions. An exception changes the normal flow of software
control. The processor uses Handler mode to handle all exceptions except for reset. See
Exception entry on page 2-26 and Exception return on page 2-28 for more information.
The NVIC registers control interrupt handling. See Nested Vectored Interrupt Controller on
page 4-3 for more information.

2.1.5

Data types
The processor:

2.1.6

•

supports the following data types:
— 32-bit words
— 16-bit halfwords
— 8-bit bytes

•

manages all data memory accesses as little-endian or big-endian. Instruction memory and
Private Peripheral Bus (PPB) accesses are always performed as little-endian. See
Memory regions, types and attributes on page 2-12 for more information.

The Cortex Microcontroller Software Interface Standard
For a Cortex-M4 microcontroller system, the Cortex Microcontroller Software Interface
Standard (CMSIS) defines:
•
a common way to:
— access peripheral registers
— define exception vectors.
•
the names of:
— the registers of the core peripherals
— the core exception vectors.
•
a device-independent interface for RTOS kernels, including a debug channel.
The CMSIS includes address definitions and data structures for the core peripherals in the
Cortex-M4 processor.

ARM DUI 0553A
ID121610

2-10

The Cortex-M4 Processor

CMSIS simplifies software development by enabling the reuse of template code and the
combination of CMSIS-compliant software components from various middleware vendors.
Software vendors can expand the CMSIS to include their peripheral definitions and access
functions for those peripherals.
This document includes the register names defined by the CMSIS, and gives short descriptions
of the CMSIS functions that address the processor core and the core peripherals.
Note
This document uses the register short names defined by the CMSIS. In a few cases these differ
from the architectural short names that might be used in other documents.
The following sections give more information about the CMSIS:
•
Power management programming hints on page 2-34
•
CMSIS functions on page 3-9
•
Accessing the Cortex-M4 NVIC registers using CMSIS on page 4-4
•
NVIC programming hints on page 4-9.

ARM DUI 0553A
ID121610

2-11

The Cortex-M4 Processor

2.2

Memory model
This section describes the processor memory map, the behavior of memory accesses, and the
optional bit-banding features. The processor has a fixed default memory map that provides up
to 4GB of addressable memory. The memory map is:
0xFFFFFFFF
Vendor-specific
memory

511MB

0xE0100000
0xE00FFFFF
Private peripheral
1.0MB
bus
0xE0000000
0xDFFFFFFF

External device

1.0GB

0xA0000000
0x9FFFFFFF

External RAM

0x43FFFFFF

1.0GB

32MB Bit band alias
0x60000000
0x5FFFFFFF

0x42000000
0x400FFFFF
1MB Bit band region
0x40000000

Peripheral

0.5GB
0x40000000
0x3FFFFFFF

0x23FFFFFF
32MB Bit band alias

SRAM

0.5GB
0x20000000
0x1FFFFFFF

0x22000000
0x200FFFFF
1MB Bit band region
0x20000000

Code

0.5GB
0x00000000

The regions for SRAM and peripherals include optional bit-band regions. Bit-banding provides
atomic operations to bit data, see Optional bit-banding on page 2-16.
The processor reserves regions of the Private Peripheral Bus (PPB) address range for core
peripheral registers, see About the Cortex-M4 peripherals on page 4-2.
2.2.1

Memory regions, types and attributes
The memory map and programming the optional MPU splits the memory map into regions.
Each region has a defined memory type, and some regions have additional memory attributes.
The memory type and attributes determine the behavior of accesses to the region.
The memory types are:
Normal

ARM DUI 0553A
ID121610

The processor can re-order transactions for efficiency, or perform
speculative reads.

2-12

The Cortex-M4 Processor

Device

The processor preserves transaction order relative to other transactions to
Device or Strongly-ordered memory.

Strongly-ordered

The processor preserves transaction order relative to all other transactions.

The different ordering requirements for Device and Strongly-ordered memory mean that the
memory system can buffer a write to Device memory, but must not buffer a write to
Strongly-ordered memory.
The additional memory attributes include:
Shareable

For a shareable memory region that is implemented, the memory system
provides data synchronization between bus masters in a system with
multiple bus masters, for example, a processor with a DMA controller.
Strongly-ordered memory is always shareable.
If multiple bus masters can access a non-shareable memory region,
software must ensure data coherency between the bus masters.
Note
This attribute is relevant only if the device is likely to be used in systems
where memory is shared between multiple processors.

Execute Never (XN) Means the processor prevents instruction accesses. A fault exception is
generated only on execution of an instruction executed from an XN
region.
2.2.2

Memory system ordering of memory accesses
For most memory accesses caused by explicit memory access instructions, the memory system
does not guarantee that the order in which the accesses complete matches the program order of
the instructions, providing this does not affect the behavior of the instruction sequence.
Normally, if correct program execution depends on two memory accesses completing in
program order, software must insert a memory barrier instruction between the memory access
instructions, see Software ordering of memory accesses on page 2-15.
However, the memory system does guarantee some ordering of accesses to Device and
Strongly-ordered memory. For two memory access instructions A1 and A2, if A1 occurs before
A2 in program order, the ordering of the memory accesses caused by two instructions is:
Normal
access

Non-shareable

Shareable

Stronglyordered
access

Normal access

Device access, non-shareable

Device access, shareable

Strongly-ordered access

A2
A1

Device access

Where:

ARM DUI 0553A
ID121610

Means that the memory system does not guarantee the ordering of the accesses.

Means that accesses are observed in program order, that is, A1 is always observed
before A2.

2-13

The Cortex-M4 Processor

2.2.3

Behavior of memory accesses
The behavior of accesses to each region in the memory map is:
Table 2-11 Memory access behavior
Memory region

Memory
typea

XNa

Description

0x000000000x1FFFFFFF

Code

Normal

Executable region for program code. You can also put data here.

0x200000000x3FFFFFFF

SRAM

Normal

Executable region for data. You can also put code here. This region
includes bit band and bit band alias areas, see Table 2-13 on
page 2-16.

0x400000000x5FFFFFFF

Peripheral

Device

This region includes bit band and bit band alias areas, see
Table 2-14 on page 2-16.

0x600000000x9FFFFFFF

External RAM

Normal

Executable region for data.

0xA00000000xDFFFFFFF

External device

Device

External Device memory.

0xE00000000xE00FFFFF

Private Peripheral Bus

Stronglyordered

This region includes the NVIC, System timer, and system control
block.

0xE01000000xFFFFFFFF

Device

Implementation-specific.

Address
range

a. See Memory regions, types and attributes on page 2-12 for more information.

The Code, SRAM, and external RAM regions can hold programs. However, ARM recommends
that programs always use the Code region. This is because the processor has separate buses that
enable instruction fetches and data accesses to occur simultaneously.
The optional MPU can override the default memory access behavior described in this section.
For more information, see Optional Memory Protection Unit on page 4-37.
Additional memory access constraints for caches and shared memory
When a system includes caches or shared memory, some memory regions might have additional
access constraints, and some regions are subdivided, as Table 2-12 shows:
Table 2-12 Memory region shareability and cache policies
Address range

Memory region

Memory type

Shareability

Cache policy

0x00000000- 0x1FFFFFFF

Code

Normal a

WT b

0x20000000- 0x3FFFFFFF

SRAM

Normal a

WBWA b

0x40000000- 0x5FFFFFFF

Peripheral

Device a

0x60000000- 0x7FFFFFFF

External RAM

Normal a

WBWA b

0x80000000- 0x9FFFFFFF

ARM DUI 0553A
ID121610

WT b

2-14

The Cortex-M4 Processor

Table 2-12 Memory region shareability and cache policies (continued)
Address range

Memory region

Memory type

Shareability

Cache policy

0xA0000000- 0xBFFFFFFF

External device

Device a

Shareable a

0xC0000000- 0xDFFFFFFF

Non-shareable a

0xE0000000- 0xE00FFFFF

Private Peripheral Bus

Strongly- ordered a

Shareable a

0xE0100000- 0xFFFFFFFF

Device

a. See Memory regions, types and attributes on page 2-12 for more information.
b. WT = Write through, no write allocate. WBWA = Write back, write allocate. See the Glossary for more information.

Instruction prefetch and branch prediction
The Cortex-M4 processor:
•
prefetches instructions ahead of execution
•
speculatively prefetches from branch target addresses.
2.2.4

Software ordering of memory accesses
The order of instructions in the program flow does not always guarantee the order of the
corresponding memory transactions. This is because:
•

the processor can reorder some memory accesses to improve efficiency, providing this
does not affect the behavior of the instruction sequence.

•

the processor has multiple bus interfaces

•

memory or devices in the memory map have different wait states

•

some memory accesses are buffered or speculative.

Memory system ordering of memory accesses on page 2-13 describes the cases where the
memory system guarantees the order of memory accesses. Otherwise, if the order of memory
accesses is critical, software must include memory barrier instructions to force that ordering.
The processor provides the following memory barrier instructions:
DMB

The Data Memory Barrier (DMB) instruction ensures that outstanding
memory transactions complete before subsequent memory transactions.
See DMB on page 3-160.

DSB

The Data Synchronization Barrier (DSB) instruction ensures that
outstanding memory transactions complete before subsequent instructions
execute. See DSB on page 3-161.

ISB

The Instruction Synchronization Barrier (ISB) ensures that the effect of all
completed memory transactions is recognizable by subsequent
instructions. See ISB on page 3-162.

MPU programming
Use a DSB followed by an ISB instruction or exception return to ensure that the new MPU
configuration is used by subsequent instructions.

ARM DUI 0553A
ID121610

2-15

The Cortex-M4 Processor

2.2.5

Optional bit-banding
A bit-band region maps each word in a bit-band alias region to a single bit in the bit-band
region. The bit-band regions occupy the lowest 1MB of the SRAM and peripheral memory
regions.
The memory map has two 32MB alias regions that map to two 1MB bit-band regions:
•

accesses to the 32MB SRAM alias region map to the 1MB SRAM bit-band region, as
shown in Table 2-13

•

accesses to the 32MB peripheral alias region map to the 1MB peripheral bit-band region,
as shown in Table 2-14.
Table 2-13 SRAM memory bit-banding regions

Address
range

Memory region

Instruction and data accesses

0x20000000-

SRAM bit-band region

Direct accesses to this memory range behave as SRAM memory accesses, but
this region is also bit addressable through bit-band alias.

SRAM bit-band alias

Data accesses to this region are remapped to bit band region. A write operation
is performed as read-modify-write. Instruction accesses are not remapped.

0x200FFFFF
0x220000000x23FFFFFF

Table 2-14 Peripheral memory bit-banding regions
Address
range

Memory region

Instruction and data accesses

0x40000000-

Peripheral bit-band alias

Direct accesses to this memory range behave as peripheral memory
accesses, but this region is also bit addressable through bit-band alias.

Peripheral bit-band region

Data accesses to this region are remapped to bit band region. A write
operation is performed as read-modify-write. Instruction accesses are not
permitted.

0x400FFFFF
0x420000000x43FFFFFF

•
•

Note
A word access to the SRAM or peripheral bit-band alias regions maps to a single bit in the
SRAM or peripheral bit-band region
Bit band accesses can use byte, halfword, or word transfers. The bit band transfer size
matches the transfer size of the instruction making the bit band access.

The following formula shows how the alias region maps onto the bit-band region:
bit_word_offset = (byte_offset x 32) + (bit_number x 4)
bit_word_addr = bit_band_base + bit_word_offset

where:
•

Bit_word_offset is the position of the target bit in the bit-band memory region

•

Bit_word_addr is the address of the word in the alias memory region that maps to the

targeted bit.

ARM DUI 0553A
ID121610

•

Bit_band_base is the starting address of the alias region

•

Byte_offset is the number of the byte in the bit-band region that contains the targeted bit

2-16

The Cortex-M4 Processor

•

Bit_number is the bit position, 0-7, of the targeted bit.

Figure 2-1 shows examples of bit-band mapping between the SRAM bit-band alias region and
the SRAM bit-band region:
•

the alias word at 0x23FFFFE0 maps to bit[0] of the bit-band byte at
0x200FFFFF: 0x23FFFFE0 = 0x22000000 + (0xFFFFF*32) + (0*4)

•

the alias word at 0x23FFFFFC maps to bit[7] of the bit-band byte at 0x200FFFFF:
0x23FFFFFC = 0x22000000 + (0xFFFFF*32) + (7*4)

•

the alias word at 0x22000000 maps to bit[0] of the bit-band byte at 0x20000000:
0x22000000 = 0x22000000 + (0*32) + (0 *4)

•

the alias word at 0x2200001C maps to bit[7] of the bit-band byte at 0x20000000:
0x2200001C = 0x22000000+ (0*32) + (7*4).
32MB alias region

0x23FFFFFC

0x23FFFFF8

0x23FFFFF4

0x23FFFFF0

0x23FFFFEC

0x23FFFFE8

0x23FFFFE4

0x23FFFFE0

0x2200001C

0x22000018

0x22000014

0x22000010

0x2200000C

0x22000008

0x22000004

0x22000000

1MB SRAM bit-band region
7

0x200FFFFF

0x20000003

0x200FFFFE

0x200FFFFD

0x20000002

0x20000001

0x200FFFFC

0x20000000

Figure 2-1 Bit-band mapping

Directly accessing an alias region
Writing to a word in the alias region updates a single bit in the bit-band region.
Bit[0] of the value written to a word in the alias region determines the value written to the
targeted bit in the bit-band region. Writing a value with bit[0] set to 1 writes a 1 to the bit-band
bit, and writing a value with bit[0] set to 0 writes a 0 to the bit-band bit.
Bits[31:1] of the alias word have no effect on the bit-band bit. Writing 0x01 has the same effect
as writing 0xFF. Writing 0x00 has the same effect as writing 0x0E.
Reading a word in the alias region:
•
0x00000000 indicates that the targeted bit in the bit-band region is set to zero
•
0x00000001 indicates that the targeted bit in the bit-band region is set to 1
Directly accessing a bit-band region
Behavior of memory accesses on page 2-14 describes the behavior of direct byte, halfword, or
word accesses to the bit-band regions.

ARM DUI 0553A
ID121610

2-17

The Cortex-M4 Processor

2.2.6

Memory endianness
The processor views memory as a linear collection of bytes numbered in ascending order from
zero. For example, bytes 0-3 hold the first stored word, and bytes 4-7 hold the second stored
word. The memory endianness used is implementation-defined, and the following subsections
describe the possible implementations:
•
Byte-invariant big-endian format
•
Little-endian format.
Read the AIRCR.ENDIANNESS field to find the implemented endianness, see Application
Interrupt and Reset Control Register on page 4-16.
Byte-invariant big-endian format
In byte-invariant big-endian format, the processor stores the most significant byte of a word at
the lowest-numbered byte, and the least significant byte at the highest-numbered byte. For
example:
Memory
7

0
31

Address A

A+1

A+2

A+3

msbyte

24 23
B0

16 15
B1

8 7
B2

0
B3

lsbyte

Little-endian format
In little-endian format, the processor stores the least significant byte of a word at the
lowest-numbered byte, and the most significant byte at the highest-numbered byte. For
example:
Memory
7

0
31

2.2.7

Address A

A+1

A+2

A+3

lsbyte

24 23
B3

16 15
B2

8 7
B1

0
B0

msbyte

Synchronization primitives
The Cortex-M4 instruction set includes pairs of synchronization primitives. These provide a
non-blocking mechanism that a thread or process can use to obtain exclusive access to a memory
location. Software can use them to perform a guaranteed read-modify-write memory update
sequence, or for a semaphore mechanism.

ARM DUI 0553A
ID121610

2-18

The Cortex-M4 Processor

A pair of synchronization primitives comprises:
A Load-Exclusive instruction
Used to read the value of a memory location, requesting exclusive access to that
location.
A Store-Exclusive instruction
Used to attempt to write to the same memory location, returning a status bit to a
register. If this bit is:
0

It indicates that the thread or process gained exclusive access to the
memory, and the write succeeds.

It indicates that the thread or process did not gain exclusive access to
the memory, and no write was performed.

The pairs of Load-Exclusive and Store-Exclusive instructions are:
•
the word instructions LDREX and STREX
•
the halfword instructions LDREXH and STREXH
•
the byte instructions LDREXB and STREXB.
Software must use a Load-Exclusive instruction with the corresponding Store-Exclusive
instruction.
To perform an exclusive read-modify-write of a memory location, software must:
1.

Use a Load-Exclusive instruction to read the value of the location.

Modify the value, as required.

Use a Store-Exclusive instruction to attempt to write the new value back to the memory
location.

Test the returned status bit. If this bit is:
0

The read-modify-write completed successfully.

No write was performed. This indicates that the value returned at step 1 might
be out of date. The software must retry the entire read-modify-write sequence.

Software can use the synchronization primitives to implement a semaphores as follows:
1.

Use a Load-Exclusive instruction to read from the semaphore address to check whether
the semaphore is free.

If the semaphore is free, use a Store-Exclusive to write the claim value to the semaphore
address.

If the returned status bit from step 2 indicates that the Store-Exclusive succeeded then the
software has claimed the semaphore. However, if the Store-Exclusive failed, another
process might have claimed the semaphore after the software performed step 1.

The Cortex-M4 includes an exclusive access monitor, that tags the fact that the processor has
executed a Load-Exclusive instruction. If the processor is part of a multiprocessor system, the
system also globally tags the memory locations addressed by exclusive accesses by each
processor.
The processor removes its exclusive access tag if:

ARM DUI 0553A
ID121610

•

It executes a CLREX instruction.

•

It executes a Store-Exclusive instruction, regardless of whether the write succeeds.

2-19

The Cortex-M4 Processor

•

An exception occurs. This means the processor can resolve semaphore conflicts between
different threads.

In a multiprocessor implementation:
•

executing a CLREX instruction removes only the local exclusive access tag for the processor

•

executing a Store-Exclusive instruction, or an exception. removes the local exclusive
access tags, and global exclusive access tags for the processor.

For more information about the synchronization primitive instructions, see LDREX and STREX
on page 3-36 and CLREX on page 3-38.
2.2.8

Programming hints for the synchronization primitives
ISO/IEC C cannot directly generate the exclusive access instructions. CMSIS provides
functions for generation of these instructions:
Table 2-15 CMSIS functions for exclusive access instructions

ARM DUI 0553A
ID121610

Instruction

CMSIS function

LDREX

uint32_t __LDREXW (uint32_t *addr)

LDREXH

uint16_t __LDREXH (uint16_t *addr)

LDREXB

uint8_t __LDREXB (uint8_t *addr)

STREX

uint32_t __STREXW (uint32_t value, uint32_t *addr)

STREXH

uint32_t __STREXH (uint16_t value, uint16_t *addr)

STREXB

uint32_t __STREXB (uint8_t value, uint8_t *addr)

CLREX

void __CLREX (void)

2-20

The Cortex-M4 Processor

2.3

Exception model
This section describes the exception model. It describes:
•
Exception states
•
Exception types
•
Exception handlers on page 2-23
•
Vector table on page 2-23
•
Exception priorities on page 2-24
•
Interrupt priority grouping on page 2-25
•
Exception entry and return on page 2-25.

2.3.1

Exception states
Each exception is in one of the following states:
Inactive

The exception is not active and not pending.

Pending

The exception is waiting to be serviced by the processor.
An interrupt request from a peripheral or from software can change the
state of the corresponding interrupt to pending.

Active

An exception that is being serviced by the processor but has not
completed.
Note
An exception handler can interrupt the execution of another exception
handler. In this case both exceptions are in the active state.

Active and pending
The exception is being serviced by the processor and there is a pending
exception from the same source.
2.3.2

Exception types
The exception types are:

ARM DUI 0553A
ID121610

Reset

Reset is invoked on power up or a warm reset. The exception model treats
reset as a special form of exception. When reset is asserted, the operation
of the processor stops, potentially at any point in an instruction. When
reset is deasserted, execution restarts from the address provided by the
reset entry in the vector table. Execution restarts as privileged execution
in Thread mode.

NMI

A Non Maskable Interrupt (NMI) can be signalled by a peripheral or
triggered by software. This is the highest priority exception other than
reset. It is permanently enabled and has a fixed priority of -2. NMIs cannot
be:
•
masked or prevented from activation by any other exception
•
preempted by any exception other than Reset.

HardFault

A HardFault is an exception that occurs because of an error during
exception processing, or because an exception cannot be managed by any
other exception mechanism. HardFaults have a fixed priority of -1,
meaning they have higher priority than any exception with configurable
priority.
Copyright © 2010 ARM. All rights reserved.
Non-Confidential

2-21

The Cortex-M4 Processor

MemManage
A MemManage fault is an exception that occurs because of a memory
protection related fault. The the fixed memory protection constraints
determines this fault, for both instruction and data memory transactions.
This fault is always used to abort instruction accesses to Execute Never
(XN) memory regions.
BusFault

A BusFault is an exception that occurs because of a memory related fault
for an instruction or data memory transaction. This might be from an error
detected on a bus in the memory system.

UsageFault

A UsageFault is an exception that occurs because of a fault related to
instruction execution. This includes:
•
an undefined instruction
•
an illegal unaligned access
•
invalid state on instruction execution
•
an error on exception return.
The following can cause a UsageFault when the core is configured to
report them:
•
an unaligned address on word and halfword memory access
•
division by zero.

SVCall

A supervisor call (SVC) is an exception that is triggered by the SVC
instruction. In an OS environment, applications can use SVC instructions to
access OS kernel functions and device drivers.

PendSV

PendSV is an interrupt-driven request for system-level service. In an OS
environment, use PendSV for context switching when no other exception
is active.

SysTick

A SysTick exception is an exception the system timer generates when it
reaches zero. Software can also generate a SysTick exception. In an OS
environment, the processor can use this exception as system tick.

Interrupt (IRQ)

A interrupt, or IRQ, is an exception signalled by a peripheral, or generated
by a software request. All interrupts are asynchronous to instruction
execution. In the system, peripherals use interrupts to communicate with
the processor.
Table 2-16 Properties of the different exception types

Exception
number a

IRQ
number a

Exception type

Priority

Vector address
or offset b

Activation

Reset

-3, the highest

0x00000004

Asynchronous

-14

NMI

-2

0x00000008

Asynchronous

-13

HardFault

-1

0x0000000C

-12

MemManage

Configurable c

0x00000010

Synchronous

-11

BusFault

Configurable c

0x00000014

Synchronous when precise,
asynchronous when imprecise

-10

UsageFault

Configurable c

0x00000018

Synchronous

7-10

Reserved

ARM DUI 0553A
ID121610

2-22

The Cortex-M4 Processor

Table 2-16 Properties of the different exception types (continued)
Exception
number a

IRQ
number a

Exception type

Priority

Vector address
or offset b

Activation

-5

SVCall

Configurable c

0x0000002C

Synchronous

12-13

Reserved

-2

PendSV

Configurable c

0x00000038

Asynchronous

-1

SysTick

Configurable c

0x0000003C

Asynchronous

Interrupt (IRQ)

Configurable d

0x00000040 e

Asynchronous

a. To simplify the software layer, the CMSIS only uses IRQ numbers and therefore uses negative values for exceptions other than
interrupts. The IPSR returns the Exception number, see Interrupt Program Status Register on page 2-6.
b. See Vector table for more information.
c. See System Handler Priority Registers on page 4-21.
d. See Interrupt Priority Registers on page 4-7.
e. Increasing in steps of 4.

For an asynchronous exception, other than reset, the processor can execute another instruction
between when the exception is triggered and when the processor enters the exception handler.
Privileged software can disable the exceptions that Table 2-16 on page 2-22 shows as having
configurable priority, see:
•
System Handler Control and State Register on page 4-23
•
Interrupt Clear-enable Registers on page 4-5.
For more information about HardFaults, MemManage faults, BusFaults, and UsageFaults, see
Fault handling on page 2-29.
2.3.3

Exception handlers
The processor handles exceptions using:
Interrupt Service Routines (ISRs)
The IRQ interrupts are the exceptions handled by ISRs.

2.3.4

Fault handlers

HardFault, MemManage fault, UsageFault, and BusFault are fault
exceptions handled by the fault handlers.

System handlers

NMI, PendSV, SVCall SysTick, and the fault exceptions are all system
exceptions that are handled by system handlers.

Vector table
The vector table contains the reset value of the stack pointer, and the start addresses, also called
exception vectors, for all exception handlers. Figure 2-2 on page 2-24 shows the order of the
exception vectors in the vector table. The least-significant bit of each vector must be 1,
indicating that the exception handler is Thumb code, see Thumb state on page 2-7.

ARM DUI 0553A
ID121610

2-23

The Cortex-M4 Processor

Exception number IRQ number
16+n

.
.
.
18

-1

-2

Offset

0x0040+4n
.
.
.
0x004C
0x0048
0x0044
0x0040
0x003C
0x0038

IRQn
.
.
.
IRQ2
IRQ1
IRQ0
Systick
PendSV
Reserved
Reserved for Debug

12
11

Vector

-5

0x002C

SVCall

Reserved

8
7
6

-10

-11

-12

-13

-14

0x0018
0x0014
0x0010
0x000C
0x0008
0x0004
0x0000

Usage fault
Bus fault
Memory management fault
Hard fault
NMI
Reset
Initial SP value

Figure 2-2 Vector table

On system reset, the vector table is fixed at address 0x00000000. Privileged software can write to
the VTOR to relocate the vector table start address to a different memory location, in the range
0x00000080 to 0x3FFFFF80, see Vector Table Offset Register on page 4-16.
2.3.5

Exception priorities
As Table 2-16 on page 2-22 shows, all exceptions have an associated priority, with:
•
a lower priority value indicating a higher priority
•
configurable priorities for all exceptions except Reset, HardFault, and NMI.
If software does not configure any priorities, then all exceptions with a configurable priority
have a priority of 0. For information about configuring exception priorities see
•
System Handler Priority Registers on page 4-21
•
Interrupt Priority Registers on page 4-7.
Note
Configurable priority values are in the range 0-. This means that the Reset, HardFault, and NMI
exceptions, with fixed negative priority values, always have higher priority than any other
exception.

ARM DUI 0553A
ID121610

2-24

The Cortex-M4 Processor

For example, assigning a higher priority value to IRQ[0] and a lower priority value to IRQ[1]
means that IRQ[1] has higher priority than IRQ[0]. If both IRQ[1] and IRQ[0] are asserted,
IRQ[1] is processed before IRQ[0].
If multiple pending exceptions have the same priority, the pending exception with the lowest
exception number takes precedence. For example, if both IRQ[0] and IRQ[1] are pending and
have the same priority, then IRQ[0] is processed before IRQ[1].
When the processor is executing an exception handler, the exception handler is preempted if a
higher priority exception occurs. If an exception occurs with the same priority as the exception
being handled, the handler is not preempted, irrespective of the exception number. However, the
status of the new interrupt changes to pending.
2.3.6

Interrupt priority grouping
To increase priority control in systems with interrupts, the NVIC supports priority grouping.
This divides each interrupt priority register entry into two fields:
•
an upper field that defines the group priority
•
a lower field that defines a subpriority within the group.
Only the group priority determines preemption of interrupt exceptions. When the processor is
executing an interrupt exception handler, another interrupt with the same group priority as the
interrupt being handled does not preempt the handler,
If multiple pending interrupts have the same group priority, the subpriority field determines the
order in which they are processed. If multiple pending interrupts have the same group priority
and subpriority, the interrupt with the lowest IRQ number is processed first.
For information about splitting the interrupt priority fields into group priority and subpriority,
see Application Interrupt and Reset Control Register on page 4-16.

2.3.7

Exception entry and return
Descriptions of exception handling use the following terms:
Preemption

When the processor is executing an exception handler, an exception can
preempt the exception handler if its priority is higher than the priority of
the exception being handled. See Interrupt priority grouping for more
information about preemption by an interrupt.
When one exception preempts another, the exceptions are called nested
exceptions. See Exception entry on page 2-26 more information.

Return

This occurs when the exception handler is completed, and:
•

there is no pending exception with sufficient priority to be serviced

•

the completed exception handler was not handling a late-arriving
exception.

The processor pops the stack and restores the processor state to the state it
had before the interrupt occurred. See Exception return on page 2-28 for
more information.
Tail-chaining

ARM DUI 0553A
ID121610

This mechanism speeds up exception servicing. On completion of an
exception handler, if there is a pending exception that meets the
requirements for exception entry, the stack pop is skipped and control
transfers to the new exception handler.

2-25

The Cortex-M4 Processor

Late-arriving

This mechanism speeds up preemption. If a higher priority exception
occurs during state saving for a previous exception, the processor switches
to handle the higher priority exception and initiates the vector fetch for
that exception. State saving is not affected by late arrival because the state
saved is the same for both exceptions. Therefore the state saving continues
uninterrupted. The processor can accept a late arriving exception until the
first instruction of the exception handler of the original exception enters
the execute stage of the processor. On return from the exception handler
of the late-arriving exception, the normal tail-chaining rules apply.

Exception entry
Exception entry occurs when there is a pending exception with sufficient priority and either:
•

the processor is in Thread mode

•

the new exception is of higher priority than the exception being handled, in which case
the new exception preempts the original exception.

When one exception preempts another, the exceptions are nested.
Sufficient priority means the exception has more priority than any limits set by the mask
registers, see Exception mask registers on page 2-7. An exception with less priority than this is
pending but is not handled by the processor.
When the processor takes an exception, unless the exception is a tail-chained or a late-arriving
exception, the processor pushes information onto the current stack. This operation is referred to
as stacking and the structure of eight data words is referred as the stack frame.
When using floating-point routines the Cortex-M4 processor automatically stacks the
architected floating-point state on exception entry. Figure 2-3 on page 2-27 shows the
Cortex-M4 stack frame layout when floating-point state is preserved on the stack as the result
of an interrupt or an exception.
Note
Where stack space for floating-point state is not allocated, the stack frame is the same as that of
ARMv7-M implementations without an FPU. Figure 2-3 on page 2-27 shows this stack frame
also.

ARM DUI 0553A
ID121610

2-26

The Cortex-M4 Processor

...
{aligner}
FPSCR
S15
S14
S13
S12
S11
S10
S9
S8
S7
S6
S5
S4
S3
S2
S1
S0
xPSR
PC
LR
R12
R3
R2
R1
R0
Exception frame with
floating-point storage

Pre-IRQ top of stack

Decreasing
memory
address

IRQ top of stack

...
{aligner}
xPSR
PC
LR
R12
R3
R2
R1
R0

Pre-IRQ top of stack

IRQ top of stack

Exception frame without
floating-point storage

Figure 2-3 Exception stack frame

Immediately after stacking, the stack pointer indicates the lowest address in the stack frame. The
alignment of the stack frame is controlled via the STKALIGN bit of the Configuration Control
Register (CCR).
The stack frame includes the return address. This is the address of the next instruction in the
interrupted program. This value is restored to the PC at exception return so that the interrupted
program resumes.
In parallel to the stacking operation, the processor performs a vector fetch that reads the
exception handler start address from the vector table. When stacking is complete, the processor
starts executing the exception handler. At the same time, the processor writes an
EXC_RETURN value to the LR. This indicates which stack pointer corresponds to the stack
frame and what operation mode the processor was in before the entry occurred.
If no higher priority exception occurs during exception entry, the processor starts executing the
exception handler and automatically changes the status of the corresponding pending interrupt
to active.
If another higher priority exception occurs during exception entry, the processor starts executing
the exception handler for this exception and does not change the pending status of the earlier
exception. This is the late arrival case.

ARM DUI 0553A
ID121610

2-27

The Cortex-M4 Processor

Exception return
Exception return occurs when the processor is in Handler mode and executes one of the
following instructions to load the EXC_RETURN value into the PC:
•
an LDM or POP instruction that loads the PC
•
an LDR instruction with PC as the destination
•
a BX instruction using any register.
EXC_RETURN is the value loaded into the LR on exception entry. The exception mechanism
relies on this value to detect when the processor has completed an exception handler. The lowest
five bits of this value provide information on the return stack and processor mode. Table 2-17
shows the EXC_RETURN values with a description of the exception return behavior.
All EXC_RETURN values have bits[31:5] set to one. When this value is loaded into the PC it
indicates to the processor that the exception is complete, and the processor initiates the
appropriate exception return sequence.
Table 2-17 Exception return behavior

ARM DUI 0553A
ID121610

EXC_RETURN[31:0]

Description

0xFFFFFFF1

Return to Handler mode, exception return uses non-floating-point state
from the MSP and execution uses MSP after return.

0xFFFFFFF9

Return to Thread mode, exception return uses non-floating-point state from
MSP and execution uses MSP after return.

0xFFFFFFFD

Return to Thread mode, exception return uses non-floating-point state from
the PSP and execution uses PSP after return.

0xFFFFFFE1

Return to Handler mode, exception return uses floating-point-state from
MSP and execution uses MSP after return.

0xFFFFFFE9

Return to Thread mode, exception return uses floating-point state from
MSP and execution uses MSP after return.

0xFFFFFFED

Return to Thread mode, exception return uses floating-point state from PSP
and execution uses PSP after return.

2-28

The Cortex-M4 Processor

2.4

Fault handling
Faults are a subset of the exceptions, see Exception model on page 2-21. Faults are generated by:

2.4.1

•

a bus error on:
— an instruction fetch or vector table load
— a data access.

•

an internally-detected error such as an undefined instruction

•

attempting to execute an instruction from a memory region marked as Execute-never
(XN).

•

If your device contains an MPU, a privilege violation or an attempt to access an
unmanaged region causing an MPU fault.

Fault types
Table 2-18 shows the types of fault, the handler used for the fault, the corresponding fault status
register, and the register bit that indicates that the fault has occurred. See Configurable Fault
Status Register on page 4-24 for more information about the fault status registers.
Table 2-18 Faults
Fault

Handler

Bit name

Fault status register

Bus error on a vector read

HardFault

VECTTBL

HardFault Status Register on
page 4-30

Fault escalated to a hard fault

FORCED

MPU or default memory map mismatch:

on instruction access

IACCVIOL a

on data access

DACCVIOL

MemManage Fault Address
Register on page 4-30

during exception stacking

MSTKERR

during exception unstacking

MUNSKERR

during lazy floating-point state preservation

MLSPERR

Bus error:

MemManage

BusFault

during exception stacking

STKERR

during exception unstacking

UNSTKERR

BusFault Status Register on
page 4-26

during instruction prefetch

IBUSERR

during lazy floating-point state preservation

LSPERR

Precise data bus error

PRECISERR

Imprecise data bus error

IMPRECISERR

ARM DUI 0553A
ID121610

2-29

The Cortex-M4 Processor

Table 2-18 Faults (continued)
Fault

Handler

Bit name

Fault status register

Attempt to access a coprocessor

UsageFault

NOCP

UsageFault Status Register on
page 4-28

Undefined instruction

UNDEFINSTR

Attempt to enter an invalid instruction set state b

INVSTATE

Invalid EXC_RETURN value

INVPC

Illegal unaligned load or store

UNALIGNED

Divide By 0

DIVBYZERO

a. Occurs on an access to an XN region even if the processor does not include an MPU or if the MPU is disabled.
b. Attempting to use an instruction set other than the Thumb instruction set or returns to a non load/store-multiple instruction with
ICI continuation.

2.4.2

Fault escalation and hard faults
All faults exceptions except for HardFault have configurable exception priority, see System
Handler Priority Registers on page 4-21. Software can disable execution of the handlers for
these faults, see System Handler Control and State Register on page 4-23.
Usually, the exception priority, together with the values of the exception mask registers,
determines whether the processor enters the fault handler, and whether a fault handler can
preempt another fault handler. as described in Exception model on page 2-21.
In some situations, a fault with configurable priority is treated as a HardFault. This is called
priority escalation, and the fault is described as escalated to HardFault. Escalation to HardFault
occurs when:
•
A fault handler causes the same kind of fault as the one it is servicing. This escalation to
HardFault occurs because a fault handler cannot preempt itself because it must have the
same priority as the current priority level.
•
A fault handler causes a fault with the same or lower priority as the fault it is servicing.
This is because the handler for the new fault cannot preempt the currently executing fault
handler.
•
An exception handler causes a fault for which the priority is the same as or lower than the
currently executing exception.
•
A fault occurs and the handler for that fault is not enabled.
If a BusFault occurs during a stack push when entering a BusFault handler, the BusFault does
not escalate to a HardFault. This means that if a corrupted stack causes a fault, the fault handler
executes even though the stack push for the handler failed. The fault handler operates but the
stack contents are corrupted.
Note
Only Reset and NMI can preempt the fixed priority HardFault. A HardFault can preempt any
exception other than Reset, NMI, or another HardFault.

ARM DUI 0553A
ID121610

2-30

The Cortex-M4 Processor

2.4.3

Fault status registers and fault address registers
The fault status registers indicate the cause of a fault. For BusFaults and MemManage faults,
the fault address register indicates the address accessed by the operation that caused the fault,
as shown in Table 2-19.
Table 2-19 Fault status and fault address registers

2.4.4

Handler

Status register
name

Address register
name

HardFault

HFSR

HardFault Status Register on page 4-30

MemManage

MMFSR

MMFAR

MemManage Fault Status Register on page 4-25
MemManage Fault Address Register on page 4-30

BusFault

BFSR

BFAR

BusFault Status Register on page 4-26
BusFault Address Register on page 4-31

UsageFault

UFSR

UsageFault Status Register on page 4-28

Lockup
The processor enters a lockup state if a fault occurs when executing the NMI or HardFault
handlers. When the processor is in lockup state it does not execute any instructions. The
processor remains in lockup state until either:
•
it is reset
•
an NMI occurs
•
it is halted by a debugger.
Note
If lockup state occurs from the NMI handler a subsequent NMI does not cause the processor to
leave lockup state.

ARM DUI 0553A
ID121610

2-31

The Cortex-M4 Processor

2.5

Power management
The Cortex-M4 processor sleep modes reduce power consumption. The sleep modes your
device implements are implementation-defined. The modes can be one or both of the following:
•
sleep mode stops the processor clock
•
deep sleep mode stops the system clock and switches off the PLL and flash memory.
If your device implements two sleep modes providing different levels of power saving, the
SLEEPDEEP bit of the SCR selects which sleep mode is used, see System Control Register on
page 4-19. For more information about the behavior of the sleep modes see the documentation
supplied by your device vendor.
This section describes the mechanisms for entering sleep mode, and the conditions for waking
up from sleep mode.

2.5.1

Entering sleep mode
This section describes the mechanisms software can use to put the processor into sleep mode.
The system can generate spurious wakeup events, for example a debug operation wakes up the
processor. Therefore software must be able to put the processor back into sleep mode after such
an event. A program might have an idle loop to put the processor back to sleep mode.
Wait for interrupt
The Wait For Interrupt instruction, WFI, causes immediate entry to sleep mode unless the
wake-up condition is true, see Wakeup from WFI or sleep-on-exit on page 2-33. When the
processor executes a WFI instruction it stops executing instructions and enters sleep mode. See
WFI on page 3-169 for more information.
Wait for event
The Wait For Event instruction, WFE, causes entry to sleep mode depending on the value of a
one-bit event register. When the processor executes a WFE instruction, it checks the value of the
event register:
0

The processor stops executing instructions and enters sleep mode.

The processor clears the register to 0 and continues executing instructions without
entering sleep mode.

See WFE on page 3-168 for more information.
If the event register is 1, this indicates that the processor must not enter sleep mode on execution
of a WFE instruction. Typically, this is because an external event signal is asserted, or a processor
in the system has executed an SEV instruction, see SEV on page 3-166. Software cannot access
this register directly.
Sleep-on-exit
If the SLEEPONEXIT bit of the SCR is set to 1, when the processor completes the execution of
all exception handlers it returns to Thread mode and immediately enters sleep mode. Use this
mechanism in applications that only require the processor to run when an exception occurs.

ARM DUI 0553A
ID121610

2-32

The Cortex-M4 Processor

2.5.2

Wakeup from sleep mode
The conditions for the processor to wakeup depend on the mechanism that cause it to enter sleep
mode.
Wakeup from WFI or sleep-on-exit
Normally, the processor wakes up only when it detects an exception with sufficient priority to
cause exception entry. Some embedded systems might have to execute system restore tasks after
the processor wakes up, and before it executes an interrupt handler. To achieve this set the
PRIMASK bit to 1 and the FAULTMASK bit to 0. If an interrupt arrives that is enabled and has
a higher priority than current exception priority, the processor wakes up but does not execute the
interrupt handler until the processor sets PRIMASK to zero. For more information about
PRIMASK and FAULTMASK see Exception mask registers on page 2-7.
Wakeup from WFE
The processor wakes up if:
•
it detects an exception with sufficient priority to cause exception entry
•
it detects an external event signal, see The external event input
•
in a multiprocessor system, another processor in the system executes an SEV instruction.
In addition, if the SEVONPEND bit in the SCR is set to 1, any new pending interrupt triggers
an event and wakes up the processor, even if the interrupt is disabled or has insufficient priority
to cause exception entry. For more information about the SCR see System Control Register on
page 4-19.

2.5.3

The optional Wakeup Interrupt Controller
Your device might include a Wakeup Interrupt Controller (WIC), an optional peripheral that can
detect an interrupt and wake the processor from deep sleep mode. The WIC is enabled only
when the DEEPSLEEP bit in the SCR is set to 1, see System Control Register on page 4-19.
The WIC is not programmable, and does not have any registers or user interface. It operates
entirely from hardware signals.
When the WIC is enabled and the processor enters deep sleep mode, the power management unit
in the system can power down most of the Cortex-M4 processor. This has the side effect of
stopping the SysTick timer. When the WIC receives an interrupt, it takes a number of clock
cycles to wakeup the processor and restore its state, before it can process the interrupt. This
means interrupt latency is increased in deep sleep mode.
Note
If the processor detects a connection to a debugger it disables the WIC.

2.5.4

The external event input
Your device might include an external event input signal, so that device peripherals can signal
the processor, to either:

ARM DUI 0553A
ID121610

•

wake the processor from WFE

•

set the internal WFE event register to one to indicate that the processor must not enter
sleep mode on a later WFE instruction.

2-33

The Cortex-M4 Processor

See Wait for event on page 2-32 and the documentation supplied by your device vendor for more
information about this signal.
2.5.5

Power management programming hints
ISO/IEC C cannot directly generate the WFI and WFE instructions. The CMSIS provides the
following functions for these instructions:
void __WFE(void) // Wait for Event
void __WFI(void) // Wait for Interrupt

ARM DUI 0553A
ID121610

2-34

Chapter 3
The Cortex-M4 Instruction Set

This chapter is the reference material for the Cortex-M4 instruction set description in a User Guide.
The following sections give general information:
•
Instruction set summary on page 3-2
•
CMSIS functions on page 3-9
•
About the instruction descriptions on page 3-11.
Each of the following sections describes a functional group of Cortex-M4 instructions. Together
they describe all the instructions supported by the Cortex-M4 processor:
•
Memory access instructions on page 3-22
•
General data processing instructions on page 3-39
•
Multiply and divide instructions on page 3-74
•
Saturating instructions on page 3-95
•
Packing and unpacking instructions on page 3-107
•
Bitfield instructions on page 3-114
•
Branch and control instructions on page 3-118
•
Miscellaneous instructions on page 3-157
•
Floating-point instructions on page 3-126.

ARM DUI 0553A
ID121610

3-1

The Cortex-M4 Instruction Set

3.1

Instruction set summary
The processor implements a version of the Thumb instruction set. Table 3-1 lists the supported
instructions.
Note
In Table 3-1:
•
angle brackets, <>, enclose alternative forms of the operand
•
braces, {}, enclose optional operands
•
the Operands column is not exhaustive
•
Op2 is a flexible second operand that can be either a register or a constant
•
most instructions can use an optional condition code suffix.
For more information on the instructions and operands, see the instruction descriptions.

Table 3-1 Cortex-M4 instructions
Mnemonic

Operands

Brief description

Flags

Page

ADC, ADCS

{Rd,} Rn, Op2

Add with Carry

N,Z,C,V

page 3-41

ADD, ADDS

{Rd,} Rn, Op2

Add

N,Z,C,V

page 3-41

ADD, ADDW

{Rd,} Rn, #imm12

Add

page 3-41

ADR

Rd, label

Load PC-relative Address

page 3-23

AND, ANDS

{Rd,} Rn, Op2

Logical AND

N,Z,C

page 3-44

ASR, ASRS

Rd, Rm,

Arithmetic Shift Right

N,Z,C

page 3-46

label

Branch

page 3-119

BFC

Rd, #lsb, #width

Bit Field Clear

page 3-115

BFI

Rd, Rn, #lsb, #width

Bit Field Insert

page 3-115

BIC, BICS

{Rd,} Rn, Op2

Bit Clear

N,Z,C

page 3-44

BKPT

#imm

Breakpoint

page 3-158

label

Branch with Link

page 3-119

BLX

Branch indirect with Link

page 3-119

Branch indirect

page 3-119

CBNZ

Rn, label

Compare and Branch if Non Zero

page 3-121

CBZ

Rn, label

Compare and Branch if Zero

page 3-121

CLREX

Clear Exclusive

page 3-38

CLZ

Rd, Rm

Count Leading Zeros

page 3-48

CMN

Rn, Op2

Compare Negative

N,Z,C,V

page 3-49

CMP

Rn, Op2

Compare

N,Z,C,V

page 3-49

CPSID

Change Processor State, Disable Interrupts

page 3-159

CPSIE

Change Processor State, Enable Interrupts

page 3-159

ARM DUI 0553A
ID121610

3-2

The Cortex-M4 Instruction Set

Table 3-1 Cortex-M4 instructions (continued)
Mnemonic

Operands

Brief description

Flags

Page

DMB

Data Memory Barrier

page 3-160

DSB

Data Synchronization Barrier

page 3-161

EOR, EORS

{Rd,} Rn, Op2

Exclusive OR

N,Z,C

page 3-44

ISB

Instruction Synchronization Barrier

page 3-162

If-Then condition block

page 3-122

LDM

Rn{!}, reglist

Load Multiple registers, increment after

page 3-32

LDMDB, LDMEA

Rn{!}, reglist

Load Multiple registers, decrement before

page 3-32

LDMFD, LDMIA

Rn{!}, reglist

Load Multiple registers, increment after

page 3-32

LDR

Rt, [Rn, #offset]

Load Register with word

page 3-22

LDRB, LDRBT

Rt, [Rn, #offset]

Load Register with byte

page 3-22

LDRD

Rt, Rt2, [Rn, #offset]

Load Register with two bytes

page 3-24

LDREX

Rt, [Rn, #offset]

Load Register Exclusive

page 3-36

LDREXB

Rt, [Rn]

Load Register Exclusive with Byte

page 3-36

LDREXH

Rt, [Rn]

Load Register Exclusive with Halfword

page 3-36

LDRH, LDRHT

Rt, [Rn, #offset]

Load Register with Halfword

page 3-22

LDRSB, LDRSBT

Rt, [Rn, #offset]

Load Register with Signed Byte

page 3-22

LDRSH, LDRSHT

Rt, [Rn, #offset]

Load Register with Signed Halfword

page 3-22

LDRT

Rt, [Rn, #offset]

Load Register with word

page 3-22

LSL, LSLS

Rd, Rm,

Logical Shift Left

N,Z,C

page 3-46

LSR, LSRS

Rd, Rm,

Logical Shift Right

N,Z,C

page 3-46

MLA

Rd, Rn, Rm, Ra

Multiply with Accumulate, 32-bit result

page 3-75

MLS

Rd, Rn, Rm, Ra

Multiply and Subtract, 32-bit result

page 3-75

MOV, MOVS

Rd, Op2

Move

N,Z,C

page 3-50

MOVT

Rd, #imm16

Move Top

page 3-52

MOVW, MOV

Rd, #imm16

Move 16-bit constant

N,Z,C

page 3-50

MRS

Rd, spec_reg

Move from Special Register to general register

page 3-163

MSR

spec_reg, Rm

Move from general register to Special Register

N,Z,C,V

page 3-164

MUL, MULS

{Rd,} Rn, Rm

Multiply, 32-bit result

N,Z

page 3-75

MVN, MVNS

Rd, Op2

Move NOT

N,Z,C

page 3-50

NOP

No Operation

page 3-165

ORN, ORNS

{Rd,} Rn, Op2

Logical OR NOT

N,Z,C

page 3-44

ORR, ORRS

{Rd,} Rn, Op2

Logical OR

N,Z,C

page 3-44

PKHTB, PKHBT

{Rd,} Rn, Rm, Op2

Pack Halfword

page 3-108

ARM DUI 0553A
ID121610

3-3

The Cortex-M4 Instruction Set

Table 3-1 Cortex-M4 instructions (continued)
Mnemonic

Operands

Brief description

Flags

Page

POP

reglist

Pop registers from stack

page 3-34

PUSH

reglist

Push registers onto stack

page 3-34

QADD

{Rd,} Rn, Rm

Saturating double and Add

page 3-98

QADD16

{Rd,} Rn, Rm

Saturating Add 16

page 3-98

QADD8

{Rd,} Rn, Rm

Saturating Add 8

page 3-98

QASX

{Rd,} Rn, Rm

Saturating Add and Subtract with Exchange

page 3-100

QDADD

{Rd,} Rn, Rm

Saturating Add

page 3-102

QDSUB

{Rd,} Rn, Rm

Saturating double and Subtract

page 3-102

QSAX

{Rd,} Rn, Rm

Saturating Subtract and Add with Exchange

page 3-100

QSUB

{Rd,} Rn, Rm

Saturating Subtract

page 3-102

QSUB16

{Rd,} Rn, Rm

Saturating Subtract 16

page 3-102

QSUB8

{Rd,} Rn, Rm

Saturating Subtract 8

page 3-102

RBIT

Rd, Rn

Reverse Bits

page 3-53

REV

Rd, Rn

Reverse byte order in a word

page 3-53

REV16

Rd, Rn

Reverse byte order in each halfword

page 3-53

REVSH

Rd, Rn

Reverse byte order in bottom halfword and sign extend

page 3-53

ROR, RORS

Rd, Rm,

Rotate Right

N,Z,C

page 3-46

RRX, RRXS

Rd, Rm

Rotate Right with Extend

N,Z,C

page 3-46

RSB, RSBS

{Rd,} Rn, Op2

Reverse Subtract

N,Z,C,V

page 3-41

SADD16

{Rd,} Rn, Rm

Signed Add 16

page 3-54

SADD8

{Rd,} Rn, Rm

Signed Add 8

page 3-54

SASX

{Rd,} Rn, Rm

Signed Add and Subtract with Exchange

page 3-60

SBC, SBCS

{Rd,} Rn, Op2

Subtract with Carry

N,Z,C,V

page 3-41

SBFX

Rd, Rn, #lsb, #width

Signed Bit Field Extract

page 3-116

SDIV

{Rd,} Rn, Rm

Signed Divide

page 3-94

SEL

{Rd,} Rn, Rm

Select bytes

page 3-70

SEV

Send Event

page 3-166

SHADD16

{Rd,} Rn, Rm

Signed Halving Add 16

page 3-55

SHADD8

{Rd,} Rn, Rm

Signed Halving Add 8

page 3-55

SHASX

{Rd,} Rn, Rm

Signed Halving Add and Subtract with Exchange

page 3-56

SHSAX

{Rd,} Rn, Rm

Signed Halving Subtract and Add with Exchange

page 3-56

SHSUB16

{Rd,} Rn, Rm

Signed Halving Subtract 16

page 3-58

SHSUB8

{Rd,} Rn, Rm

Signed Halving Subtract 8

page 3-58

ARM DUI 0553A
ID121610

3-4

The Cortex-M4 Instruction Set

Table 3-1 Cortex-M4 instructions (continued)
Mnemonic

Operands

Brief description

Flags

Page

SMLABB, SMLABT,
SMLATB, SMLATT

Rd, Rn, Rm, Ra

Signed Multiply Accumulate Long (halfwords)

page 3-79

SMLAD, SMLADX

Rd, Rn, Rm, Ra

Signed Multiply Accumulate Dual

page 3-81

SMLAL

RdLo, RdHi, Rn, Rm

Signed Multiply with Accumulate (32 x 32 + 64), 64-bit
result

page 3-93

SMLALBB, SMLALBT,
SMLALTB, SMLALTT

RdLo, RdHi, Rn, Rm

Signed Multiply Accumulate Long, halfwords

page 3-82

SMLALD, SMLALDX

RdLo, RdHi, Rn, Rm

Signed Multiply Accumulate Long Dual

page 3-82

SMLAWB, SMLAWT

Rd, Rn, Rm, Ra

Signed Multiply Accumulate, word by halfword

page 3-79

SMLSD

Rd, Rn, Rm, Ra

Signed Multiply Subtract Dual

page 3-84

SMLSLD

RdLo, RdHi, Rn, Rm

Signed Multiply Subtract Long Dual

SMMLA

Rd, Rn, Rm, Ra

Signed Most significant word Multiply Accumulate

page 3-86

SMMLS, SMMLR

Rd, Rn, Rm, Ra

Signed Most significant word Multiply Subtract

page 3-86

SMMUL, SMMULR

{Rd,} Rn, Rm

Signed Most significant word Multiply

page 3-88

SMUAD

{Rd,} Rn, Rm

Signed dual Multiply Add

page 3-89

SMULBB, SMULBT
SMULTB, SMULTT

{Rd,} Rn, Rm

Signed Multiply (halfwords)

page 3-91

SMULL

RdLo, RdHi, Rn, Rm

Signed Multiply (32 x 32), 64-bit result

page 3-93

SMULWB, SMULWT

{Rd,} Rn, Rm

Signed Multiply word by halfword

page 3-91

SMUSD, SMUSDX

{Rd,} Rn, Rm

Signed dual Multiply Subtract

page 3-89

SSAT

Rd, #n, Rm {,shift #s}

Signed Saturate

page 3-96

SSAT16

Rd, #n, Rm

Signed Saturate 16

page 3-97

SSAX

{Rd,} Rn, Rm

Signed Subtract and Add with Exchange

page 3-60

SSUB16

{Rd,} Rn, Rm

Signed Subtract 16

page 3-59

SSUB8

{Rd,} Rn, Rm

Signed Subtract 8

page 3-59

STM

Rn{!}, reglist

Store Multiple registers, increment after

page 3-32

STMDB, STMEA

Rn{!}, reglist

Store Multiple registers, decrement before

page 3-32

STMFD, STMIA

Rn{!}, reglist

Store Multiple registers, increment after

page 3-32

STR

Rt, [Rn, #offset]

Store Register word

page 3-22

STRB, STRBT

Rt, [Rn, #offset]

Store Register byte

page 3-22

STRD

Rt, Rt2, [Rn, #offset]

Store Register two words

page 3-24

STREX

Rd, Rt, [Rn, #offset]

Store Register Exclusive

page 3-36

STREXB

Rd, Rt, [Rn]

Store Register Exclusive Byte

page 3-36

STREXH

Rd, Rt, [Rn]

Store Register Exclusive Halfword

page 3-36

ARM DUI 0553A
ID121610

page 3-84

3-5

The Cortex-M4 Instruction Set

Table 3-1 Cortex-M4 instructions (continued)
Mnemonic

Operands

Brief description

Flags

Page

STRH, STRHT

Rt, [Rn, #offset]

Store Register Halfword

page 3-22

STRT

Rt, [Rn, #offset]

Store Register word

page 3-22

SUB, SUBS

{Rd,} Rn, Op2

Subtract

N,Z,C,V

page 3-41

SUB, SUBW

{Rd,} Rn, #imm12

Subtract

page 3-41

SVC

#imm

Supervisor Call

page 3-167

SXTAB

{Rd,} Rn, Rm,{,ROR #}

Extend 8 bits to 32 and add

page 3-112

SXTAB16

{Rd,} Rn, Rm,{,ROR #}

Dual extend 8 bits to 16 and add

page 3-112

SXTAH

{Rd,} Rn, Rm,{,ROR #}

Extend 16 bits to 32 and add

page 3-112

SXTB16

{Rd,} Rm {,ROR #n}

Signed Extend Byte 16

page 3-110

SXTB

{Rd,} Rm {,ROR #n}

Sign extend a byte

page 3-117

SXTH

{Rd,} Rm {,ROR #n}

Sign extend a halfword

page 3-117

TBB

[Rn, Rm]

Table Branch Byte

page 3-124

TBH

[Rn, Rm, LSL #1]

Table Branch Halfword

page 3-124

TEQ

Rn, Op2

Test Equivalence

N,Z,C

page 3-62

TST

Rn, Op2

Test

N,Z,C

page 3-62

UADD16

{Rd,} Rn, Rm

Unsigned Add 16

page 3-63

UADD8

{Rd,} Rn, Rm

Unsigned Add 8

page 3-63

USAX

{Rd,} Rn, Rm

Unsigned Subtract and Add with Exchange

page 3-64

UHADD16

{Rd,} Rn, Rm

Unsigned Halving Add 16

page 3-66

UHADD8

{Rd,} Rn, Rm

Unsigned Halving Add 8

page 3-66

UHASX

{Rd,} Rn, Rm

Unsigned Halving Add and Subtract with Exchange

page 3-67

UHSAX

{Rd,} Rn, Rm

Unsigned Halving Subtract and Add with Exchange

page 3-67

UHSUB16

{Rd,} Rn, Rm

Unsigned Halving Subtract 16

page 3-69

UHSUB8

{Rd,} Rn, Rm

Unsigned Halving Subtract 8

page 3-69

UBFX

Rd, Rn, #lsb, #width

Unsigned Bit Field Extract

page 3-116

UDIV

{Rd,} Rn, Rm

Unsigned Divide

page 3-94

UMAAL

RdLo, RdHi, Rn, Rm

Unsigned Multiply Accumulate Accumulate Long (32 x
32 + 32 +32), 64-bit result

page 3-77

UMLAL

RdLo, RdHi, Rn, Rm

Unsigned Multiply with Accumulate (32 x 32 + 64),
64-bit result

page 3-93

UMULL

RdLo, RdHi, Rn, Rm

Unsigned Multiply (32 x 32), 64-bit result

page 3-93

UQADD16

{Rd,} Rn, Rm

Unsigned Saturating Add 16

page 3-105

UQADD8

{Rd,} Rn, Rm

Unsigned Saturating Add 8

page 3-105

ARM DUI 0553A
ID121610

3-6

The Cortex-M4 Instruction Set

Table 3-1 Cortex-M4 instructions (continued)
Mnemonic

Operands

Brief description

Flags

Page

UQASX

{Rd,} Rn, Rm

Unsigned Saturating Add and Subtract with Exchange

page 3-103

UQSAX

{Rd,} Rn, Rm

Unsigned Saturating Subtract and Add with Exchange

page 3-103

UQSUB16

{Rd,} Rn, Rm

Unsigned Saturating Subtract 16

page 3-105

UQSUB8

{Rd,} Rn, Rm

Unsigned Saturating Subtract 8

page 3-105

USAD8

{Rd,} Rn, Rm

Unsigned Sum of Absolute Differences

page 3-71

USADA8

{Rd,} Rn, Rm, Ra

Unsigned Sum of Absolute Differences and Accumulate

page 3-72

USAT

Rd, #n, Rm {,shift #s}

Unsigned Saturate

page 3-96

USAT16

Rd, #n, Rm

Unsigned Saturate 16

page 3-97

UASX

{Rd,} Rn, Rm

Unsigned Add and Subtract with Exchange

page 3-64

USUB16

{Rd,} Rn, Rm

Unsigned Subtract 16

page 3-73

USUB8

{Rd,} Rn, Rm

Unsigned Subtract 8

page 3-73

UXTAB

{Rd,} Rn, Rm,{,ROR #}

Rotate, extend 8 bits to 32 and Add

page 3-112

UXTAB16

{Rd,} Rn, Rm,{,ROR #}

Rotate, dual extend 8 bits to 16 and Add

page 3-112

UXTAH

{Rd,} Rn, Rm,{,ROR #}

Rotate, unsigned extend and Add Halfword

page 3-112

UXTB

{Rd,} Rm {,ROR #n}

Zero extend a Byte

page 3-117

UXTB16

{Rd,} Rm {,ROR #n}

Unsigned Extend Byte 16

page 3-117

UXTH

{Rd,} Rm {,ROR #n}

Zero extend a Halfword

page 3-117

VABS.F32

Sd, Sm

Floating-point Absolute

page 3-128

VADD.F32

{Sd,} Sn, Sm

Floating-point Add

page 3-129

VCMP.F32

Sd,

Compare two floating-point registers, or one
floating-point register and zero

FPSCR

page 3-130

VCMPE.F32

Sd,

Compare two floating-point registers, or one
floating-point register and zero with Invalid Operation
check

FPSCR

page 3-130

VCVT.S32.F32

Sd, Sm

Convert between floating-point and integer

page 3-131

VCVT.S16.F32

Sd, Sd, #fbits

Convert between floating-point and fixed point

page 3-132

VCVTR.S32.F32

Sd, Sm

Convert between floating-point and integer with
rounding

page 3-131

VCVT.F32.F16

Sd, Sm

Converts half-precision value to single-precision

page 3-133

VCVTT.F32.F16

Sd, Sm

Converts single-precision register to half-precision

page 3-133

VDIV.F32

{Sd,} Sn, Sm

Floating-point Divide

page 3-134

VFMA.F32

{Sd,} Sn, Sm

Floating-point Fused Multiply Accumulate

page 3-135

VFNMA.F32

{Sd,} Sn, Sm

Floating-point Fused Negate Multiply Accumulate

page 3-136

VFMS.F32

{Sd,} Sn, Sm

Floating-point Fused Multiply Subtract

page 3-135

ARM DUI 0553A
ID121610

3-7

The Cortex-M4 Instruction Set

Table 3-1 Cortex-M4 instructions (continued)
Mnemonic

Operands

Brief description

Flags

Page

VFNMS.F32

{Sd,} Sn, Sm

Floating-point Fused Negate Multiply Subtract

page 3-136

VLDM.F<32|64>

Rn{!}, list

Load Multiple extension registers

page 3-137

VLDR.F<32|64>

, [Rn]

Load an extension register from memory

page 3-138

VLMA.F32

{Sd,} Sn, Sm

Floating-point Multiply Accumulate

page 3-139

VLMS.F32

{Sd,} Sn, Sm

Floating-point Multiply Subtract

page 3-139

VMOV.F32

Sd, #imm

Floating-point Move immediate

page 3-140

VMOV

Sd, Sm

Floating-point Move register

page 3-141

VMOV

Sn, Rt

Copy ARM core register to single precision

page 3-143

VMOV

Sm, Sm1, Rt, Rt2

Copy 2 ARM core registers to 2 single precision

page 3-144

VMOV

Dd[x], Rt

Copy ARM core register to scalar

page 3-145

VMOV

Rt, Dn[x]

Copy scalar to ARM core register

page 3-142

VMRS

Rt, FPSCR

Move FPSCR to ARM core register or APSR

N,Z,C,V

page 3-146

VMSR

FPSCR, Rt

Move to FPSCR from ARM Core register

FPSCR

page 3-147

VMUL.F32

{Sd,} Sn, Sm

Floating-point Multiply

page 3-148

VNEG.F32

Sd, Sm

Floating-point Negate

page 3-149

VNMLA.F32

Sd, Sn, Sm

Floating-point Multiply and Add

page 3-150

VNMLS.F32

Sd, Sn, Sm

Floating-point Multiply and Subtract

page 3-150

VNMUL

{Sd,} Sn, Sm

Floating-point Multiply

page 3-150

VPOP

list

Pop extension registers

page 3-151

VPUSH

list

Push extension registers

page 3-152

VSQRT.F32

Sd, Sm

Calculates floating-point Square Root

page 3-153

VSTM

Rn{!}, list

Floating-point register Store Multiple

page 3-154

VSTR.F<32|64>

Sd, [Rn]

Stores an extension register to memory

page 3-155

VSUB.F<32|64>

{Sd,} Sn, Sm

Floating-point Subtract

page 3-156

WFE

Wait For Event

page 3-168

WFI

Wait For Interrupt

page 3-169

ARM DUI 0553A
ID121610

3-8

The Cortex-M4 Instruction Set

3.2

CMSIS functions
ISO/IEC C code cannot directly access some Cortex-M4 instructions. This section describes
intrinsic functions that can generate these instructions, provided by the CMSIS and that might
be provided by a C compiler. If a C compiler does not support an appropriate intrinsic function,
you might have to use inline assembler to access some instructions.
The CMSIS provides the following intrinsic functions to generate instructions that ISO/IEC C
code cannot directly access:
Table 3-2 CMSIS functions to generate some Cortex-M4 instructions
Instruction

CMSIS function

CPSIE I

void __enable_irq(void)

CPSID I

void __disable_irq(void)

CPSIE F

void __enable_fault_irq(void)

CPSID F

void __disable_fault_irq(void)

ISB

void __ISB(void)

DSB

void __DSB(void)

DMB

void __DMB(void)

REV

uint32_t __REV(uint32_t int value)

REV16

uint32_t __REV16(uint32_t int value)

REVSH

uint32_t __REVSH(uint32_t int value)

RBIT

uint32_t __RBIT(uint32_t int value)

SEV

void __SEV(void)

WFE

void __WFE(void)

WFI

void __WFI(void)

The CMSIS also provides a number of functions for accessing the special registers using MRS and
MSR instructions:
Table 3-3 CMSIS functions to access the special registers
Special register

Access

CMSIS function

PRIMASK

Read

uint32_t __get_PRIMASK (void)

Write

void __set_PRIMASK (uint32_t value)

Read

uint32_t __get_FAULTMASK (void)

Write

void __set_FAULTMASK (uint32_t value)

Read

uint32_t __get_BASEPRI (void)

Write

void __set_BASEPRI (uint32_t value)

Read

uint32_t __get_CONTROL (void)

Write

void __set_CONTROL (uint32_t value)

FAULTMASK

BASEPRI

CONTROL

ARM DUI 0553A
ID121610

3-9

The Cortex-M4 Instruction Set

Table 3-3 CMSIS functions to access the special registers (continued)
Special register

Access

CMSIS function

MSP

Read

uint32_t __get_MSP (void)

Write

void __set_MSP (uint32_t TopOfMainStack)

Read

uint32_t __get_PSP (void)

Write

void __set_PSP (uint32_t TopOfProcStack)

PSP

ARM DUI 0553A
ID121610

3-10

The Cortex-M4 Instruction Set

3.3

About the instruction descriptions
The following sections give more information about using the instructions:
•
Operands on page 3-12
•
Restrictions when using PC or SP on page 3-12
•
Flexible second operand on page 3-12
•
Shift Operations on page 3-13
•
Address alignment on page 3-17
•
PC-relative expressions on page 3-17
•
Conditional execution on page 3-18
•
Instruction width selection on page 3-21.

ARM DUI 0553A
ID121610

3-11

The Cortex-M4 Instruction Set

3.3.1

Operands
An instruction operand can be an ARM register, a constant, or another instruction-specific
parameter. Instructions act on the operands and often store the result in a destination register.
When there is a destination register in the instruction, it is usually specified before the operands.
Operands in some instructions are flexible in that they can either be a register or a constant. See
Flexible second operand.

3.3.2

Restrictions when using PC or SP
Many instructions have restrictions on whether you can use the Program Counter (PC) or Stack
Pointer (SP) for the operands or destination register. See instruction descriptions for more
information.
Note
Bit[0] of any address you write to the PC with a BX, BLX, LDM, LDR, or POP instruction must be 1
for correct execution, because this bit indicates the required instruction set, and the Cortex-M4
processor only supports Thumb instructions.

3.3.3

Flexible second operand
Many general data processing instructions have a flexible second operand. This is shown as
Operand2 in the descriptions of the syntax of each instruction.
Operand2 can be a:

•
•

Constant
Register with optional shift on page 3-13

Constant
You specify an Operand2 constant in the form:
#constant

where constant can be:
•

any constant that can be produced by shifting an 8-bit value left by any number of bits
within a 32-bit word

•

any constant of the form 0x00XY00XY

•

any constant of the form 0xXY00XY00

•

any constant of the form 0xXYXYXYXY.

Note
In the constants shown above, X and Y are hexadecimal digits.
In addition, in a small number of instructions, constant can take a wider range of values. These
are described in the individual instruction descriptions.
When an Operand2 constant is used with the instructions MOVS, MVNS, ANDS, ORRS, ORNS, EORS, BICS,
TEQ or TST, the carry flag is updated to bit[31] of the constant, if the constant is greater than 255
and can be produced by shifting an 8-bit value. These instructions do not affect the carry flag if
Operand2 is any other constant.

ARM DUI 0553A
ID121610

3-12

The Cortex-M4 Instruction Set

Instruction substitution

Your assembler might be able to produce an equivalent instruction in cases where you specify a
constant that is not permitted. For example, an assembler might assemble the instruction CMP Rd,
#0xFFFFFFFE as the equivalent instruction CMN Rd, #0x2.
Register with optional shift
You specify an Operand2 register in the form:
Rm {, shift}

where:
Rm

The register holding the data for the second operand.

shift

An optional shift to be applied to Rm. It can be one of:
ASR #n

Arithmetic shift right n bits, 1 ≤ n ≤ 32.

LSL #n

Logical shift left n bits, 1 ≤ n ≤ 31.

LSR #n

Logical shift right n bits, 1 ≤ n ≤ 32.

ROR #n

Rotate right n bits, 1 ≤ n ≤ 31.

RRX

Rotate right one bit, with extend.

If omitted, no shift occurs, equivalent to LSL #0.

If you omit the shift, or specify LSL #0, the instruction uses the value in Rm.
If you specify a shift, the shift is applied to the value in Rm, and the resulting 32-bit value is used
by the instruction. However, the contents in the register Rm remains unchanged. Specifying a
register with shift also updates the carry flag when used with certain instructions. For
information on the shift operations and how they affect the carry flag, see Shift Operations.
3.3.4

Shift Operations
Register shift operations move the bits in a register left or right by a specified number of bits,
the shift length. Register shift can be performed:
•

directly by the instructions ASR, LSR, LSL, ROR, and RRX, and the result is written to a
destination register

•

during the calculation of Operand2 by the instructions that specify the second operand as a
register with shift, see Flexible second operand on page 3-12. The result is used by the
instruction.

The permitted shift lengths depend on the shift type and the instruction, see the individual
instruction description or Flexible second operand on page 3-12. If the shift length is 0, no shift
occurs. Register shift operations update the carry flag except when the specified shift length is
0. The following sub-sections describe the various shift operations and how they affect the carry
flag. In these descriptions, Rm is the register containing the value to be shifted, and n is the shift
length.
ASR
Arithmetic shift right by n bits moves the left-hand 32-n bits of the register Rm, to the right by n
places, into the right-hand 32-n bits of the result. And it copies the original bit[31] of the register
into the left-hand n bits of the result. See Figure 3-1 on page 3-14.

ARM DUI 0553A
ID121610

3-13

The Cortex-M4 Instruction Set

You can use the ASR #n operation to divide the value in the register Rm by 2n, with the result being
rounded towards negative-infinity.
When the instruction is ASRS or when ASR #n is used in Operand2 with the instructions MOVS, MVNS,
ANDS, ORRS, ORNS, EORS, BICS, TEQ or TST, the carry flag is updated to the last bit shifted out, bit[n-1],
of the register Rm.

•

Note
If n is 32 or more, then all the bits in the result are set to the value of bit[31] of Rm.

•

If n is 32 or more and the carry flag is updated, it is updated to the value of bit[31] of Rm.

Carry
Flag
31

5 4 3 2 1 0

Figure 3-1 ASR #3

LSR
Logical shift right by n bits moves the left-hand 32-n bits of the register Rm, to the right by n
places, into the right-hand 32-n bits of the result. And it sets the left-hand n bits of the result to
0. See Figure 3-2.
You can use the LSR #n operation to divide the value in the register Rm by 2n, if the value is
regarded as an unsigned integer.
When the instruction is LSRS or when LSR #n is used in Operand2 with the instructions MOVS, MVNS,
ANDS, ORRS, ORNS, EORS, BICS, TEQ or TST, the carry flag is updated to the last bit shifted out, bit[n-1],
of the register Rm.

•

Note
If n is 32 or more, then all the bits in the result are cleared to 0.

•

If n is 33 or more and the carry flag is updated, it is updated to 0.

Carry
Flag

0 0 0
31

5 4 3 2 1 0

Figure 3-2 LSR #3

LSL
Logical shift left by n bits moves the right-hand 32-n bits of the register Rm, to the left by n places,
into the left-hand 32-n bits of the result. And it sets the right-hand n bits of the result to 0. See
Figure 3-3 on page 3-15.

ARM DUI 0553A
ID121610

3-14

The Cortex-M4 Instruction Set

You can use he LSL #n operation to multiply the value in the register Rm by 2n, if the value is
regarded as an unsigned integer or a two’s complement signed integer. Overflow can occur
without warning.
When the instruction is LSLS or when LSL #n, with non-zero n, is used in Operand2 with the
instructions MOVS, MVNS, ANDS, ORRS, ORNS, EORS, BICS, TEQ or TST, the carry flag is updated to the
last bit shifted out, bit[32-n], of the register Rm. These instructions do not affect the carry flag
when used with LSL #0.

•

Note
If n is 32 or more, then all the bits in the result are cleared to 0.

•

If n is 33 or more and the carry flag is updated, it is updated to 0.

0 0 0
31

5 4 3 2 1 0

Carry
Flag

Figure 3-3 LSL #3

ROR
Rotate right by n bits moves the left-hand 32-n bits of the register Rm, to the right by n places, into
the right-hand 32-n bits of the result. And it moves the right-hand n bits of the register into the
left-hand n bits of the result. See Figure 3-4.
When the instruction is RORS or when ROR #n is used in Operand2 with the instructions MOVS, MVNS,
ANDS, ORRS, ORNS, EORS, BICS, TEQ or TST, the carry flag is updated to the last bit rotation, bit[n-1],
of the register Rm.

•

Note
If n is 32, then the value of the result is same as the value in Rm, and if the carry flag is
updated, it is updated to bit[31] of Rm.

•

ROR with shift length, n, more than 32 is the same as ROR with shift length n-32.

Carry
Flag
31

5 4 3 2 1 0

Figure 3-4 ROR #3

RRX
Rotate right with extend moves the bits of the register Rm to the right by one bit. And it copies
the carry flag into bit[31] of the result. See Figure 3-5 on page 3-16.
When the instruction is RRXS or when RRX is used in Operand2 with the instructions MOVS, MVNS,
ANDS, ORRS, ORNS, EORS, BICS, TEQ or TST, the carry flag is updated to bit[0] of the register Rm.
ARM DUI 0553A
ID121610

3-15

The Cortex-M4 Instruction Set

Carry
Flag
31 30

Figure 3-5 RRX

ARM DUI 0553A
ID121610

3-16

The Cortex-M4 Instruction Set

3.3.5

Address alignment
An aligned access is an operation where a word-aligned address is used for a word, dual word,
or multiple word access, or where a halfword-aligned address is used for a halfword access. Byte
accesses are always aligned.
The Cortex-M4 processor supports unaligned access only for the following instructions:
•
LDR, LDRT
•
LDRH, LDRHT
•
LDRSH, LDRSHT
•
STR, STRT
•
STRH, STRHT
All other load and store instructions generate a UsageFault exception if they perform an
unaligned access, and therefore their accesses must be address aligned. For more information
about UsageFaults see Fault handling on page 2-29.
Unaligned accesses are usually slower than aligned accesses. In addition, some memory regions
might not support unaligned accesses. Therefore, ARM recommends that programmers ensure
that accesses are aligned. To trap accidental generation of unaligned accesses, use the
UNALIGN_TRP bit in the Configuration and Control Register, see Configuration and Control
Register on page 4-19.

3.3.6

PC-relative expressions
A PC-relative expression or label is a symbol that represents the address of an instruction or
literal data. It is represented in the instruction as the PC value plus or minus a numeric offset.
The assembler calculates the required offset from the label and the address of the current
instruction. If the offset is too big, the assembler produces an error.

•

ARM DUI 0553A
ID121610

Note
For B, BL, CBNZ, and CBZ instructions, the value of the PC is the address of the current
instruction plus 4 bytes.

•

For all other instructions that use labels, the value of the PC is the address of the current
instruction plus 4 bytes, with bit[1] of the result cleared to 0 to make it word-aligned.

•

Your assembler might permit other syntaxes for PC-relative expressions, such as a label
plus or minus a number, or an expression of the form [PC, #number].

3-17

The Cortex-M4 Instruction Set

3.3.7

Conditional execution
Most data processing instructions can optionally update the condition flags in the Application
Program Status Register (APSR) according to the result of the operation, see Application
Program Status Register on page 2-5. Some instructions update all flags, and some only update
a subset. If a flag is not updated, the original value is preserved. See the instruction descriptions
for the flags they affect.
You can execute an instruction conditionally, based on the condition flags set in another
instruction, either:
•
immediately after the instruction that updated the flags
•
after any number of intervening instructions that have not updated the flags.
Conditional execution is available by using conditional branches or by adding condition code
suffixes to instructions. See Table 3-4 on page 3-19 for a list of the suffixes to add to instructions
to make them conditional instructions. The condition code suffix enables the processor to test a
condition based on the flags. If the condition test of a conditional instruction fails, the
instruction:
•
does not execute
•
does not write any value to its destination register
•
does not affect any of the flags
•
does not generate any exception.
Conditional instructions, except for conditional branches, must be inside an If-Then instruction
block. See IT on page 3-122 for more information and restrictions when using the IT instruction.
Depending on the vendor, the assembler might automatically insert an IT instruction if you have
conditional instructions outside the IT block.
Use the CBZ and CBNZ instructions to compare the value of a register against zero and branch on
the result.
This section describes:
•
The condition flags on page 3-19
•
Condition code suffixes on page 3-19.

ARM DUI 0553A
ID121610

3-18

The Cortex-M4 Instruction Set

The condition flags
The APSR contains the following condition flags:
N
Set to 1 when the result of the operation was negative, cleared to 0 otherwise.
Z
Set to 1 when the result of the operation was zero, cleared to 0 otherwise.
C
Set to 1 when the operation resulted in a carry, cleared to 0 otherwise.
V
Set to 1 when the operation caused overflow, cleared to 0 otherwise.
For more information about the APSR see Program Status Register on page 2-4.
A carry occurs:
•
if the result of an addition is greater than or equal to 232
•
if the result of a subtraction is positive or zero
•
as the result of an inline barrel shifter operation in a move or logical instruction.
Overflow occurs when the sign of the result, in bit[31], does not match the sign of the result had
the operation been performed at infinite precision, for example:
•
if adding two negative values results in a positive value
•
if adding two positive values results in a negative value
•
if subtracting a positive value from a negative value generates a positive value
•
if subtracting a negative value from a positive value generates a negative value.
The Compare operations are identical to subtracting, for CMP, or adding, for CMN, except that the
result is discarded. See the instruction descriptions for more information.
Note
Most instructions update the status flags only if the S suffix is specified. See the instruction
descriptions for more information.

Condition code suffixes
The instructions that can be conditional have an optional condition code, shown in syntax
descriptions as {cond}. Conditional execution requires a preceding IT instruction. An instruction
with a condition code is only executed if the condition code flags in the APSR meet the specified
condition. Table 3-4 shows the condition codes to use.
You can use conditional execution with the IT instruction to reduce the number of branch
instructions in code.
Table 3-4 also shows the relationship between condition code suffixes and the N, Z, C, and V
flags.
Table 3-4 Condition code suffixes

ARM DUI 0553A
ID121610

Suffix

Flags

Meaning

Z=1

Equal

Z=0

Not equal

CS or HS

C=1

Higher or same, unsigned

CC or LO

C=0

Lower, unsigned

N=1

Negative

N=0

Positive or zero

3-19

The Cortex-M4 Instruction Set

Table 3-4 Condition code suffixes (continued)
Suffix

Flags

Meaning

V=1

Overflow

V=0

No overflow

C = 1 and Z = 0

Higher, unsigned

C = 0 or Z = 1

Lower or same, unsigned

N=V

Greater than or equal, signed

N != V

Less than, signed

Z = 0 and N = V

Greater than, signed

Z = 1 and N != V

Less than or equal, signed

Can have any value

Always. This is the default when no suffix is specified.

Example 3-1 shows the use of a conditional instruction to find the absolute value of a number.
R0 = abs(R1).
Example 3-1 Absolute value

MOVS
IT
RSBMI

R0, R1
MI
R0, R0, #0

; R0 = R1, setting flags
; skipping next instruction if value 0 or positive
; If negative, R0 = -R0

Example 3-2 shows the use of conditional instructions to update the value of R4 if the signed
values R0 is greater than R1 and R2 is greater than R3.
Example 3-2 Compare and update value

CMP
ITT
CMPGT
MOVGT

ARM DUI 0553A
ID121610

R0, R1
GT
R2, R3
R4, R5

;
;
;
;

Compare R0 and R1, setting flags
Skip next two instructions unless GT condition holds
If 'greater than', compare R2 and R3, setting flags
If still 'greater than', do R4 = R5

3-20

The Cortex-M4 Instruction Set

3.3.8

Instruction width selection
There are many instructions that can generate either a 16-bit encoding or a 32-bit encoding
depending on the operands and destination register specified. For some of these instructions,
you can force a specific instruction size by using an instruction width suffix. The .W suffix forces
a 32-bit instruction encoding. The .N suffix forces a 16-bit instruction encoding.
If you specify an instruction width suffix and the assembler cannot generate an instruction
encoding of the requested width, it generates an error.
Note
In some cases it might be necessary to specify the .W suffix, for example if the operand is the
label of an instruction or literal data, as in the case of branch instructions. This is because the
assembler might not automatically generate the right size encoding.
To use an instruction width suffix, place it immediately after the instruction mnemonic and
condition code, if any. Example 3-3 shows instructions with the instruction width suffix.
Example 3-3 Instruction width selection

BCS.W

label

; creates a 32-bit instruction even for a short branch

ADDS.W R0, R0, R1 ; creates a 32-bit instruction even though the same
; operation can be done by a 16-bit instruction

ARM DUI 0553A
ID121610

3-21

The Cortex-M4 Instruction Set

3.4

Memory access instructions
Table 3-5 shows the memory access instructions:
Table 3-5 Memory access instructions

ARM DUI 0553A
ID121610

Mnemonic

Brief description

See

ADR

Generate PC-relative address

ADR on page 3-23

CLREX

Clear Exclusive

CLREX on page 3-38

LDM{mode}

Load Multiple registers

LDM and STM on page 3-32

LDR{type}

Load Register using immediate offset

LDR and STR, immediate offset on page 3-24

LDR{type}

Load Register using register offset

LDR and STR, register offset on page 3-27

LDR{type}T

Load Register with unprivileged access

LDR and STR, unprivileged on page 3-29

LDR

Load Register using PC-relative address

LDR, PC-relative on page 3-30

LDREX{type}

Load Register Exclusive

LDREX and STREX on page 3-36

POP

Pop registers from stack

PUSH and POP on page 3-34

PUSH

Push registers onto stack

PUSH and POP on page 3-34

STM{mode}

Store Multiple registers

LDM and STM on page 3-32

STR{type}

Store Register using immediate offset

LDR and STR, immediate offset on page 3-24

STR{type}

Store Register using register offset

LDR and STR, register offset on page 3-27

STR{type}T

Store Register with unprivileged access

LDR and STR, unprivileged on page 3-29

STREX{type}

Store Register Exclusive

LDREX and STREX on page 3-36

3-22

The Cortex-M4 Instruction Set

3.4.1

ADR
Generate PC-relative address.
Syntax
ADR{cond} Rd, label

where:
cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

label

Is a PC-relative expression. See PC-relative expressions on page 3-17.

Operation
ADR generates an address by adding an immediate value to the PC, and writes the result to the

destination register.
ADR provides the means by which position-independent code can be generated, because the

address is PC-relative.
If you use ADR to generate a target address for a BX or BLX instruction, you must ensure that bit[0]
of the address you generate is set to 1 for correct execution.
Values of label must be within the range of −4095 to +4095 from the address in the PC.
Note
You might have to use the .W suffix to get the maximum offset range or to generate addresses
that are not word-aligned. See Instruction width selection on page 3-21.

Restrictions
Rd must not be SP and must not be PC.

Condition flags
This instruction does not change the flags.
Examples
ADR

ARM DUI 0553A
ID121610

R1, TextMessage

; Write address value of a location labelled as
; TextMessage to R1.

3-23

The Cortex-M4 Instruction Set

3.4.2

LDR and STR, immediate offset
Load and Store with immediate offset, pre-indexed immediate offset, or post-indexed
immediate offset.
Syntax
op{type}{cond} Rt, [Rn {, #offset}]

; immediate offset

op{type}{cond} Rt, [Rn, #offset]!

; pre-indexed

op{type}{cond} Rt, [Rn], #offset

; post-indexed

opD{cond} Rt, Rt2, [Rn {, #offset}]

; immediate offset, two words

opD{cond} Rt, Rt2, [Rn, #offset]!

; pre-indexed, two words

opD{cond} Rt, Rt2, [Rn], #offset

; post-indexed, two words

where:
op

Is one of:
LDR
STR

type

Is one of:
B
SB
H
SH
-

ARM DUI 0553A
ID121610

Load Register.
Store Register.
unsigned byte, zero extend to 32 bits on loads.
signed byte, sign extend to 32 bits (LDR only).
unsigned halfword, zero extend to 32 bits on loads.
signed halfword, sign extend to 32 bits (LDR only).
omit, for word.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the register to load or store.

Specifies the register on which the memory address is based.

offset

Specifies an offset from Rn. If offset is omitted, the address is the contents of Rn.

Rt2

Specifies the additional register to load or store for two-word operations.

3-24

The Cortex-M4 Instruction Set

Operation
LDR instructions load one or two registers with a value from memory.
STR instructions store one or two register values to memory.

Load and store instructions with immediate offset can use the following addressing modes:
Offset addressing
The offset value is added to or subtracted from the address obtained from the
register Rn. The result is used as the address for the memory access. The register
Rn is unaltered. The assembly language syntax for this mode is:
[Rn, #offset]

Pre-indexed addressing
The offset value is added to or subtracted from the address obtained from the
register Rn. The result is used as the address for the memory access and written
back into the register Rn. The assembly language syntax for this mode is:
[Rn, #offset]!

Post-indexed addressing
The address obtained from the register Rn is used as the address for the memory
access. The offset value is added to or subtracted from the address, and written
back into the register Rn. The assembly language syntax for this mode is:
[Rn], #offset

The value to load or store can be a byte, halfword, word, or two words. Bytes and halfwords can
either be signed or unsigned. See Address alignment on page 3-17.
Table 3-6 shows the ranges of offset for immediate, pre-indexed and post-indexed forms.
Table 3-6 Offset ranges

ARM DUI 0553A
ID121610

Instruction type

Immediate offset

Pre-indexed

Post-indexed

Word, halfword, signed
halfword, byte, or signed byte

−255 to 4095

−255 to 255

Two words

multiple of 4 in the
range −1020 to 1020

3-25

The Cortex-M4 Instruction Set

Restrictions
For load instructions:
•
Rt can be SP or PC for word loads only
•
Rt must be different from Rt2 for two-word loads
•
Rn must be different from Rt and Rt2 in the pre-indexed or post-indexed forms.
When Rt is PC in a word load instruction:
•
bit[0] of the loaded value must be 1 for correct execution
•
a branch occurs to the address created by changing bit[0] of the loaded value to 0
•
if the instruction is conditional, it must be the last instruction in the IT block.
For store instructions:
•
Rt can be SP for word stores only
•
Rt must not be PC
•
Rn must not be PC
•
Rn must be different from Rt and Rt2 in the pre-indexed or post-indexed forms.
Condition flags
These instructions do not change the flags.
Examples

ARM DUI 0553A
ID121610

LDR
LDRNE

R8, [R10]
R2, [R5, #960]!

STR

R2, [R9,#const-struc]

STRH

R3, [R4], #4

LDRD

R8, R9, [R3, #0x20]

STRD

R0, R1, [R8], #-16

;
;
;
;
;
;
;
;
;
;
;
;
;
;

Loads R8 from the address in R10.
Loads (conditionally) R2 from a word
960 bytes above the address in R5, and
increments R5 by 960
const-struc is an expression evaluating
to a constant in the range 0-4095.
Store R3 as halfword data into address in
R4, then increment R4 by 4
Load R8 from a word 32 bytes above the
address in R3, and load R9 from a word 36
bytes above the address in R3
Store R0 to address in R8, and store R1 to
a word 4 bytes above the address in R8,
and then decrement R8 by 16.

3-26

The Cortex-M4 Instruction Set

3.4.3

LDR and STR, register offset
Load and Store with register offset.
Syntax
op{type}{cond} Rt, [Rn, Rm {, LSL #n}]

where:
Is one of:

LDR
STR

Load Register.
Store Register.

Is one of:

type

B
SB
H
SH
-

unsigned byte, zero extend to 32 bits on loads.
signed byte, sign extend to 32 bits (LDR only).
unsigned halfword, zero extend to 32 bits on loads.
signed halfword, sign extend to 32 bits (LDR only).
omit, for word.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the register to load or store.

Specifies the register on which the memory address is based.

Specifies the register containing a value to be used as the offset.

LSL #n

Is an optional shift, with n in the range 0 to 3.

Operation
LDR instructions load a register with a value from memory.
STR instructions store a register value into memory.

The memory address to load from or store to is at an offset from the register Rn. The offset is
specified by the register Rm and can be shifted left by up to 3 bits using LSL.
The value to load or store can be a byte, halfword, or word. For load instructions, bytes and
halfwords can either be signed or unsigned. See Address alignment on page 3-17.
Restrictions
In these instructions:
•
Rn must not be PC
•
Rm must not be SP and must not be PC
•
Rt can be SP only for word loads and word stores
•
Rt can be PC only for word loads.
When Rt is PC in a word load instruction:

ARM DUI 0553A
ID121610

•

bit[0] of the loaded value must be 1 for correct execution, and a branch occurs to this
halfword-aligned address

•

if the instruction is conditional, it must be the last instruction in the IT block.

3-27

The Cortex-M4 Instruction Set

Condition flags
These instructions do not change the flags.
Examples
STR
LDRSB

STR

ARM DUI 0553A
ID121610

R0, [R5, R1]

;
;
R0, [R5, R1, LSL #1] ;
;
;
R0, [R1, R2, LSL #2] ;
;

Store value of R0 into an address equal to
sum of R5 and R1
Read byte value from an address equal to
sum of R5 and two times R1, sign extended it
to a word value and put it in R0
Stores R0 to an address equal to sum of R1
and four times R2.

3-28

The Cortex-M4 Instruction Set

3.4.4

LDR and STR, unprivileged
Load and Store with unprivileged access.
Syntax
op{type}T{cond} Rt, [Rn {, #offset}]

; immediate offset

where:
Is one of:

LDR
STR

Load Register.
Store Register.

Is one of:

type

B
SB
H
SH
-

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the register to load or store.

Specifies the register on which the memory address is based.

offset

Specifies an offset from Rn and can be 0 to 255. If offset is omitted, the address
is the value in Rn.

Operation
These load and store instructions perform the same function as the memory access instructions
with immediate offset, see LDR and STR, immediate offset on page 3-24. The difference is that
these instructions have only unprivileged access even when used in privileged software.
When used in unprivileged software, these instructions behave in exactly the same way as
normal memory access instructions with immediate offset.
Restrictions
In these instructions:
•
Rn must not be PC
•
Rt must not be SP and must not be PC.
Condition flags
These instructions do not change the flags.
Examples

ARM DUI 0553A
ID121610

STRBTEQ

R4, [R7]

LDRHT

R2, [R2, #8]

;
;
;
;

Conditionally store least significant byte in
R4 to an address in R7, with unprivileged access
Load halfword value from an address equal to
sum of R2 and 8 into R2, with unprivileged access.

3-29

The Cortex-M4 Instruction Set

3.4.5

LDR, PC-relative
Load register from memory.
Syntax
LDR{type}{cond} Rt, label
LDRD{cond} Rt, Rt2, label

; Load two words

where:
type

Is one of:
B
SB
H
SH
-

unsigned byte, zero extend to 32 bits.
signed byte, sign extend to 32 bits.
unsigned halfword, zero extend to 32 bits.
signed halfword, sign extend to 32 bits.
omit, for word.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the register to load or store.

Rt2

Specifies the second register to load or store.

label

Is a PC-relative expression. See PC-relative expressions on page 3-17.

Operation
LDR loads a register with a value from a PC-relative memory address. The memory address is

specified by a label or by an offset from the PC.
The value to load or store can be a byte, halfword, or word. For load instructions, bytes and
halfwords can either be signed or unsigned. See Address alignment on page 3-17.
label must be within a limited range of the current instruction. Table 3-7 shows the possible
offsets between label and the PC.

Table 3-7 Offset ranges
Instruction type

Offset range

Word, halfword, signed halfword, byte, signed byte

−4095 to 4095

Two words

−1020 to 1020

Note
You might have to use the .W suffix to get the maximum offset range. See Instruction width
selection on page 3-21.

Restrictions
In these instructions:
•
Rt can be SP or PC only for word loads
•
Rt2 must not be SP and must not be PC
•
Rt must be different from Rt2.

ARM DUI 0553A
ID121610

3-30

The Cortex-M4 Instruction Set

When Rt is PC in a word load instruction:
•
bit[0] of the loaded value must be 1 for correct execution, and a branch occurs to this
halfword-aligned address
•
if the instruction is conditional, it must be the last instruction in the IT block.
Condition flags
These instructions do not change the flags.
Examples

ARM DUI 0553A
ID121610

LDR

R0, LookUpTable

LDRSB

R7, localdata

;
;
;
;
;

Load R0 with a word of data from an address
labelled as LookUpTable
Load a byte value from an address labelled
as localdata, sign extend it to a word
value, and put it in R7.

3-31

The Cortex-M4 Instruction Set

3.4.6

LDM and STM
Load and Store Multiple registers.
Syntax
op{addr_mode}{cond} Rn{!}, reglist

where:
op

Is one of:
LDM
STM

Load Multiple registers.
Store Multiple registers.

addr_mode

Is any one of the following:
IA
Increment address After each access. This is the default.
DB
Decrement address Before each access.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the register on which the memory addresses are based.

Is an optional writeback suffix. If ! is present the final address, that is loaded from
or stored to, is written back into Rn.

reglist

Is a list of one or more registers to be loaded or stored, enclosed in braces. It can
contain register ranges. It must be comma separated if it contains more than one
register or register range, see Examples on page 3-33.

LDM and LDMFD are synonyms for LDMIA. LDMFD refers to its use for popping data from Full
Descending stacks.
LDMEA is a synonym for LDMDB, and refers to its use for popping data from Empty Ascending

stacks.
STM and STMEA are synonyms for STMIA. STMEA refers to its use for pushing data onto Empty
Ascending stacks.
STMFD is s synonym for STMDB, and refers to its use for pushing data onto Full Descending stacks

Operation
LDM instructions load the registers in reglist with word values from memory addresses based on
Rn.
STM instructions store the word values in the registers in reglist to memory addresses based on
Rn.

For LDM, LDMIA, LDMFD, STM, STMIA, and STMEA the memory addresses used for the accesses are at
4-byte intervals ranging from Rn to Rn + 4 * (n-1), where n is the number of registers in reglist.
The accesses happens in order of increasing register numbers, with the lowest numbered register
using the lowest memory address and the highest number register using the highest memory
address. If the writeback suffix is specified, the value of Rn + 4 * (n-1) is written back to Rn.
For LDMDB, LDMEA, STMDB, and STMFD the memory addresses used for the accesses are at 4-byte
intervals ranging from Rn to Rn - 4 * (n-1), where n is the number of registers in reglist. The
accesses happen in order of decreasing register numbers, with the highest numbered register
using the highest memory address and the lowest number register using the lowest memory
address. If the writeback suffix is specified, the value of Rn - 4 * (n-1) is written back to Rn.

ARM DUI 0553A
ID121610

3-32

The Cortex-M4 Instruction Set

The PUSH and POP instructions can be expressed in this form. See PUSH and POP on page 3-34
for details.
Restrictions
In these instructions:
•
Rn must not be PC
•
reglist must not contain SP
•
in any STM instruction, reglist must not contain PC
•
in any LDM instruction, reglist must not contain PC if it contains LR
•
reglist must not contain Rn if you specify the writeback suffix.
When PC is in reglist in an LDM instruction:
•

bit[0] of the value loaded to the PC must be 1 for correct execution, and a branch occurs
to this halfword-aligned address

•

if the instruction is conditional, it must be the last instruction in the IT block.

Condition flags
These instructions do not change the flags.
Examples
LDM
STMDB

R8,{R0,R2,R9}
; LDMIA is a synonym for LDM
R1!,{R3-R6,R11,R12}

Incorrect examples
STM
LDM

ARM DUI 0553A
ID121610

R5!,{R5,R4,R9} ; Value stored for R5 is unpredictable
R2, {}
; There must be at least one register in the list.

3-33

The Cortex-M4 Instruction Set

3.4.7

PUSH and POP
Push registers onto, and pop registers off a full-descending stack.
Syntax
PUSH{cond} reglist
POP{cond} reglist

where:
cond

Is an optional condition code, see Conditional execution on page 3-18.

reglist

Is a non-empty list of registers, enclosed in braces. It can contain register ranges.
It must be comma separated if it contains more than one register or register range.

PUSH and POP are synonyms for STMDB and LDM (or LDMIA) with the memory addresses for the access
based on SP, and with the final address for the access written back to the SP. PUSH and POP are

the preferred mnemonics in these cases.
Operation
PUSH stores registers on the stack, with the lowest numbered register using the lowest memory

address and the highest numbered register using the highest memory address.
POP loads registers from the stack, with the lowest numbered register using the lowest memory

address and the highest numbered register using the highest memory address.
PUSH uses the value in the SP register minus four as the highest memory address, POP uses the
value in the SP register as the lowest memory address, implementing a full-descending stack.
On completion, PUSH updates the SP register to point to the location of the lowest store value,
POP updates the SP register to point to the location above the highest location loaded.

If a POP instruction includes PC in its reglist, a branch to this location is performed when the POP
instruction has completed. Bit[0] of the value read for the PC is used to update the APSR T-bit.
This bit must be 1 to ensure correct operation.
See LDM and STM on page 3-32 for more information.
Restrictions
In these instructions:
•
reglist must not contain SP
•
for the PUSH instruction, reglist must not contain PC
•
for the POP instruction, reglist must not contain PC if it contains LR.
When PC is in reglist in a POP instruction:
•

bit[0] of the value loaded to the PC must be 1 for correct execution, and a branch occurs
to this halfword-aligned address

•

if the instruction is conditional, it must be the last instruction in the IT block.

Condition flags
These instructions do not change the flags.

ARM DUI 0553A
ID121610

3-34

The Cortex-M4 Instruction Set

Examples
PUSH {R0,R4-R7} ; Push R0,R4,R5,R6,R7 onto the stack
PUSH {R2,LR}
; Push R2 and the link-register onto the stack
POP {R0,R6,PC} ; Pop r0,r6 and PC from the stack, then branch to the new PC.

ARM DUI 0553A
ID121610

3-35

The Cortex-M4 Instruction Set

3.4.8

LDREX and STREX
Load and Store Register Exclusive.
Syntax
LDREX{cond} Rt, [Rn {, #offset}]
STREX{cond} Rd, Rt, [Rn {, #offset}]
LDREXB{cond} Rt, [Rn]
STREXB{cond} Rd, Rt, [Rn]
LDREXH{cond} Rt, [Rn]
STREXH{cond} Rd, Rt, [Rn]

where:
cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register for the returned status.

Specifies the register to load or store.

Specifies the register on which the memory address is based.

offset

Is an optional offset applied to the value in Rn. If offset is omitted, the address is
the value in Rn.

Operation
LDREX, LDREXB, and LDREXH load a word, byte, and halfword respectively from a memory address.
STREX, STREXB, and STREXH attempt to store a word, byte, and halfword respectively to a memory
address. The address used in any Store-Exclusive instruction must be the same as the address in
the most recently executed Load-exclusive instruction. The value stored by the Store-Exclusive
instruction must also have the same data size as the value loaded by the preceding
Load-exclusive instruction. This means software must always use a Load-exclusive instruction
and a matching Store-Exclusive instruction to perform a synchronization operation, see
Synchronization primitives on page 2-18.

If an Store-Exclusive instruction performs the store, it writes 0 to its destination register. If it
does not perform the store, it writes 1 to its destination register. If the Store-Exclusive
instruction writes 0 to the destination register, it is guaranteed that no other process in the system
has accessed the memory location between the Load-exclusive and Store-Exclusive
instructions.
For reasons of performance, keep the number of instructions between corresponding
Load-Exclusive and Store-Exclusive instruction to a minimum.
Note
The result of executing a Store-Exclusive instruction to an address that is different from that
used in the preceding Load-Exclusive instruction is unpredictable.

ARM DUI 0553A
ID121610

3-36

The Cortex-M4 Instruction Set

Restrictions
In these instructions:
•
do not use PC
•
do not use SP for Rd and Rt
•
for STREX, Rd must be different from both Rt and Rn
•
the value of offset must be a multiple of four in the range 0-1020.
Condition flags
These instructions do not change the flags.
Examples
MOV

R1, #0x1

; Initialize the ‘lock taken’ value

LDREX
CMP
ITT
STREXEQ
CMPEQ
BNE
....

R0,
R0,
EQ
R0,
R0,
try

;
;
;
;
;
;
;

try

ARM DUI 0553A
ID121610

[LockAddr]
#0
R1, [LockAddr]
#0

Load the lock value
Is the lock free?
IT instruction for STREXEQ and CMPEQ
Try and claim the lock
Did this succeed?
No – try again
Yes – we have the lock.

3-37

The Cortex-M4 Instruction Set

3.4.9

CLREX
Clear Exclusive.
Syntax
CLREX{cond}

where:
cond

Is an optional condition code, see Conditional execution on page 3-18.

Operation
Use CLREX to make the next STREX, STREXB, or STREXH instruction write 1 to its destination register
and fail to perform the store. It is useful in exception handler code to force the failure of the store
exclusive if the exception occurs between a load exclusive instruction and the matching store
exclusive instruction in a synchronization operation.
See Synchronization primitives on page 2-18 for more information.
Condition flags
This instruction does not change the flags.
Examples
CLREX

ARM DUI 0553A
ID121610

3-38

The Cortex-M4 Instruction Set

3.5

General data processing instructions
Table 3-8 shows the data processing instructions:
Table 3-8 Data processing instructions
Mnemonic

Brief description

See

ADC

Add with Carry

ADD, ADC, SUB, SBC, and RSB on page 3-41

ADD

Add

ADD, ADC, SUB, SBC, and RSB on page 3-41

ADDW

Add

ADD, ADC, SUB, SBC, and RSB on page 3-41

AND

Logical AND

AND, ORR, EOR, BIC, and ORN on page 3-44

ASR

Arithmetic Shift Right

ASR, LSL, LSR, ROR, and RRX on page 3-46

BIC

Bit Clear

AND, ORR, EOR, BIC, and ORN on page 3-44

CLZ

Count leading zeros

CLZ on page 3-48

CMN

Compare Negative

CMP and CMN on page 3-49

CMP

Compare

CMP and CMN on page 3-49

EOR

Exclusive OR

AND, ORR, EOR, BIC, and ORN on page 3-44

LSL

Logical Shift Left

ASR, LSL, LSR, ROR, and RRX on page 3-46

LSR

Logical Shift Right

ASR, LSL, LSR, ROR, and RRX on page 3-46

MOV

Move

MOV and MVN on page 3-50

MOVT

Move Top

MOVT on page 3-52

MOVW

Move 16-bit constant

MOV and MVN on page 3-50

MVN

Move NOT

MOV and MVN on page 3-50

ORN

Logical OR NOT

AND, ORR, EOR, BIC, and ORN on page 3-44

ORR

Logical OR

AND, ORR, EOR, BIC, and ORN on page 3-44

RBIT

Reverse Bits

REV, REV16, REVSH, and RBIT on page 3-53

REV

Reverse byte order in a word

REV, REV16, REVSH, and RBIT on page 3-53

REV16

Reverse byte order in each halfword

REV, REV16, REVSH, and RBIT on page 3-53

REVSH

Reverse byte order in bottom halfword and sign extend

REV, REV16, REVSH, and RBIT on page 3-53

ROR

Rotate Right

ASR, LSL, LSR, ROR, and RRX on page 3-46

RRX

Rotate Right with Extend

ASR, LSL, LSR, ROR, and RRX on page 3-46

RSB

Reverse Subtract

ADD, ADC, SUB, SBC, and RSB on page 3-41

SADD16

Signed Add 16

SADD16 and SADD8 on page 3-54

SADD8

Signed Add 8

SADD16 and SADD8 on page 3-54

SASX

Signed Add and Subtract with Exchange

SASX and SSAX on page 3-60

SSAX

Signed Subtract and Add with Exchange

SASX and SSAX on page 3-60

SBC

Subtract with Carry

ADD, ADC, SUB, SBC, and RSB on page 3-41

ARM DUI 0553A
ID121610

3-39

The Cortex-M4 Instruction Set

Table 3-8 Data processing instructions (continued)
Mnemonic

Brief description

See

SHADD16

Signed Halving Add 16

SHADD16 and SHADD8 on page 3-55

SHADD8

Signed Halving Add 8

SHADD16 and SHADD8 on page 3-55

SHASX

Signed Halving Add and Subtract with Exchange

SHASX and SHSAX on page 3-56

SHSAX

Signed Halving Subtract and Add with Exchange

SHASX and SHSAX on page 3-56

SHSUB16

Signed Halving Subtract 16

SHSUB16 and SHSUB8 on page 3-58

SHSUB8

Signed Halving Subtract 8

SHSUB16 and SHSUB8 on page 3-58

SSUB16

Signed Subtract 16

SSUB16 and SSUB8 on page 3-59

SSUB8

Signed Subtract 8

SSUB16 and SSUB8 on page 3-59

SUB

Subtract

ADD, ADC, SUB, SBC, and RSB on page 3-41

SUBW

Subtract

ADD, ADC, SUB, SBC, and RSB on page 3-41

TEQ

Test Equivalence

TST and TEQ on page 3-62

TST

Test

TST and TEQ on page 3-62

UADD16

Unsigned Add 16

UADD16 and UADD8 on page 3-63

UADD8

Unsigned Add 8

UADD16 and UADD8 on page 3-63

UASX

Unsigned Add and Subtract with Exchange

UASX and USAX on page 3-64

USAX

Unsigned Subtract and Add with Exchange

UASX and USAX on page 3-64

UHADD16

Unsigned Halving Add 16

UHADD16 and UHADD8 on page 3-66

UHADD8

Unsigned Halving Add 8

UHADD16 and UHADD8 on page 3-66

UHASX

Unsigned Halving Add and Subtract with Exchange

UHASX and UHSAX on page 3-67

UHSAX

Unsigned Halving Subtract and Add with Exchange

UHASX and UHSAX on page 3-67

UHSUB16

Unsigned Halving Subtract 16

UHSUB16 and UHSUB8 on page 3-69

UHSUB8

Unsigned Halving Subtract 8

UHSUB16 and UHSUB8 on page 3-69

USAD8

Unsigned Sum of Absolute Differences

USAD8 on page 3-71

USADA8

Unsigned Sum of Absolute Differences and Accumulate

USADA8 on page 3-72

USUB16

Unsigned Subtract 16

USUB16 and USUB8 on page 3-73

USUB8

Unsigned Subtract 8

USUB16 and USUB8 on page 3-73

ARM DUI 0553A
ID121610

3-40

The Cortex-M4 Instruction Set

3.5.1

ADD, ADC, SUB, SBC, and RSB
Add, Add with carry, Subtract, Subtract with carry, and Reverse Subtract.
Syntax
op{S}{cond} {Rd,} Rn, Operand2
op{cond} {Rd,} Rn, #imm12

; ADD and SUB only

where:
op

Is one of:
ADD
ADC
SUB
SBC
RSB

Add.
Add with Carry.
Subtract.
Subtract with Carry.
Reverse Subtract.

Is an optional suffix. If S is specified, the condition code flags are updated on the
result of the operation, see Conditional execution on page 3-18.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register. If Rd is omitted, the destination register is Rn.

Specifies the register holding the first operand.

Operand2

Is a flexible second operand. See Flexible second operand on page 3-12 for
details of the options.

imm12

Is any value in the range 0-4095.

Operation
The ADD instruction adds the value of Operand2 or imm12 to the value in Rn.
The ADC instruction adds the values in Rn and Operand2, together with the carry flag.
The SUB instruction subtracts the value of Operand2 or imm12 from the value in Rn.
The SBC instruction subtracts the value of Operand2 from the value in Rn. If the carry flag is clear,
the result is reduced by one.
The RSB instruction subtracts the value in Rn from the value of Operand2. This is useful because
of the wide range of options for Operand2.
Use ADC and SBC to synthesize multiword arithmetic, see Multiword arithmetic examples on
page 3-42.
See also ADR on page 3-23.
Note
ADDW is equivalent to the ADD syntax that uses the imm12 operand. SUBW is equivalent to the SUB
syntax that uses the imm12 operand.

ARM DUI 0553A
ID121610

3-41

The Cortex-M4 Instruction Set

Restrictions
In these instructions:
•
•

Operand2 must not be SP and must not be PC
Rd can be SP only in ADD and SUB, and only with the additional restrictions:

—
—

Rn must also be SP

any shift in Operand2 must be limited to a maximum of 3 bits using LSL

•

Rn can be SP only in ADD and SUB

•

Rd can be PC only in the ADD{cond} PC, PC, Rm instruction where:

—
—
—
•

you must not specify the S suffix
Rm must not be PC and must not be SP
if the instruction is conditional, it must be the last instruction in the IT block

with the exception of the ADD{cond} PC, PC, Rm instruction, Rn can be PC only in ADD and
SUB, and only with the additional restrictions:
— you must not specify the S suffix
— the second operand must be a constant in the range 0 to 4095.
—

—

Note
When using the PC for an addition or a subtraction, bits[1:0] of the PC are rounded
to 0b00 before performing the calculation, making the base address for the
calculation word-aligned.
If you want to generate the address of an instruction, you have to adjust the constant
based on the value of the PC. ARM recommends that you use the ADR instruction
instead of ADD or SUB with Rn equal to the PC, because your assembler automatically
calculates the correct constant for the ADR instruction.

When Rd is PC in the ADD{cond} PC, PC, Rm instruction:
•
bit[0] of the value written to the PC is ignored
•
a branch occurs to the address created by forcing bit[0] of that value to 0.
Condition flags
If S is specified, these instructions update the N, Z, C and V flags according to the result.
Examples
ADD
SUBS
RSB
ADCHI

R2, R1, R3
R8, R6, #240
R4, R4, #1280
R11, R0, R3

;
;
;
;

Sets the flags on the result
Subtracts contents of R4 from 1280
Only executed if C flag set and Z
flag clear.

Multiword arithmetic examples
Example 3-4 on page 3-43 shows two instructions that add a 64-bit integer contained in R2 and
R3 to another 64-bit integer contained in R0 and R1, and place the result in R4 and R5.

ARM DUI 0553A
ID121610

3-42

The Cortex-M4 Instruction Set

Example 3-4 64-bit addition

ADDS
ADC

R4, R0, R2
R5, R1, R3

; add the least significant words
; add the most significant words with carry

Multiword values do not have to use consecutive registers. Example 3-5 shows instructions that
subtract a 96-bit integer contained in R9, R1, and R11 from another contained in R6, R2, and
R8. The example stores the result in R6, R9, and R2.
Example 3-5 96-bit subtraction

SUBS
SBCS
SBC

ARM DUI 0553A
ID121610

R6, R6, R9
R9, R2, R1
R2, R8, R11

; subtract the least significant words
; subtract the middle words with carry
; subtract the most significant words with carry

3-43

The Cortex-M4 Instruction Set

3.5.2

AND, ORR, EOR, BIC, and ORN
Logical AND, OR, Exclusive OR, Bit Clear, and OR NOT.
Syntax
op{S}{cond} {Rd,} Rn, Operand2

where:
Is one of:

AND
ORR
EOR
BIC
ORN

logical AND.
logical OR, or bit set.
logical Exclusive OR.
logical AND NOT, or bit clear.
logical OR NOT.

Is an optional suffix. If S is specified, the condition code flags are updated on the
result of the operation, see Conditional execution on page 3-18.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Specifies the register holding the first operand.

Operand2

Is a flexible second operand. See Flexible second operand on page 3-12 for
details of the options.

Operation
The AND, EOR, and ORR instructions perform bitwise AND, Exclusive OR, and OR operations on
the values in Rn and Operand2.
The BIC instruction performs an AND operation on the bits in Rn with the complements of the
corresponding bits in the value of Operand2.
The ORN instruction performs an OR operation on the bits in Rn with the complements of the
corresponding bits in the value of Operand2.
Restrictions
Do not use SP and do not use PC.
Condition flags
If S is specified, these instructions:

ARM DUI 0553A
ID121610

•

update the N and Z flags according to the result

•

can update the C flag during the calculation of Operand2, see Flexible second operand on
page 3-12

•

do not affect the V flag.

3-44

The Cortex-M4 Instruction Set

Examples
AND
ORREQ
ANDS
EORS
BIC
ORN
ORNS

ARM DUI 0553A
ID121610

R9,
R2,
R9,
R7,
R0,
R7,
R7,

R2, #0xFF00
R0, R5
R8, #0x19
R11, #0x18181818
R1, #0xab
R11, R14, ROR #4
R11, R14, ASR #32

3-45

The Cortex-M4 Instruction Set

3.5.3

ASR, LSL, LSR, ROR, and RRX
Arithmetic Shift Right, Logical Shift Left, Logical Shift Right, Rotate Right, and Rotate Right
with Extend.
Syntax
op{S}{cond} Rd, Rm, Rs
op{S}{cond} Rd, Rm, #n
RRX{S}{cond} Rd, Rm

where:
op

Is one of:
ASR

Arithmetic Shift Right.

LSL

Logical Shift Left.

LSR

Logical Shift Right.

ROR

Rotate Right.

Is an optional suffix. If S is specified, the condition code flags are updated on the
result of the operation, see Conditional execution on page 3-18.

Specifies the destination register.

Specifies the register holding the value to be shifted.

Specifies the register holding the shift length to apply to the value in Rm. Only the
least significant byte is used and can be in the range 0 to 255.

Specifies the shift length. The range of shift length depends on the instruction:
ASR
shift length from 1 to 32
LSL
shift length from 0 to 31
LSR
shift length from 1 to 32
ROR
shift length from 1 to 31.

Note
MOVS Rd, Rm is the preferred syntax for LSLS Rd, Rm, #0.

ARM DUI 0553A
ID121610

3-46

The Cortex-M4 Instruction Set

Operation
ASR, LSL, LSR, and ROR move the bits in the register Rm to the left or right by the number of places
specified by constant n or register Rs.
RRX moves the bits in register Rm to the right by 1.

In all these instructions, the result is written to Rd, but the value in register Rm remains
unchanged. For details on what result is generated by the different instructions, see Shift
Operations on page 3-13.
Restrictions
Do not use SP and do not use PC.
Condition flags
If S is specified:
•

these instructions update the N and Z flags according to the result

•

the C flag is updated to the last bit shifted out, except when the shift length is 0, see Shift
Operations on page 3-13.

Examples
ASR
LSLS
LSR
ROR
RRX

ARM DUI 0553A
ID121610

R7,
R1,
R4,
R4,
R4,

R8,
R2,
R5,
R5,
R5

#9
#3
#6
R6

;
;
;
;
;

Arithmetic shift right by 9 bits
Logical shift left by 3 bits with flag update
Logical shift right by 6 bits
Rotate right by the value in the bottom byte of R6
Rotate right with extend.

3-47

The Cortex-M4 Instruction Set

3.5.4

CLZ
Count Leading Zeros.
Syntax
CLZ{cond} Rd, Rm

where:
Is an optional condition code, see Conditional execution on page 3-18.
Specifies the destination register.
Specifies the operand register.

cond
Rd
Rm

Operation
The CLZ instruction counts the number of leading zeros in the value in Rm and returns the result
in Rd. The result value is 32 if no bits are set and zero if bit[31] is set.
Restrictions
Do not use SP and do not use PC.
Condition flags
This instruction does not change the flags.
Examples
CLZ
CLZNE

ARM DUI 0553A
ID121610

R4,R9
R2,R3

3-48

The Cortex-M4 Instruction Set

3.5.5

CMP and CMN
Compare and Compare Negative.
Syntax
CMP{cond} Rn, Operand2
CMN{cond} Rn, Operand2

where:
cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the register holding the first operand.

Operand2

Is a flexible second operand. See Flexible second operand on page 3-12 for
details of the options.

Operation
These instructions compare the value in a register with Operand2. They update the condition
flags on the result, but do not write the result to a register.
The CMP instruction subtracts the value of Operand2 from the value in Rn. This is the same as a
SUBS instruction, except that the result is discarded.
The CMN instruction adds the value of Operand2 to the value in Rn. This is the same as an ADDS
instruction, except that the result is discarded.
Restrictions
In these instructions:
•
do not use PC
•
Operand2 must not be SP.
Condition flags
These instructions update the N, Z, C and V flags according to the result.
Examples
CMP
CMN
CMPGT

ARM DUI 0553A
ID121610

R2, R9
R0, #6400
SP, R7, LSL #2

3-49

The Cortex-M4 Instruction Set

3.5.6

MOV and MVN
Move and Move NOT.
Syntax
MOV{S}{cond} Rd, Operand2
MOV{cond} Rd, #imm16
MVN{S}{cond} Rd, Operand2

where:
S

Is an optional suffix. If S is specified, the condition code flags are updated on the
result of the operation, see Conditional execution on page 3-18.

cond

Is an optional condition code. See Conditional execution on page 3-18.

Specifies the destination register.

Operand2

Is a flexible second operand, see Flexible second operand on page 3-12 for details
of the options.

imm16

Is any value in the range 0-65535.

Operation
The MOV instruction copies the value of Operand2 into Rd.
When Operand2 in a MOV instruction is a register with a shift other than LSL #0, the preferred
syntax is the corresponding shift instruction:
•

ASR{S}{cond} Rd, Rm, #n is the preferred syntax for MOV{S}{cond} Rd, Rm, ASR #n

•

LSL{S}{cond} Rd, Rm, #n is the preferred syntax for MOV{S}{cond} Rd, Rm, LSL #n if n != 0

•

LSR{S}{cond} Rd, Rm, #n is the preferred syntax for MOV{S}{cond} Rd, Rm, LSR #n

•

ROR{S}{cond} Rd, Rm, #n is the preferred syntax for MOV{S}{cond} Rd, Rm, ROR #n

•

RRX{S}{cond} Rd, Rm is the preferred syntax for MOV{S}{cond} Rd, Rm, RRX.

Also, the MOV instruction permits additional forms of Operand2 as synonyms for shift instructions:
•
MOV{S}{cond} Rd, Rm, ASR Rs is a synonym for ASR{S}{cond} Rd, Rm, Rs
•
MOV{S}{cond} Rd, Rm, LSL Rs is a synonym for LSL{S}{cond} Rd, Rm, Rs
•
MOV{S}{cond} Rd, Rm, LSR Rs is a synonym for LSR{S}{cond} Rd, Rm, Rs
•
MOV{S}{cond} Rd, Rm, ROR Rs is a synonym for ROR{S}{cond} Rd, Rm, Rs
See ASR, LSL, LSR, ROR, and RRX on page 3-46.
The MVN instruction takes the value of Operand2, performs a bitwise logical NOT operation on the
value, and places the result into Rd.
Note
The MOVW instruction provides the same function as MOV, but is restricted to using the imm16
operand.

ARM DUI 0553A
ID121610

3-50

The Cortex-M4 Instruction Set

Restrictions
You can use SP and PC only in the MOV instruction, with the following restrictions:
•
the second operand must be a register without shift
•
you must not specify the S suffix.
When Rd is PC in a MOV instruction:
•
bit[0] of the value written to the PC is ignored
•
a branch occurs to the address created by forcing bit[0] of that value to 0.
Note
Though it is possible to use MOV as a branch instruction, ARM strongly recommends the use of
a BX or BLX instruction to branch for software portability to the ARM instruction set.

Condition flags
If S is specified, these instructions:
•

update the N and Z flags according to the result

•

can update the C flag during the calculation of Operand2, see Flexible second operand on
page 3-12

•

do not affect the V flag.

Example
MOVS
MOV
MOVS
MOV
MOV
MVNS

ARM DUI 0553A
ID121610

R11, #0x000B
R1, #0xFA05
R10, R12
R3, #23
R8, SP
R2, #0xF

;
;
;
;
;
;
;

Write value of 0x000B to R11, flags get updated
Write value of 0xFA05 to R1, flags are not updated
Write value in R12 to R10, flags get updated
Write value of 23 to R3
Write value of stack pointer to R8
Write value of 0xFFFFFFF0 (bitwise inverse of 0xF)
to the R2 and update flags.

3-51

The Cortex-M4 Instruction Set

3.5.7

MOVT
Move Top.
Syntax
MOVT{cond} Rd, #imm16

where:
cond
Rd
imm16

Is an optional condition code, see Conditional execution on page 3-18.
Specifies the destination register.
Is a 16-bit immediate constant.

Operation
MOVT writes a 16-bit immediate value, imm16, to the top halfword, Rd[31:16], of its destination
register. The write does not affect Rd[15:0].

The MOV, MOVT instruction pair enables you to generate any 32-bit constant.
Restrictions
Rd must not be SP and must not be PC.

Condition flags
This instruction does not change the flags.
Examples
MOVT

ARM DUI 0553A
ID121610

R3, #0xF123 ; Write 0xF123 to upper halfword of R3, lower halfword
; and APSR are unchanged.

3-52

The Cortex-M4 Instruction Set

3.5.8

REV, REV16, REVSH, and RBIT
Reverse bytes and Reverse bits.
Syntax
op{cond} Rd, Rn

where:
Is any of:

REV

Reverse byte order in a word.

REV16

Reverse byte order in each halfword independently.

REVSH

Reverse byte order in the bottom halfword, and sign extend to 32 bits.

RBIT

Reverse the bit order in a 32-bit word.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Specifies the register holding the operand.

Operation
Use these instructions to change endianness of data:
Converts either:

REV

REV16

REVSH

•

32-bit big-endian data into little-endian data

•

32-bit little-endian data into big-endian data.

Converts either:
•

16-bit big-endian data into little-endian data

•

16-bit little-endian data into big-endian data.

Converts either:
•

16-bit signed big-endian data into 32-bit signed little-endian data

•

16-bit signed little-endian data into 32-bit signed big-endian data.

Restrictions
Do not use SP and do not use PC.
Condition flags
These instructions do not change the flags.
Examples
REV
REV16
REVSH
REVHS
RBIT

ARM DUI 0553A
ID121610

R3,
R0,
R0,
R3,
R7,

R7
R0
R5
R7
R8

;
;
;
;
;

Reverse
Reverse
Reverse
Reverse
Reverse

byte order of value in R7 and write it to R3
byte order of each 16-bit halfword in R0
Signed Halfword
with Higher or Same condition
bit order of value in R8 and write the result to R7.

3-53

The Cortex-M4 Instruction Set

3.5.9

SADD16 and SADD8
Signed Add 16 and Signed Add 8.
Syntax
op{cond}{Rd,} Rn, Rm

where:
Is any of:

SADD16

Performs two 16-bit signed integer additions.

SADD8

Performs four 8-bit signed integer additions.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Specifies the first register holding the operand.

Specifies the second register holding the operand.

Operation
Use these instructions to perform a halfword or byte add in parallel:
The SADD16 instruction:
1.
Adds each halfword from the first operand to the corresponding halfword
of the second operand.
2.
Writes the result in the corresponding halfwords of the destination register.
The SADD8 instruction:
1.
Adds each byte of the first operand to the corresponding byte of the second
operand.
2.
Writes the result in the corresponding bytes of the destination register.
Restrictions
Do not use SP and do not use PC.
Condition flags
These instructions do not change the flags.
Examples
SADD16 R1, R0
SADD8

ARM DUI 0553A
ID121610

R4, R0, R5

;
;
;
;

Adds the halfwords in R0 to the corresponding halfwords of
R1 and writes to corresponding halfword of R1.
Adds bytes of R0 to the corresponding byte in R5 and writes
to the corresponding byte in R4.

3-54

The Cortex-M4 Instruction Set

3.5.10

SHADD16 and SHADD8
Signed Halving Add 16 and Signed Halving Add 8.
Syntax
op{cond}{Rd,} Rn, Rm

where:
Is any of:

SHADD16

Signed Halving Add 16

SHADD8

Signed Halving Add 8

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Specifies the first operand register.

Specifies the second operand register.

Operation
Use these instructions to add 16-bit and 8-bit data and then to halve the result before writing the
result to the destination register:
The SHADD16 instruction:
1.
Adds each halfword from the first operand to the corresponding halfword
of the second operand.
2.
Shuffles the result by one bit to the right, halving the data.
3.
Writes the halfword results in the destination register.
The SHADDB8 instruction:
1.
Adds each byte of the first operand to the corresponding byte of the second
operand.
2.
Shuffles the result by one bit to the right, halving the data.
3.
Writes the byte results in the destination register.
Restrictions
Do not use SP and do not use PC.
Condition flags
These instructions do not change the flags.
Examples
SHADD16 R1, R0
SHADD8

ARM DUI 0553A
ID121610

;
;
R4, R0, R5 ;
;

Adds halfwords in R0 to corresponding halfword of R1 and
writes halved result to corresponding halfword in R1
Adds bytes of R0 to corresponding byte in R5 and
writes halved result to corresponding byte in R4.

3-55

The Cortex-M4 Instruction Set

3.5.11

SHASX and SHSAX
Signed Halving Add and Subtract with Exchange and Signed Halving Subtract and Add with
Exchange.
Syntax
op{cond} {Rd}, Rn, Rm

where:
op

Is any of:
SHASX
SHSAX

Add and Subtract with Exchange and Halving.
Subtract and Add with Exchange and Halving.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Rn, Rm

Are registers holding the first and second operands.

Operation
The SHASX instruction:
1.
Adds the top halfword of the first operand with the bottom halfword of the second
operand.
2.
Writes the halfword result of the addition to the top halfword of the destination register,
shifted by one bit to the right causing a divide by two, or halving.
3.
Subtracts the top halfword of the second operand from the bottom highword of the first
operand.
4.
Writes the halfword result of the division in the bottom halfword of the destination
register, shifted by one bit to the right causing a divide by two, or halving.
The SHSAX instruction:
1.
Subtracts the bottom halfword of the second operand from the top highword of the first
operand.
2.
Writes the halfword result of the addition to the bottom halfword of the destination
register, shifted by one bit to the right causing a divide by two, or halving.
3.
Adds the bottom halfword of the first operand with the top halfword of the second
operand.
4.
Writes the halfword result of the division in the top halfword of the destination register,
shifted by one bit to the right causing a divide by two, or halving.
Restrictions
Do not use SP and do not use PC.
Condition flags
These instructions do not affect the condition code flags.

ARM DUI 0553A
ID121610

3-56

The Cortex-M4 Instruction Set

Examples

ARM DUI 0553A
ID121610

SHASX

R7, R4, R2

SHSAX

R0, R3, R5

;
;
;
;
;
;
;
;

Adds top halfword of R4 to bottom halfword of R2
and writes halved result to top halfword of R7
Subtracts top halfword of R2 from bottom halfword of
R4 and writes halved result to bottom halfword of R7
Subtracts bottom halfword of R5 from top halfword
of R3 and writes halved result to top halfword of R0
Adds top halfword of R5 to bottom halfword of R3 and
writes halved result to bottom halfword of R0.

3-57

The Cortex-M4 Instruction Set

3.5.12

SHSUB16 and SHSUB8
Signed Halving Subtract 16 and Signed Halving Subtract 8.
Syntax
op{cond}{Rd,} Rn, Rm

where:
Is any of:

SHSUB16

Signed Halving Subtract 16

SHSUB8

Signed Halving Subtract 8

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Specifies the first operand register.

Specifies the second operand register

Operation
Use these instructions to add 16-bit and 8-bit data and then to halve the result before writing the
result to the destination register:
The SHSUB16 instruction:
1.
Subtracts each halfword of the second operand from the corresponding
halfwords of the first operand.
2.
Shuffles the result by one bit to the right, halving the data.
3.
Writes the halved halfword results in the destination register.
The SHSUBB8 instruction:
1.
Subtracts each byte of the second operand from the corresponding byte of
the first operand,
2.
Shuffles the result by one bit to the right, halving the data,
3.
Writes the corresponding signed byte results in the destination register.
Restrictions
Do not use SP and do not use PC.
Condition flags
These instructions do not change the flags.
Examples
SHSUB16 R1, R0
SHSUB8

ARM DUI 0553A
ID121610

R4, R0, R5

;
;
;
;

Subtracts halfwords in R0 from corresponding halfword
of R1 and writes to corresponding halfword of R1
Subtracts bytes of R0 from corresponding byte in R5,
and writes to corresponding byte in R4.

3-58

The Cortex-M4 Instruction Set

3.5.13

SSUB16 and SSUB8
Signed Subtract 16 and Signed Subtract 8.
Syntax
op{cond}{Rd,} Rn, Rm

where:
Is any of:

SSUB16

Performs two 16-bit signed integer subtractions.

SSUB8

Performs four 8-bit signed integer subtractions.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Specifies the first operand register.

Specifies the second operand register.

Operation
Use these instructions to change endianness of data:
The SSUB16 instruction:
1.
Subtracts each halfword from the second operand from the corresponding
halfword of the first operand.
2.
Writes the difference result of two signed halfwords in the corresponding
halfword of the destination register.
The SSUB8 instruction:
1.
Subtracts each byte of the second operand from the corresponding byte of
the first operand
2.
Writes the difference result of four signed bytes in the corresponding byte
of the destination register.
Restrictions
Do not use SP and do not use PC.
Condition flags
These instructions do not change the flags.
Examples
SSUB16 R1, R0
SSUB8

ARM DUI 0553A
ID121610

R4, R0, R5

;
;
;
;

Subtracts halfwords in R0 from corresponding halfword of R1
and writes to corresponding halfword of R1
Subtracts bytes of R5 from corresponding byte in
R0, and writes to corresponding byte of R4.

3-59

The Cortex-M4 Instruction Set

3.5.14

SASX and SSAX
Signed Add and Subtract with Exchange and Signed Subtract and Add with Exchange.
Syntax
op{cond} {Rd}, Rm, Rn

where:
op

Is any of:
SASX

Signed Add and Subtract with Exchange.

SSAX

Signed Subtract and Add with Exchange.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Rn, Rm

Are registers holding the first and second operands.

Operation
The SASX instruction:
1.
Adds the signed top halfword of the first operand with the signed bottom halfword of the
second operand.
2.
Writes the signed result of the addition to the top halfword of the destination register.
3.
Subtracts the signed bottom halfword of the second operand from the top signed highword
of the first operand.
4.
Writes the signed result of the subtraction to the bottom halfword of the destination
register.
The SSAX instruction:
1.
Subtracts the signed bottom halfword of the second operand from the top signed highword
of the first operand.
2.
Writes the signed result of the addition to the bottom halfword of the destination register.
3.
Adds the signed top halfword of the first operand with the signed bottom halfword of the
second operand.
4.
Writes the signed result of the subtraction to the top halfword of the destination register.
Restrictions
Do not use SP and do not use PC.
Condition flags
These instructions do not affect the condition code flags.

ARM DUI 0553A
ID121610

3-60

The Cortex-M4 Instruction Set

Examples

ARM DUI 0553A
ID121610

SASX

R0, R4, R5

SSAX

R7, R3, R2

;
;
;
;
;
;
;
;

Adds top halfword of R4 to bottom halfword of R5 and
writes to top halfword of R0
Subtracts bottom halfword of R5 from top halfword of R4
and writes to bottom halfword of R0
Subtracts top halfword of R2 from bottom halfword of R3
and writes to bottom halfword of R7
Adds top halfword of R3 with bottom halfword of R2 and
writes to top halfword of R7.

3-61

The Cortex-M4 Instruction Set

3.5.15

TST and TEQ
Test bits and Test Equivalence.
Syntax
TST{cond} Rn, Operand2
TEQ{cond} Rn, Operand2

where:
cond

Is an optional condition code. See Conditional execution on page 3-18.

Specifies the register holding the first operand.

Operand2

Is a flexible second operand. See Flexible second operand on page 3-12 for
details of the options.

Operation
These instructions test the value in a register against Operand2. They update the condition flags
based on the result, but do not write the result to a register.
The TST instruction performs a bitwise AND operation on the value in Rn and the value of
Operand2. This is the same as the ANDS instruction, except that it discards the result.
To test whether a bit of Rn is 0 or 1, use the TST instruction with an Operand2 constant that has
that bit set to 1 and all other bits cleared to 0.
The TEQ instruction performs a bitwise Exclusive OR operation on the value in Rn and the value
of Operand2. This is the same as the EORS instruction, except that it discards the result.
Use the TEQ instruction to test if two values are equal without affecting the V or C flags.
TEQ is also useful for testing the sign of a value. After the comparison, the N flag is the logical

Exclusive OR of the sign bits of the two operands.
Restrictions
Do not use SP and do not use PC.
Condition flags
These instructions:
•

update the N and Z flags according to the result

•

can update the C flag during the calculation of Operand2, see Flexible second operand on
page 3-12

•

do not affect the V flag.

Examples

ARM DUI 0553A
ID121610

TST

R0, #0x3F8

TEQEQ

R10, R9

;
;
;
;

Perform bitwise AND of R0 value to 0x3F8,
APSR is updated but result is discarded
Conditionally test if value in R10 is equal to
value in R9, APSR is updated but result is discarded.

3-62

The Cortex-M4 Instruction Set

3.5.16

UADD16 and UADD8
Unsigned Add 16 and Unsigned Add 8.
Syntax
op{cond}{Rd,} Rn, Rm

where:
Is any of:

UADD16

Performs two 16-bit unsigned integer additions.

UADD8

Performs four 8-bit unsigned integer additions.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Specifies the first register holding the operand.

Specifies the second register holding the operand.

Operation
Use these instructions to add 16- and 8-bit unsigned data:
The UADD16 instruction:
1.
Adds each halfword from the first operand to the corresponding halfword of the second
operand.
2.
Writes the unsigned result in the corresponding halfwords of the destination register.
The UADD16 instruction:
1.
Adds each byte of the first operand to the corresponding byte of the second operand.
2.
Writes the unsigned result in the corresponding byte of the destination register.
Restrictions
Do not use SP and do not use PC.
Condition flags
These instructions do not change the flags.
Examples
UADD16 R1, R0
UADD8

ARM DUI 0553A
ID121610

R4, R0, R5

;
;
;
;

Adds halfwords in R0 to corresponding halfword of R1,
writes to corresponding halfword of R1
Adds bytes of R0 to corresponding byte in R5 and writes
to corresponding byte in R4.

3-63

The Cortex-M4 Instruction Set

3.5.17

UASX and USAX
Add and Subtract with Exchange and Subtract and Add with Exchange.
Syntax
op{cond} {Rd}, Rn, Rm

where:
op

Is one of:
Add and Subtract with Exchange.
USAX
Subtract and Add with Exchange.
Is an optional condition code, see Conditional execution on page 3-18.
Specifies the destination register.
Are registers holding the first and second operands.
UASX

cond
Rd
Rn, Rm

Operation
The UASX instruction:
1.
Subtracts the top halfword of the second operand from the bottom halfword of the first
operand.
2.
Writes the unsigned result from the subtraction to the bottom halfword of the destination
register.
3.
Adds the top halfword of the first operand with the bottom halfword of the second
operand.
4.
Writes the unsigned result of the addition to the top halfword of the destination register.
The USAX instruction:
1.
Adds the bottom halfword of the first operand with the top halfword of the second
operand.
2.
Writes the unsigned result of the addition to the bottom halfword of the destination
register.
3.
Subtracts the bottom halfword of the second operand from the top halfword of the first
operand.
4.
Writes the unsigned result from the subtraction to the top halfword of the destination
register.
Restrictions
Do not use SP and do not use PC.
Condition flags
These instructions do not affect the condition code flags.

ARM DUI 0553A
ID121610

3-64

The Cortex-M4 Instruction Set

Examples

ARM DUI 0553A
ID121610

UASX

R0, R4, R5

USAX

R7, R3, R2

;
;
;
;
;
;
;
;

Adds top halfword of R4 to bottom halfword of R5 and
writes to top halfword of R0
Subtracts bottom halfword of R5 from top halfword of R0
and writes to bottom halfword of R0
Subtracts top halfword of R2 from bottom halfword of R3
and writes to bottom halfword of R7
Adds top halfword of R3 to bottom halfword of R2 and
writes to top halfword of R7.

3-65

The Cortex-M4 Instruction Set

3.5.18

UHADD16 and UHADD8
Unsigned Halving Add 16 and Unsigned Halving Add 8.
Syntax
op{cond}{Rd,} Rn, Rm

where:
Is any of:

Unsigned Halving Add 16.
UHADD8
Unsigned Halving Add 8.
Is an optional condition code, see Conditional execution on page 3-18.
Specifies the destination register.
Specifies the register holding the first operand.
Specifies the register holding the second operand.
UHADD16

cond
Rd
Rn
Rm

Operation
Use these instructions to add 16- and 8-bit data and then to halve the result before writing the
result to the destination register:
The UHADD16 instruction:
1.
Adds each halfword from the first operand to the corresponding halfword of the second
operand.
2.
Shuffles the halfword result by one bit to the right, halving the data.
3.
Writes the unsigned results to the corresponding halfword in the destination register.
The UHADD8 instruction:
1.
Adds each byte of the first operand to the corresponding byte of the second operand.
2.
Shuffles the byte result by one bit to the right, halving the data.
3.
Writes the unsigned results in the corresponding byte in the destination register.
Restrictions
Do not use SP and do not use PC.
Condition flags
These instructions do not change the flags.
Examples
UHADD16 R7, R3
UHADD8

ARM DUI 0553A
ID121610

R4, R0, R5

;
;
;
;

Adds halfwords in R7 to corresponding halfword of R3
and writes halved result to corresponding halfword in R7
Adds bytes of R0 to corresponding byte in R5 and writes
halved result to corresponding byte in R4.

3-66

The Cortex-M4 Instruction Set

3.5.19

UHASX and UHSAX
Unsigned Halving Add and Subtract with Exchange and Unsigned Halving Subtract and Add
with Exchange.
Syntax
op{cond} {Rd}, Rn, Rm

where:
op

Is one of:
Add and Subtract with Exchange and Halving.
UHSAX
Subtract and Add with Exchange and Halving.
Is an optional condition code, see Conditional execution on page 3-18.
Specifies the destination register.
Are registers holding the first and second operands.
UHASX

cond
Rd
Rn, Rm

Operation
The UHASX instruction:
1.
Adds the top halfword of the first operand with the bottom halfword of the second
operand.
2.
Shifts the result by one bit to the right causing a divide by two, or halving.
3.
Writes the halfword result of the addition to the top halfword of the destination register.
4.
Subtracts the top halfword of the second operand from the bottom highword of the first
operand.
5.
Shifts the result by one bit to the right causing a divide by two, or halving.
6.
Writes the halfword result of the division in the bottom halfword of the destination
register.
The UHSAX instruction:
1.
Subtracts the bottom halfword of the second operand from the top highword of the first
operand.
2.
Shifts the result by one bit to the right causing a divide by two, or halving.
3.
Writes the halfword result of the subtraction in the top halfword of the destination register.
4.
Adds the bottom halfword of the first operand with the top halfword of the second
operand.
5.
Shifts the result by one bit to the right causing a divide by two, or halving.
6.
Writes the halfword result of the addition to the bottom halfword of the destination
register.
Restrictions
Do not use SP and do not use PC.
Condition flags
These instructions do not affect the condition code flags.

ARM DUI 0553A
ID121610

3-67

The Cortex-M4 Instruction Set

Examples

ARM DUI 0553A
ID121610

UHASX

R7, R4, R2

UHSAX

R0, R3, R5

;
;
;
;
;
;
;
;

Adds top halfword of R4 with bottom halfword of R2
and writes halved result to top halfword of R7
Subtracts top halfword of R2 from bottom halfword of
R7 and writes halved result to bottom halfword of R7
Subtracts bottom halfword of R5 from top halfword of
R3 and writes halved result to top halfword of R0
Adds top halfword of R5 to bottom halfword of R3 and
writes halved result to bottom halfword of R0.

3-68

The Cortex-M4 Instruction Set

3.5.20

UHSUB16 and UHSUB8
Unsigned Halving Subtract 16 and Unsigned Halving Subtract 8.
Syntax
op{cond}{Rd,} Rn, Rm

where:
Is any of:

Performs two unsigned 16-bit integer additions, halves the results, and
writes the results to the destination register.
UHSUB8
Performs four unsigned 8-bit integer additions, halves the results, and
writes the results to the destination register.
Is an optional condition code, see Conditional execution on page 3-18.
Specifies the destination register.
Specifies the first register holding the operand.
Specifies the second register holding the operand.
UHSUB16

cond
Rd
Rn
Rm

Operation
Use these instructions to add 16-bit and 8-bit data and then to halve the result before writing the
result to the destination register:
The UHSUB16 instruction:
1.
Subtracts each halfword of the second operand from the corresponding halfword of the
first operand.
2.
Shuffles each halfword result to the right by one bit, halving the data.
3.
Writes each unsigned halfword result to the corresponding halfwords in the destination
register.
The UHSUB8 instruction:
1.
Subtracts each byte of second operand from the corresponding byte of the first operand.
2.
Shuffles each byte result by one bit to the right, halving the data.
3.
Writes the unsigned byte results to the corresponding byte of the destination register.
Restrictions
Do not use SP and do not use PC.
Condition flags
These instructions do not change the flags.
Examples
UHSUB16

UHSUB8

ARM DUI 0553A
ID121610

R1, R0

R4, R0, R5

;
;
;
;
;

Subtracts halfwords in R0 from corresponding halfword of
R1 and writes halved result to corresponding halfword in
R1
Subtracts bytes of R5 from corresponding byte in R0 and
writes halved result to corresponding byte in R4.

3-69

The Cortex-M4 Instruction Set

3.5.21

SEL
Select Bytes. Selects each byte of its result from either its first operand or its second operand,
according to the values of the GE flags.
Syntax
SEL{}{} {,} ,

where:
,

Is a standard assembler syntax fields.
Specifies the destination register.
Specifies the first operand register.
Specifies the second operand register.

Operation
The SEL instruction:
1.

Reads the value of each bit of APSR.GE.

Depending on the value of APSR.GE, assigns the destination register the value of either
the first or second operand register.

Restrictions
None.
Condition flags
These instructions do not change the flags.
Examples
SADD16 R0, R1, R2
SEL R0, R0, R3

ARM DUI 0553A
ID121610

; Set GE bits based on result
; Select bytes from R0 or R3, based on GE.

3-70

The Cortex-M4 Instruction Set

3.5.22

USAD8
Unsigned Sum of Absolute Differences.
Syntax
USAD8{cond}{Rd,} Rn, Rm

where:
cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Specifies the first operand register.

Specifies the second operand register.

Operation
The USAD8 instruction:
1.
Subtracts each byte of the second operand register from the corresponding byte of the first
operand register.
2.
Adds the absolute values of the differences together.
3.
Writes the result to the destination register.
Restrictions
Do not use SP and do not use PC.
Condition flags
These instructions do not change the flags.
Examples
USAD8 R1, R4, R0
USAD8

ARM DUI 0553A
ID121610

R0, R5

;
;
;
;

Subtracts each byte in R0 from corresponding byte of R4
adds the differences and writes to R1
Subtracts bytes of R5 from corresponding byte in R0,
adds the differences and writes to R0.

3-71

The Cortex-M4 Instruction Set

3.5.23

USADA8
Unsigned Sum of Absolute Differences and Accumulate.
Syntax
USADA8{cond}{Rd,} Rn, Rm, Ra

where:
cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Specifies the first operand register.

Specifies the second operand register.

Specifies the register that contains the accumulation value.

Operation
The USADA8 instruction:
1.
Subtracts each byte of the second operand register from the corresponding byte of the first
operand register.
2.
Adds the unsigned absolute differences together.
3.
Adds the accumulation value to the sum of the absolute differences.
4.
Writes the result to the destination register.
Restrictions
Do not use SP and do not use PC.
Condition flags
These instructions do not change the flags.
Examples
USADA8 R1, R0, R6
USADA8

ARM DUI 0553A
ID121610

;
;
R4, R0, R5, R2 ;
;

Subtracts bytes in R0 from corresponding halfword of R1
adds differences, adds value of R6, writes to R1
Subtracts bytes of R5 from corresponding byte in R0
adds differences, adds value of R2 writes to R4.

3-72

The Cortex-M4 Instruction Set

3.5.24

USUB16 and USUB8
Unsigned Subtract 16 and Unsigned Subtract 8.
Syntax
op{cond}{Rd,} Rn, Rm

where:
Is any of:

USUB16

Unsigned Subtract 16.

USUB8

Unsigned Subtract 8.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Specifies the first operand register.

Specifies the second operand register.

Operation
Use these instructions to subtract 16-bit and 8-bit data before writing the result to the destination
register:
The USUB16 instruction:
1.
Subtracts each halfword from the second operand register from the corresponding
halfword of the first operand register.
2.
Writes the unsigned result in the corresponding halfwords of the destination register.
The USUB8 instruction:
1.
Subtracts each byte of the second operand register from the corresponding byte of the first
operand register.
2.
Writes the unsigned byte result in the corresponding byte of the destination register.
Restrictions
Do not use SP and do not use PC.
Condition flags
These instructions do not change the flags.
Examples
USUB16 R1, R0

ARM DUI 0553A
ID121610

;
;
;
;

Subtracts halfwords in R0 from corresponding halfword of R1
and writes to corresponding halfword in R1USUB8 R4, R0, R5
Subtracts bytes of R5 from corresponding byte in R0 and
writes to the corresponding byte in R4.

3-73

The Cortex-M4 Instruction Set

3.6

Multiply and divide instructions
Table 3-9 shows the multiply and divide instructions:
Table 3-9 Multiply and divide instructions
Mnemonic

Brief description

See

MLA

Multiply with Accumulate, 32-bit result

MUL, MLA, and MLS on page 3-75

MLS

Multiply and Subtract, 32-bit result

MUL, MLA, and MLS on page 3-75

MUL

Multiply, 32-bit result

MUL, MLA, and MLS on page 3-75

SDIV

Signed Divide

SDIV and UDIV on page 3-94

SMLA[B,T]

Signed Multiply Accumulate (halfwords)

SMLA and SMLAW on page 3-79

SMLAD, SMLADX

Signed Multiply Accumulate Dual

SMLAD on page 3-81

SMLAL

Signed Multiply with Accumulate (32x32+64),
64-bit result

UMULL, UMLAL, SMULL, and SMLAL on page 3-93

SMLAL[B,T]

Signed Multiply Accumulate Long (halfwords)

SMLAL and SMLALD on page 3-82

SMLALD, SMLALDX

Signed Multiply Accumulate Long Dual

SMLAL and SMLALD on page 3-82

SMLAW[B|T]

Signed Multiply Accumulate (word by halfword)

SMLA and SMLAW on page 3-79

SMLSD

Signed Multiply Subtract Dual

SMLSD and SMLSLD on page 3-84

SMLSLD

Signed Multiply Subtract Long Dual

SMLSD and SMLSLD on page 3-84

SMMLA

Signed Most Significant Word Multiply
Accumulate

SMMLA and SMMLS on page 3-86

SMMLS, SMMLSR

Signed Most Significant Word Multiply Subtract

SMMLA and SMMLS on page 3-86

SMUAD, SMUADX

Signed Dual Multiply Add

SMUAD and SMUSD on page 3-89

SMUL[B,T]

Signed Multiply (word by halfword)

SMUL and SMULW on page 3-91

SMMUL, SMMULR

Signed Most Significant Word Multiply

SMMUL on page 3-88

SMULL

Signed Multiply (32x32), 64-bit result

UMULL, UMLAL, SMULL, and SMLAL on page 3-93

SMULWB, SMULWT

Signed Multiply (word by halfword)

SMUL and SMULW on page 3-91

SMUSD, SMUSDX

Signed Dual Multiply Subtract

SMUAD and SMUSD on page 3-89

UDIV

Unsigned Divide

SDIV and UDIV on page 3-94

UMAAL

Unsigned Multiply Accumulate Accumulate Long
(32x32+32+32), 64-bit result

UMULL, UMAAL, UMLAL on page 3-77

UMLAL

Unsigned Multiply with Accumulate (32x32+64),
64-bit result

UMULL, UMLAL, SMULL, and SMLAL on page 3-93

UMULL

Unsigned Multiply (32x32), 64-bit result

UMULL, UMLAL, SMULL, and SMLAL on page 3-93

ARM DUI 0553A
ID121610

3-74

The Cortex-M4 Instruction Set

3.6.1

MUL, MLA, and MLS
Multiply, Multiply with Accumulate, and Multiply with Subtract, using 32-bit operands, and
producing a 32-bit result.
Syntax
MUL{S}{cond} {Rd,} Rn, Rm ; Multiply
MLA{cond} Rd, Rn, Rm, Ra

; Multiply with accumulate

MLS{cond} Rd, Rn, Rm, Ra

; Multiply with subtract

where:
cond

Is an optional condition code, see Conditional execution on page 3-18.

Is an optional suffix. If S is specified, the condition code flags are updated on the
result of the operation, see Conditional execution on page 3-18.

Specifies the destination register. If Rd is omitted, the destination register is Rn.

Rn, Rm

Are registers holding the values to be multiplied.

Is a register holding the value to be added or subtracted from.

Operation
The MUL instruction multiplies the values from Rn and Rm, and places the least significant 32 bits
of the result in Rd.
The MLA instruction multiplies the values from Rn and Rm, adds the value from Ra, and places the
least significant 32 bits of the result in Rd.
The MLS instruction multiplies the values from Rn and Rm, subtracts the product from the value
from Ra, and places the least significant 32 bits of the result in Rd.
The results of these instructions do not depend on whether the operands are signed or unsigned.

ARM DUI 0553A
ID121610

3-75

The Cortex-M4 Instruction Set

Restrictions
In these instructions, do not use SP and do not use PC.
If you use the S suffix with the MUL instruction:
•
Rd, Rn, and Rm must all be in the range R0 to R7
•
Rd must be the same as Rm
•
you must not use the cond suffix.
Condition flags
If S is specified, the MUL instruction:
•
updates the N and Z flags according to the result
•
does not affect the C and V flags.
Examples
MUL
MLA
MULS
MULLT
MLS

ARM DUI 0553A
ID121610

R10, R2, R5
R10, R2, R1, R5
R0, R2, R2
R2, R3, R2
R4, R5, R6, R7

;
;
;
;
;

Multiply, R10
Multiply with
Multiply with
Conditionally
Multiply with

= R2 x R5
accumulate, R10 =
flag update, R0 =
multiply, R2 = R3
subtract, R4 = R7

(R2 x R1) + R5
R2 x R2
x R2
- (R5 x R6)

3-76

The Cortex-M4 Instruction Set

3.6.2

UMULL, UMAAL, UMLAL
Unsigned Long Multiply, with optional Accumulate, using 32-bit operands and producing a
64-bit result.
Syntax
op{cond} RdLo, RdHi, Rn, Rm

where:
op

Is one of:
UMULL

Unsigned Long Multiply.

UMAAL

Unsigned Long Multiply with Accumulate Accumulate.

UMLAL

Unsigned Long Multiply, with Accumulate.

cond

Is an optional condition code, see Conditional execution on page 3-18.

RdHi, RdLo

Are the destination registers. For UMAAL, UMLAL and UMLAL they also hold the
accumulating value.

Rn, Rm

Are registers holding the first and second operands.

Operation
These instructions interpret the values from Rn and Rm as unsigned 32-bit integers. The UMULL
instruction:
•
Multiplies the two unsigned integers in the first and second operands.
•
Writes the least significant 32 bits of the result in RdLo.
•
Writes the most significant 32 bits of the result in RdHi.
The UMAAL instruction:
•
Multiplies the two unsigned 32-bit integers in the first and second operands.
•
Adds the unsigned 32-bit integer in RdHi to the 64-bit result of the multiplication.
•
Adds the unsigned 32-bit integer in RdLo to the 64-bit result of the addition.
•
Writes the top 32-bits of the result to RdHi.
•
Writes the lower 32-bits of the result to RdLo.
The UMLAL instruction:
•
multiplies the two unsigned integers in the first and second operands.
•
Adds the 64-bit result to the 64-bit unsigned integer contained in RdHi and RdLo.
•
Writes the result back to RdHi and RdLo.
Restrictions
In these instructions:
•
do not use SP and do not use PC.
•
RdHi and RdLo must be different registers.
Condition flags
These instructions do not affect the condition code flags.

ARM DUI 0553A
ID121610

3-77

The Cortex-M4 Instruction Set

Examples

ARM DUI 0553A
ID121610

UMULL

R0, R4, R5, R6

UMAAL

R3, R6, R2, R7

UMLAL

R2, R1, R3, R5

;
;
;
;
;

Multiplies R5 and R6, writes the top 32 bits to R4
and the bottom 32 bits to R0
Multiplies R2 and R7, adds R6, adds R3, writes the
top 32 bits to R6, and the bottom 32 bits to R3
Multiplies R5 and R3, adds R1:R2, writes to R1:R2.

3-78

The Cortex-M4 Instruction Set

3.6.3

SMLA and SMLAW
Signed Multiply Accumulate (halfwords).
Syntax
op{XY}{cond} Rd, Rn, Rm
op{Y}{cond} Rd, Rn, Rm, Ra

where:
op

Is one of:
SMLA

Signed Multiply Accumulate Long (halfwords)
X and Y specifies which half of the source registers Rn and Rm are used
as the first and second multiply operand.
If X is B, then the bottom halfword, bits [15:0], of Rn is used. If X is T,
then the top halfword, bits [31:16], of Rn is used.
If Y is B, then the bottom halfword, bits [15:0], of Rm is used. If Y is T,
then the top halfword, bits [31:16], of Rm is used.

SMLAW

Signed Multiply Accumulate (word by halfword)
Y specifies which half of the source register Rm is used as the second
multiply operand.
If Y is T, then the top halfword, bits [31:16] of Rm is used.
If Y is B, then the bottom halfword, bits [15:0] of Rm is used.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register. If Rd is omitted, the destination register is Rn.

Rn, Rm

Are registers holding the values to be multiplied.

Is a register holding the value to be added or subtracted from.

Operation
The SMALBB, SMLABT, SMLATB, SMLATT instructions:
•
Multiplies the specified signed halfword, top or bottom, values from Rn and Rm.
•
Adds the value in Ra to the resulting 32-bit product.
•
Writes the result of the multiplication and addition in Rd.
The non-specified halfwords of the source registers are ignored.
The SMLAWB and SMLAWT instructions:
•
Multiply the 32-bit signed values in Rn with:
— The top signed halfword of Rm, T instruction suffix.
— The bottom signed halfword of Rm, B instruction suffix.
•
Add the 32-bit signed value in Ra to the top 32 bits of the 48-bit product
•
Writes the result of the multiplication and addition in Rd.
The bottom 16 bits of the 48-bit product are ignored.
If overflow occurs during the addition of the accumulate value, the instruction sets the Q flag in
the APSR. No overflow can occur during the multiplication.

ARM DUI 0553A
ID121610

3-79

The Cortex-M4 Instruction Set

Restrictions
In these instructions, do not use SP and do not use PC.
Condition flags
If an overflow is detected, the Q flag is set.
Examples
SMLABB
SMLATB
SMLATT
SMLABT
SMLABT
SMLAWB
SMLAWT

ARM DUI 0553A
ID121610

R5, R6, R4, R1

;
;
R5, R6, R4, R1 ;
;
R5, R6, R4, R1 ;
;
R5, R6, R4, R1 ;
;
R4, R3, R2
;
;
R10, R2, R5, R3 ;
;
R10, R2, R1, R5 ;
;

Multiplies bottom halfwords of R6 and R4, adds
R1 and writes to R5
Multiplies top halfword of R6 with bottom halfword
of R4, adds R1 and writes to R5
Multiplies top halfwords of R6 and R4, adds
R1 and writes the sum to R5
Multiplies bottom halfword of R6 with top halfword
of R4, adds R1 and writes to R5
Multiplies bottom halfword of R4 with top halfword of
R3, adds R2 and writes to R4
Multiplies R2 with bottom halfword of R5, adds
R3 to the result and writes top 32-bits to R10
Multiplies R2 with top halfword of R1, adds R5
and writes top 32-bits to R10.

3-80

The Cortex-M4 Instruction Set

3.6.4

SMLAD
Signed Multiply Accumulate Long Dual.
Syntax
op{X}{cond} Rd, Rn, Rm, Ra

;

where:
op

Is one of:

SMLAD

Signed Multiply Accumulate Dual

SMLADX

Signed Multiply Accumulate Dual Reverse
X specifies which halfword of the source register Rn is used as the multiply
operand. If X is omitted, the multiplications are bottom × bottom and top × top. If
X is present, the multiplications are bottom × top and top × bottom.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Specifies the first operand register holding the values to be multiplied.

Specifies the second operand register.

Specifies the accumulate value.

Operation
The SMLAD and SMLADX instructions regard the two operands as four halfword 16-bit values. The
SMLAD and SMLADX instructions:
•
If X is not present, multiply the top signed halfword value in Rn with the top signed
halfword of Rm and the bottom signed halfword values in Rn with the bottom signed
halfword of Rm.
•
Or if X is present, multiply the top signed halfword value in Rn with the bottom signed
halfword of Rm and the bottom signed halfword values in Rn with the top signed halfword
of Rm.
•
Add both multiplication results to the signed 32-bit value in Ra.
•
Writes the 32-bit signed result of the multiplication and addition to Rd.
Restrictions
Do not use SP and do not use PC.
Condition flags
These instructions do not change the flags.
Examples
SMLAD

R10, R2, R1, R5 ;
;
;
SMLALDX R0, R2, R4, R6 ;
;
;

ARM DUI 0553A
ID121610

Multiplies two halfword values in R2 with
corresponding halfwords in R1, adds R5 and writes to
R10
Multiplies top halfword of R2 with bottom halfword
of R4, multiplies bottom halfword of R2 with top
halfword of R4, adds R6 and writes to R0.

3-81

The Cortex-M4 Instruction Set

3.6.5

SMLAL and SMLALD
Signed Multiply Accumulate Long, Signed Multiply Accumulate Long (halfwords) and Signed
Multiply Accumulate Long Dual.
Syntax
op{cond} RdLo, RdHi, Rn, Rm
op{XY}{cond} RdLo, RdHi, Rn, Rm
op{X}{cond} RdLo, RdHi, Rn, Rm

where:
op

Is one of:

SMLAL

Signed Multiply Accumulate Long

SMLAL

Signed Multiply Accumulate Long (halfwords, X and Y)
X and Y specify which halfword of the source registers Rn and Rm are used as the
first and second multiply operand:
If X is B, then the bottom halfword, bits [15:0], of Rn is used. If X is T, then the top
halfword, bits [31:16], of Rn is used.
If Y is B, then the bottom halfword, bits [15:0], of Rm is used. If Y is T, then the top
halfword, bits [31:16], of Rm is used.

SMLALD

Signed Multiply Accumulate Long Dual

SMLALDX

Signed Multiply Accumulate Long Dual Reversed
If the X is omitted, the multiplications are bottom × bottom and top × top.
If X is present, the multiplications are bottom × top and top × bottom.

cond

Is an optional condition code, see Conditional execution on page 3-18.

RdHi, RdLo

Are the destination registers. RdLo is the lower 32 bits and RdHi is the upper 32 bits
of the 64-bit integer. For SMLAL, SMLALBB, SMLALBT, SMLALTB, SMLALTT, SMLALD and
SMLALDX, they also hold the accumulating value.

Rn, Rm

Are registers holding the first and second operands.

Operation
The SMLAL instruction:
•
Multiplies the two’s complement signed word values from Rn and Rm.
•
Adds the 64-bit value in RdLo and RdHi to the resulting 64-bit product.
•
Writes the 64-bit result of the multiplication and addition in RdLo and RdHi.
The SMLALBB, SMLALBT, SMLALTB and SMLALTT instructions:
•
Multiplies the specified signed halfword, Top or Bottom, values from Rn and Rm.
•
Adds the resulting sign-extended 32-bit product to the 64-bit value in RdLo and RdHi.
•
Writes the 64-bit result of the multiplication and addition in RdLo and RdHi.
The non-specified halfwords of the source registers are ignored.

ARM DUI 0553A
ID121610

3-82

The Cortex-M4 Instruction Set

The SMLALD and SMLALDX instructions interpret the values from Rn and Rm as four halfword two’s
complement signed 16-bit integers. These instructions:
•
if X is not present, multiply the top signed halfword value of Rn with the top signed
halfword of Rm and the bottom signed halfword values of Rn with the bottom signed
halfword of Rm.
•
Or if X is present, multiply the top signed halfword value of Rn with the bottom signed
halfword of Rm and the bottom signed halfword values of Rn with the top signed halfword
of Rm.
•
Add the two multiplication results to the signed 64-bit value in RdLo and RdHi to create the
resulting 64-bit product.
•
Write the 64-bit product in RdLo and RdHi.
Restrictions
In these instructions:
•
do not use SP and do not use PC.
•
RdHi and RdLo must be different registers.
Condition flags
These instructions do not affect the condition code flags.
Examples

ARM DUI 0553A
ID121610

SMLAL

R4, R5, R3, R8

SMLALBT

R2, R1, R6, R7

SMLALTB

R2, R1, R6, R7

SMLALD

R6, R8, R5, R1

SMLALDX

R6, R8, R5, R1

;
;
;
;
;
;
;
;
;
;
;
;
;
;
;

Multiplies R3 and R8, adds R5:R4 and writes to
R5:R4
Multiplies bottom halfword of R6 with top
halfword of R7, sign extends to 32-bit, adds
R1:R2 and writes to R1:R2
Multiplies top halfword of R6 with bottom
halfword of R7, sign extends to 32-bit, adds R1:R2
and writes to R1:R2
Multiplies top halfwords in R5 and R1 and bottom
halfwords of R5 and R1, adds R8:R6 and writes to
R8:R6
Multiplies top halfword in R5 with bottom
halfword of R1, and bottom halfword of R5 with
top halfword of R1, adds R8:R6 and writes to
R8:R6.

3-83

The Cortex-M4 Instruction Set

3.6.6

SMLSD and SMLSLD
Signed Multiply Subtract Dual and Signed Multiply Subtract Long Dual.
Syntax
op{X}{cond} Rd, Rn, Rm, Ra

where:
op

Is one of:
SMLSD

Signed Multiply Subtract Dual.

SMLSDX

Signed Multiply Subtract Dual Reversed.

SMLSLD

Signed Multiply Subtract Long Dual.

SMLSLDX

Signed Multiply Subtract Long Dual Reversed.

If X is present, the multiplications are bottom × top and top × bottom. If the X is
omitted, the multiplications are bottom × bottom and top × top.
cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Rn, Rm

Are registers holding the first and second operands.

Specifies the register holding the accumulate value.

Operation
The SMLSD instruction interprets the values from the first and second operands as four signed
halfwords. This instruction:
•
Optionally rotates the halfwords of the second operand.
•
Performs two signed 16 × 16-bit halfword multiplications.
•
Subtracts the result of the upper halfword multiplication from the result of the lower
halfword multiplication.
•
Adds the signed accumulate value to the result of the subtraction.
•
Writes the result of the addition to the destination register.
The SMLSLD instruction interprets the values from Rn and Rm as four signed halfwords. This
instruction:
•
Optionally rotates the halfwords of the second operand.
•
Performs two signed 16 × 16-bit halfword multiplications.
•
Subtracts the result of the upper halfword multiplication from the result of the lower
halfword multiplication.
•
Adds the 64-bit value in RdHi and RdLo to the result of the subtraction.
•
Writes the 64-bit result of the addition to the RdHi and RdLo.
Restrictions
In these instructions:
•
Do not use SP and do not use PC.

ARM DUI 0553A
ID121610

3-84

The Cortex-M4 Instruction Set

Condition flags
This instruction sets the Q flag if the accumulate operation overflows. Overflow cannot occur
during the multiplications or subtraction.
For the Thumb instruction set, these instructions do not affect the condition code flags.
Examples

ARM DUI 0553A
ID121610

SMLSD

R0, R4, R5, R6

SMLSDX

R1, R3, R2, R0

SMLSLD

R3, R6, R2, R7

SMLSLDX

R3, R6, R2, R7

;
;
;
;
;
;
;
;
;
;
;
;
;
;
;
;

Multiplies bottom halfword of R4 with bottom
halfword of R5, multiplies top halfword of R4
with top halfword of R5, subtracts second from
first, adds R6, writes to R0
Multiplies bottom halfword of R3 with top
halfword of R2, multiplies top halfword of R3
with bottom halfword of R2, subtracts second from
first, adds R0, writes to R1
Multiplies bottom halfword of R6 with bottom
halfword of R2, multiplies top halfword of R6
with top halfword of R2, subtracts second from
first, adds R6:R3, writes to R6:R3
Multiplies bottom halfword of R6 with top
halfword of R2, multiplies top halfword of R6
with bottom halfword of R2, subtracts second from
first, adds R6:R3, writes to R6:R3.

3-85

The Cortex-M4 Instruction Set

3.6.7

SMMLA and SMMLS
Signed Most Significant Word Multiply Accumulate and Signed Most Significant Word
Multiply Subtract.
Syntax
op{R}{cond} Rd, Rn, Rm, Ra

where:
op

Is one of:
SMMLA

Signed Most Significant Word Multiply Accumulate.

SMMLS

Signed Most Significant Word Multiply Subtract.

Is a rounding error flag. If R is specified, the result is rounded instead of being
truncated. In this case the constant 0x80000000 is added to the product before the
high word is extracted.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Rn, Rm

Are registers holding the first and second multiply operands.

Specifies the register holding the accumulate value.

Operation
The SMMLA instruction interprets the values from Rn and Rm as signed 32-bit words.
The SMMLA instruction:
•
Multiplies the values in Rn and Rm.
•
Optionally rounds the result by adding 0x80000000.
•
Extracts the most significant 32 bits of the result.
•
Adds the value of Ra to the signed extracted value.
•
Writes the result of the addition in Rd.
The SMMLS instruction interprets the values from Rn and Rm as signed 32-bit words.
The SMMLS instruction:
•
Multiplies the values in Rn and Rm.
•
Optionally rounds the result by adding 0x80000000.
•
Extracts the most significant 32 bits of the result.
•
Subtracts the extracted value of the result from the value in Ra.
•
Writes the result of the subtraction in Rd.
Restrictions
In these instructions:
•
Do not use SP and do not use PC.
Condition flags
These instructions do not affect the condition code flags.

ARM DUI 0553A
ID121610

3-86

The Cortex-M4 Instruction Set

Examples

ARM DUI 0553A
ID121610

SMMLA

R0, R4, R5, R6

SMMLAR

R6, R2, R1, R4

SMMLSR

R3, R6, R2, R7

SMMLS

R4, R5, R3, R8

;
;
;
;
;
;
;
;

Multiplies R4 and R5, extracts top
R6, truncates and writes to R0
Multiplies R2 and R1, extracts top
R4, rounds and writes to R6
Multiplies R6 and R2, extracts top
subtracts R7, rounds and writes to
Multiplies R5 and R3, extracts top
subtracts R8, truncates and writes

32 bits, adds
32 bits, adds
32 bits,
R3
32 bits,
to R4.

3-87

The Cortex-M4 Instruction Set

3.6.8

SMMUL
Signed Most Significant Word Multiply.
Syntax
op{R}{cond} Rd, Rn, Rm

where:
op

Is one of:

SMMUL

Signed Most Significant Word Multiply

Is a rounding error flag. If R is specified, the result is rounded instead of being
truncated. In this case the constant 0x80000000 is added to the product before the
high word is extracted.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Rn, Rm

Are registers holding the first and second operands.

Operation
The SMMUL instruction interprets the values from Rn and Rm as two’s complement 32-bit signed
integers. The SMMUL instruction:
•
Multiplies the values from Rn and Rm.
•
Optionally rounds the result, otherwise truncates the result.
•
Writes the most significant signed 32 bits of the result in Rd.
Restrictions
In this instruction:
•
do not use SP and do not use PC.
Condition flags
This instruction does not affect the condition code flags.
Examples

ARM DUI 0553A
ID121610

SMULL

R0, R4, R5

SMULLR

R6, R2

;
;
;
;

Multiplies
and writes
Multiplies
and writes

R4
to
R6
to

and R5, truncates top 32 bits
R0
and R2, rounds the top 32 bits
R6.

3-88

The Cortex-M4 Instruction Set

3.6.9

SMUAD and SMUSD
Signed Dual Multiply Add and Signed Dual Multiply Subtract.
Syntax
op{X}{cond} Rd, Rn, Rm

where:
Is one of:

SMUAD

Signed Dual Multiply Add.

SMUADX

Signed Dual Multiply Add Reversed.

SMUSD

Signed Dual Multiply Subtract.

SMUSDX

Signed Dual Multiply Subtract Reversed.

If X is present, the multiplications are bottom × top and top × bottom. If the X is
omitted, the multiplications are bottom × bottom and top × top.
cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Rn, Rm

Are registers holding the first and the second operands.

Operation
The SMUAD instruction interprets the values from the first and second operands as two signed
halfwords in each operand. This instruction:
•
Optionally rotates the halfwords of the second operand.
•
Performs two signed 16 × 16-bit multiplications.
•
Adds the two multiplication results together.
•
Writes the result of the addition to the destination register.
The SMUSD instruction interprets the values from the first and second operands as two’s
complement signed integers. This instruction:
•
Optionally rotates the halfwords of the second operand.
•
Performs two signed 16 × 16-bit multiplications.
•
Subtracts the result of the top halfword multiplication from the result of the bottom
halfword multiplication.
•
Writes the result of the subtraction to the destination register.
Restrictions
In these instructions:
•
Do not use SP and do not use PC.
Condition flags
Sets the Q flag if the addition overflows. The multiplications cannot overflow.
Examples
SMUAD

ARM DUI 0553A
ID121610

R0, R4, R5

; Multiplies bottom halfword of R4 with the bottom
; halfword of R5, adds multiplication of top halfword
; of R4 with top halfword of R5, writes to R0

3-89

The Cortex-M4 Instruction Set

ARM DUI 0553A
ID121610

SMUADX

R3, R7, R4

SMUSD

R3, R6, R2

SMUSDX

R4, R5, R3

;
;
;
;
;
;
;
;
;

Multiplies bottom halfword of R7 with top halfword
of R4, adds multiplication of top halfword of R7
with bottom halfword of R4, writes to R3
Multiplies bottom halfword of R4 with bottom halfword
of R6, subtracts multiplication of top halfword of R6
with top halfword of R3, writes to R3
Multiplies bottom halfword of R5 with top halfword of
R3, subtracts multiplication of top halfword of R5
with bottom halfword of R3, writes to R4.

3-90

The Cortex-M4 Instruction Set

3.6.10

SMUL and SMULW
Signed Multiply (halfwords) and Signed Multiply (word by halfword).
Syntax
op{XY}{cond} Rd,Rn, Rm
op{Y}{cond} Rd. Rn, Rm

For SMULXY only:
op

Is one of:
SMUL{XY} Signed Multiply (halfwords)
X and Y specify which halfword of the source registers Rn and Rm is used as the first
and second multiply operand. If X is B, then the bottom halfword, bits [15:0] of Rn
is used. If X is T, then the top halfword, bits [31:16] of Rn is used. If Y is B, then the
bottom halfword, bits [15:0], of Rm is used. If Y is T, then the top halfword, bits
[31:16], of Rm is used.
SMULW{Y} Signed Multiply (word by halfword)

Y specifies which halfword of the source register Rm is used as the second multiply
operand. If Y is B, then the bottom halfword (bits [15:0]) of Rm is used. If Y is T,
then the top halfword (bits [31:16]) of Rm is used.
cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Rn, Rm

Are registers holding the first and second operands.

Operation
The SMULBB, SMULTB, SMULBT and SMULTT instructions interprets the values from Rn and Rm as four
signed 16-bit integers. These instructions:
•
Multiplies the specified signed halfword, Top or Bottom, values from Rn and Rm.
•
Writes the 32-bit result of the multiplication in Rd.
The SMULWT and SMULWB instructions interprets the values from Rn as a 32-bit signed integer and
Rm as two halfword 16-bit signed integers. These instructions:
•
Multiplies the first operand and the top, T suffix, or the bottom, B suffix, halfword of the
second operand.
•
Writes the signed most significant 32 bits of the 48-bit result in the destination register.
Restrictions
In these instructions:
•
Do not use SP and do not use PC.
•
RdHi and RdLo must be different registers.

ARM DUI 0553A
ID121610

3-91

The Cortex-M4 Instruction Set

Examples

ARM DUI 0553A
ID121610

SMULBT

R0, R4, R5

SMULBB

R0, R4, R5

SMULTT

R0, R4, R5

SMULTB

R0, R4, R5

SMULWT

R4, R5, R3

SMULWB

R4, R5, R3

;
;
;
;
;
;
;
;
;
;
;
;
;
;
;
;

Multiplies the bottom halfword of R4 with the
top halfword of R5, multiplies results and
writes to R0
Multiplies the bottom halfword of R4 with the
bottom halfword of R5, multiplies results and
writes to R0
Multiplies the top halfword of R4 with the top
halfword of R5, multiplies results and writes
to R0
Multiplies the top halfword of R4 with the
bottom halfword of R5, multiplies results and
and writes to R0
Multiplies R5 with the top halfword of R3,
extracts top 32 bits and writes to R4
Multiplies R5 with the bottom halfword of R3,
extracts top 32 bits and writes to R4.

3-92

The Cortex-M4 Instruction Set

3.6.11

UMULL, UMLAL, SMULL, and SMLAL
Signed and Unsigned Long Multiply, with optional Accumulate, using 32-bit operands and
producing a 64-bit result.
Syntax
op{cond} RdLo, RdHi, Rn, Rm

where:
Is one of:

UMULL

Unsigned Long Multiply.

UMLAL

Unsigned Long Multiply, with Accumulate.

SMULL

Signed Long Multiply.

SMLAL

Signed Long Multiply, with Accumulate.

cond

Is an optional condition code, see Conditional execution on page 3-18.

RdHi, RdLo

Are the destination registers. For UMLAL and SMLAL they also hold the accumulating
value.

Rn, Rm

Are registers holding the operands.

Operation
The UMULL instruction interprets the values from Rn and Rm as unsigned integers. It multiplies
these integers and places the least significant 32 bits of the result in RdLo, and the most
significant 32 bits of the result in RdHi.
The UMLAL instruction interprets the values from Rn and Rm as unsigned integers. It multiplies
these integers, adds the 64-bit result to the 64-bit unsigned integer contained in RdHi and RdLo,
and writes the result back to RdHi and RdLo.
The SMULL instruction interprets the values from Rn and Rm as two’s complement signed integers.
It multiplies these integers and places the least significant 32 bits of the result in RdLo, and the
most significant 32 bits of the result in RdHi.
The SMLAL instruction interprets the values from Rn and Rm as two’s complement signed integers.
It multiplies these integers, adds the 64-bit result to the 64-bit signed integer contained in RdHi
and RdLo, and writes the result back to RdHi and RdLo.
Restrictions
In these instructions:
•
do not use SP and do not use PC
•
RdHi and RdLo must be different registers.
Condition flags
These instructions do not affect the condition code flags.
Examples
UMULL
SMLAL

ARM DUI 0553A
ID121610

R0, R4, R5, R6
R4, R5, R3, R8

; Unsigned (R4,R0) = R5 x R6
; Signed (R5,R4) = (R5,R4) + R3 x R8

3-93

The Cortex-M4 Instruction Set

3.6.12

SDIV and UDIV
Signed Divide and Unsigned Divide.
Syntax
SDIV{cond} {Rd,} Rn, Rm
UDIV{cond} {Rd,} Rn, Rm

where:
cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register. If Rd is omitted, the destination register is Rn.

Specifies the register holding the value to be divided.

Is a register holding the divisor.

Operation
SDIV performs a signed integer division of the value in Rn by the value in Rm.
UDIV performs an unsigned integer division of the value in Rn by the value in Rm.

For both instructions, if the value in Rn is not divisible by the value in Rm, the result is rounded
towards zero.
Restrictions
Do not use SP and do not use PC.
Condition flags
These instructions do not change the flags.
Examples
SDIV
UDIV

ARM DUI 0553A
ID121610

R0, R2, R4
R8, R8, R1

; Signed divide, R0 = R2/R4
; Unsigned divide, R8 = R8/R1

3-94

The Cortex-M4 Instruction Set

3.7

Saturating instructions
Table 3-10 shows the saturating instructions:
Table 3-10 Saturating instructions
Mnemonic

Brief description

See

SSAT

Signed Saturate

SSAT and USAT on page 3-96

SSAT16

Signed Saturate Halfword

SSAT16 and USAT16 on page 3-97

USAT

Unsigned Saturate

SSAT and USAT on page 3-96

USAT16

Unsigned Saturate Halfword

SSAT16 and USAT16 on page 3-97

QADD

Saturating Add

QADD and QSUB on page 3-98

QSUB

Saturating Subtract

QADD and QSUB on page 3-98

QSUB16

Saturating Subtract 16

QADD and QSUB on page 3-98

QASX

Saturating Add and Subtract with Exchange

QASX and QSAX on page 3-100

QSAX

Saturating Subtract and Add with Exchange

QASX and QSAX on page 3-100

QDADD

Saturating Double and Add

QDADD and QDSUB on page 3-102

QDSUB

Saturating Double and Subtract

QDADD and QDSUB on page 3-102

UQADD16

Unsigned Saturating Add 16

UQADD and UQSUB on page 3-105

UQADD8

Unsigned Saturating Add 8

UQADD and UQSUB on page 3-105

UQASX

Unsigned Saturating Add and Subtract with Exchange

UQASX and UQSAX on page 3-103

UQSAX

Unsigned Saturating Subtract and Add with Exchange

UQASX and UQSAX on page 3-103

UQSUB16

Unsigned Saturating Subtract 16

UQADD and UQSUB on page 3-105

UQSUB8

Unsigned Saturating Subtract 8

UQADD and UQSUB on page 3-105

For signed n-bit saturation, this means that:
•
if the value to be saturated is less than −2n−1, the result returned is −2n-1
•
if the value to be saturated is greater than 2n−1−1, the result returned is 2n-1−1
•
otherwise, the result returned is the same as the value to be saturated.
For unsigned n-bit saturation, this means that:
•
if the value to be saturated is less than 0, the result returned is 0
•
if the value to be saturated is greater than 2n−1, the result returned is 2n−1
•
otherwise, the result returned is the same as the value to be saturated.
If the returned result is different from the value to be saturated, it is called saturation. If
saturation occurs, the instruction sets the Q flag to 1 in the APSR. Otherwise, it leaves the Q flag
unchanged. To clear the Q flag to 0, you must use the MSR instruction, see MSR on page 3-164.
To read the state of the Q flag, use the MRS instruction, see MRS on page 3-163.

ARM DUI 0553A
ID121610

3-95

The Cortex-M4 Instruction Set

3.7.1

SSAT and USAT
Signed Saturate and Unsigned Saturate to any bit position, with optional shift before saturating.
Syntax
op{cond} Rd, #n, Rm {, shift #s}

where:
Is one of:

SSAT
USAT

Saturates a signed value to a signed range.
Saturates a signed value to an unsigned range.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Specifies the bit position to saturate to:
•
n ranges from 1 to 32 for SSAT
•
n ranges from 0 to 31 for USAT.

Specifies the register containing the value to saturate.

shift #s

Is an optional shift applied to Rm before saturating. It must be one of the following:
ASR #s
where s is in the range 1 to 31
LSL #s
where s is in the range 0 to 31.

Operation
These instructions saturate to a signed or unsigned n-bit value.
The SSAT instruction applies the specified shift, then saturates to the signed range
−2n–1 ≤ x ≤ 2n–1−1.
The USAT instruction applies the specified shift, then saturates to the unsigned range
0 ≤ x ≤ 2n−1.
Restrictions
Do not use SP and do not use PC.
Condition flags
These instructions do not affect the condition code flags.
If saturation occurs, these instructions set the Q flag to 1.
Examples

ARM DUI 0553A
ID121610

SSAT

R7, #16, R7, LSL #4

USATNE

R0, #7, R5

;
;
;
;
;

Logical shift left value in R7 by 4, then
saturate it as a signed 16-bit value and
write it back to R7
Conditionally saturate value in R5 as an
unsigned 7 bit value and write it to R0.

3-96

The Cortex-M4 Instruction Set

3.7.2

SSAT16 and USAT16
Signed Saturate and Unsigned Saturate to any bit position for two halfwords.
Syntax
op{cond} Rd, #n, Rm

where:
Is one of:

SSAT16
USAT16

Saturates a signed halfword value to a signed range.
Saturates a signed halfword value to an unsigned range.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Specifies the bit position to saturate to:
•
n ranges from 1 to 16 for SSAT.
•
n ranges from 0 to 15 for USAT.

Specifies the register containing the value to saturate.

Operation
The SSAT16 instruction:
1.
Saturates two signed 16-bit halfword values of the register with the value to saturate from
selected by the bit position in n.
2.
Writes the results as two signed 16-bit halfwords to the destination register.
The USAT16 instruction:
1.
Saturates two unsigned 16-bit halfword values of the register with the value to saturate
from selected by the bit position in n.
2.
Writes the results as two unsigned halfwords in the destination register.
Restrictions
Do not use SP and do not use PC.
Condition flags
These instructions do not affect the condition code flags.
If saturation occurs, these instructions set the Q flag to 1.
Examples

ARM DUI 0553A
ID121610

SSAT16

R7, #9, R2

USAT16NE

R0, #13, R5

;
;
;
;
;
;

Saturates the top and bottom highwords of R2
as 9-bit values, writes to corresponding halfword
of R7
Conditionally saturates the top and bottom
halfwords of R5 as 13-bit values, writes to
corresponding halfword of R0.

3-97

The Cortex-M4 Instruction Set

3.7.3

QADD and QSUB
Saturating Add and Saturating Subtract, signed.
Syntax
op{cond} {Rd}, Rn, Rm
op{cond} {Rd}, Rn, Rm

where:
op

Is one of:
QADD

Saturating 32-bit add.

QADD8

Saturating four 8-bit integer additions.

QADD16

Saturating two 16-bit integer additions.

QSUB

Saturating 32-bit subtraction.

QSUB8

Saturating four 8-bit integer subtraction.

QSUB16

Saturating two 16-bit integer subtraction.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Rn, Rm

Are registers holding the first and second operands.

Operation
These instructions add or subtract two, four or eight values from the first and second operands
and then writes a signed saturated value in the destination register.
The QADD and QSUB instructions apply the specified add or subtract, and then saturate the result to
the signed range −2n–1 ≤ x ≤ 2n–1−1, where x is given by the number of bits applied in the
instruction, 32, 16 or 8.
If the returned result is different from the value to be saturated, it is called saturation. If
saturation occurs, the QADD and QSUB instructions set the Q flag to 1 in the APSR. Otherwise, it
leaves the Q flag unchanged. The 8-bit and 16-bit QADD and QSUB instructions always leave the Q
flag unchanged.
To clear the Q flag to 0, you must use the MSR instruction, see MSR on page 3-164.
To read the state of the Q flag, use the MRS instruction, see MRS on page 3-163.
Restrictions
Do not use SP and do not use PC.
Condition flags
These instructions do not affect the condition code flags.
If saturation occurs, these instructions set the Q flag to 1.

ARM DUI 0553A
ID121610

3-98

The Cortex-M4 Instruction Set

Examples

ARM DUI 0553A
ID121610

QADD16

R7, R4, R2

QADD8

R3, R1, R6

QSUB16

R4, R2, R3

QSUB8

R4, R2, R5

;
;
;
;
;
;
;
;
;
;
;
;

Adds halfwords of R4 with corresponding halfword of
R2, saturates to 16 bits and writes to corresponding
halfword of R7
Adds bytes of R1 to the corresponding bytes of R6,
saturates to 8 bits and writes to corresponding byte of
R3
Subtracts halfwords of R3 from corresponding halfword
of R2, saturates to 16 bits, writes to corresponding
halfword of R4
Subtracts bytes of R5 from the corresponding byte in
R2, saturates to 8 bits, writes to corresponding byte of
R4.

3-99

The Cortex-M4 Instruction Set

3.7.4

QASX and QSAX
Saturating Add and Subtract with Exchange and Saturating Subtract and Add with Exchange,
signed.
Syntax
op{cond} {Rd}, Rm, Rn

where:
op

Is one of:
QASX
QSAX

Add and Subtract with Exchange and Saturate.
Subtract and Add with Exchange and Saturate.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Rn, Rm

Are registers holding the first and second operands.

Operation
The QASX instruction:
1.
Adds the top halfword of the source operand with the bottom halfword of the second
operand.
2.
Subtracts the top halfword of the second operand from the bottom highword of the first
operand.
3.
Saturates the result of the subtraction and writes a 16-bit signed integer in the range –215
≤ x ≤ 215 – 1, where x equals 16, to the bottom halfword of the destination register.
4.
Saturates the results of the sum and writes a 16-bit signed integer in the range –215 ≤ x ≤
215 – 1, where x equals 16, to the top halfword of the destination register.
The QSAX instruction:
1.
Subtracts the bottom halfword of the second operand from the top highword of the first
operand.
2.
Adds the bottom halfword of the source operand with the top halfword of the second
operand.
3.
Saturates the results of the sum and writes a 16-bit signed integer in the range –215 ≤ x ≤
215 – 1, where x equals 16, to the bottom halfword of the destination register.
4.
Saturates the result of the subtraction and writes a 16-bit signed integer in the range –215
≤ x ≤ 215 – 1, where x equals 16, to the top halfword of the destination register.
Restrictions
Do not use SP and do not use PC.
Condition flags
These instructions do not affect the condition code flags.

ARM DUI 0553A
ID121610

3-100

The Cortex-M4 Instruction Set

Examples

ARM DUI 0553A
ID121610

QASX

R7, R4, R2

QSAX

R0, R3, R5

;
;
;
;
;
;
;
;
;

Adds top halfword of R4 to bottom halfword of R2,
saturates to 16 bits, writes to top halfword of R7
Subtracts top highword of R2 from bottom halfword of
R4, saturates to 16 bits and writes to bottom halfword
of R7
Subtracts bottom halfword of R5 from top halfword of
R3, saturates to 16 bits, writes to top halfword of R0
Adds bottom halfword of R3 to top halfword of R5,
saturates to 16 bits, writes to bottom halfword of R0.

3-101

The Cortex-M4 Instruction Set

3.7.5

QDADD and QDSUB
Saturating Double and Add and Saturating Double and Subtract, signed.
Syntax
op{cond} {Rd}, Rm, Rn

where:
Is one of:

QDADD
QDSUB

Saturating Double and Add.
Saturating Double and Subtract.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Rm, Rn

Are registers holding the first and second operands.

Operation
The QDADD instruction:
•
Doubles the second operand value.
•
Adds the result of the doubling to the signed saturated value in the first operand.
•
Writes the result to the destination register.
The QDSUB instruction:
•
Doubles the second operand value.
•
Subtracts the doubled value from the signed saturated value in the first operand.
•
Writes the result to the destination register.
Both the doubling and the addition or subtraction have their results saturated to the 32-bit signed
integer range –231 ≤ x ≤ 231– 1. If saturation occurs in either operation, it sets the Q flag in the
APSR.
Restrictions
Do not use SP and do not use PC.
Condition flags
If saturation occurs, these instructions set the Q flag to 1.
Examples

ARM DUI 0553A
ID121610

QDADD

R7, R4, R2

QDSUB

R0, R3, R5

;
;
;
;

Doubles and saturates R4 to 32 bits, adds R2,
saturates to 32 bits, writes to R7
Subtracts R3 doubled and saturated to 32 bits
from R5, saturates to 32 bits, writes to R0.

3-102

The Cortex-M4 Instruction Set

3.7.6

UQASX and UQSAX
Saturating Add and Subtract with Exchange and Saturating Subtract and Add with Exchange,
unsigned.
Syntax
op{cond} {Rd}, Rm, Rn

where:
type

Is one of:
UQASX
UQSAX

Add and Subtract with Exchange and Saturate.
Subtract and Add with Exchange and Saturate.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Rn, Rm

Are registers holding the first and second operands.

Operation
The UQASX instruction:
1.
Adds the bottom halfword of the source operand with the top halfword of the second
operand.
2.
Subtracts the bottom halfword of the second operand from the top highword of the first
operand.
3.
Saturates the results of the sum and writes a 16-bit unsigned integer in the range 0 ≤ x ≤
216 – 1, where x equals 16, to the top halfword of the destination register.
4.
Saturates the result of the subtraction and writes a 16-bit unsigned integer in the range 0
≤ x ≤ 216 – 1, where x equals 16, to the bottom halfword of the destination register.
The UQSAX instruction:
1.
Subtracts the bottom halfword of the second operand from the top highword of the first
operand.
2.
Adds the bottom halfword of the first operand with the top halfword of the second
operand.
3.
Saturates the result of the subtraction and writes a 16-bit unsigned integer in the range 0
≤ x ≤ 216 – 1, where x equals 16, to the top halfword of the destination register.
4.
Saturates the results of the addition and writes a 16-bit unsigned integer in the range 0 ≤
x ≤ 216 – 1, where x equals 16, to the bottom halfword of the destination register.
Restrictions
Do not use SP and do not use PC.
Condition flags
These instructions do not affect the condition code flags.

ARM DUI 0553A
ID121610

3-103

The Cortex-M4 Instruction Set

Examples

ARM DUI 0553A
ID121610

UQASX

R7, R4, R2

UQSAX

R0, R3, R5

;
;
;
;
;
;
;
;

Adds top halfword of R4 with bottom halfword of R2,
saturates to 16 bits, writes to top halfword of R7
Subtracts top halfword of R2 from bottom halfword of
R4, saturates to 16 bits, writes to bottom halfword of R7
Subtracts bottom halfword of R5 from top halfword of R3,
saturates to 16 bits, writes to top halfword of R0
Adds bottom halfword of R4 to top halfword of R5
saturates to 16 bits, writes to bottom halfword of R0.

3-104

The Cortex-M4 Instruction Set

3.7.7

UQADD and UQSUB
Saturating Add and Saturating Subtract Unsigned.
Syntax
op{cond} {Rd}, Rn, Rm
op{cond} {Rd}, Rn, Rm

where:
op

Is one of:
UQADD8

Saturating four unsigned 8-bit integer additions.

UQADD16

Saturating two unsigned 16-bit integer additions.

UDSUB8

Saturating four unsigned 8-bit integer subtractions.

UQSUB16

Saturating two unsigned 16-bit integer subtractions.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Rn, Rm

Are registers holding the first and second operands.

Operation
These instructions add or subtract two or four values and then writes an unsigned saturated value
in the destination register.
The UQADD16 instruction:
•
Adds the respective top and bottom halfwords of the first and second operands.
•
Saturates the result of the additions for each halfword in the destination register to the
unsigned range 0 ≤ x ≤ 216−1, where x is 16.
The UQADD8 instruction:
•
Adds each respective byte of the first and second operands.
•
Saturates the result of the addition for each byte in the destination register to the unsigned
range 0 ≤ x ≤ 28−1, where x is 8.
The UQSUB16 instruction:
•
Subtracts both halfwords of the second operand from the respective halfwords of the first
operand.
•
Saturates the result of the differences in the destination register to the unsigned range
0 ≤ x ≤ 216−1, where x is 16.
The UQSUB8 instructions:
•
Subtracts the respective bytes of the second operand from the respective bytes of the first
operand.
•
Saturates the results of the differences for each byte in the destination register to the
unsigned range 0 ≤ x ≤ 28−1, where x is 8.
Restrictions
Do not use SP and do not use PC.

ARM DUI 0553A
ID121610

3-105

The Cortex-M4 Instruction Set

Condition flags
These instructions do not affect the condition code flags.
Examples

ARM DUI 0553A
ID121610

UQADD16

R7, R4, R2

UQADD8

R4, R2, R5

UQSUB16

R6, R3, R0

UQSUB8

R1, R5, R6

;
;
;
;
;
;
;
;
;
;

Adds halfwords in R4 to corresponding halfword in R2,
saturates to 16 bits, writes to corresponding halfword
of R7
Adds bytes of R2 to corresponding byte of R5, saturates
to 8 bits, writes to corresponding bytes of R4
Subtracts halfwords in R0 from corresponding halfword
in R3, saturates to 16 bits, writes to corresponding
halfword in R6
Subtracts bytes in R6 from corresponding byte of R5,
saturates to 8 bits, writes to corresponding byte of R1.

3-106

The Cortex-M4 Instruction Set

3.8

Packing and unpacking instructions
Table 3-11shows the instructions that operate on packing and unpacking data:
Table 3-11 Packing and unpacking instructions

ARM DUI 0553A
ID121610

Mnemonic

Brief description

See

PKH

Pack Halfword

PKHBT and PKHTB on page 3-108

SXTAB

Extend 8 bits to 32 and add

SXTA and UXTA on page 3-112

SXTAB16

Dual extend 8 bits to 16 and add

SXTA and UXTA on page 3-112

SXTAH

Extend 16 bits to 32 and add

SXTA and UXTA on page 3-112

SXTB

Sign extend a byte

SXT and UXT on page 3-117

SXTB16

Dual extend 8 bits to 16 and add

SXT and UXT on page 3-117

SXTH

Sign extend a halfword

SXT and UXT on page 3-117

UXTAB

Extend 8 bits to 32 and add

SXTA and UXTA on page 3-112

UXTAB16

Dual extend 8 bits to 16 and add

SXTA and UXTA on page 3-112

UXTAH

Extend 16 bits to 32 and add

SXTA and UXTA on page 3-112

UXTB

Zero extend a byte

SXT and UXT on page 3-117

UXTB16

Dual zero extend 8 bits to 16 and add

SXT and UXT on page 3-117

UXTH

Zero extend a halfword

SXT and UXT on page 3-117

3-107

The Cortex-M4 Instruction Set

3.8.1

PKHBT and PKHTB
Pack Halfword.
Syntax
op{cond} {Rd}, Rn, Rm {, LSL #imm}
op{cond} {Rd}, Rn, Rm {, ASR #imm}

where:
Is one of:

PKHBT

Pack Halfword, bottom and top with shift.

PKHTB

Pack Halfword, top and bottom with shift.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Specifies the first operand register.

Specifies the second operand register holding the value to be optionally shifted.

imm

Specifies the shift length. The type of shift length depends on the instruction:
For PKHBT:
LSL

A left shift with a shift length from 1 to 31, 0 means no shift.

For PKHTB:
ASR

An arithmetic shift right with a shift length from 1 to 32, a shift of
32-bits is encoded as 0b00000.

Operation
The PKHBT instruction:

1.
2.

Writes the value of the bottom halfword of the first operand to the bottom halfword of the
destination register.
If shifted, the shifted value of the second operand is written to the top halfword of the
destination register.

The PKHTB instruction:
1.
Writes the value of the top halfword of the first operand to the top halfword of the
destination register.
2.
If shifted, the shifted value of the second operand is written to the bottom halfword of the
destination register.
Restrictions
Rd must not be SP and must not be PC.

Condition flags
This instruction does not change the flags.

ARM DUI 0553A
ID121610

3-108

The Cortex-M4 Instruction Set

Examples

ARM DUI 0553A
ID121610

PKHBT

R3, R4, R5 LSL #0

PKHTB

R4, R0, R2 ASR #1

;
;
;
;
;
;

Writes bottom halfword of R4 to bottom halfword of
R3, writes top halfword of R5, unshifted, to top
halfword of R3
Writes R2 shifted right by 1 bit to bottom halfword
of R4, and writes top halfword of R0 to top
halfword of R4.

3-109

The Cortex-M4 Instruction Set

3.8.2

SXT and UXT
Sign extend and Zero extend.
Syntax
op{cond} {Rd,} Rm {, ROR #n}
op{cond} {Rd}, Rm {, ROR #n}

where:
Is one of:

SXTB
SXTH
SXTB16
UXTB
UXTH
UXTB16

Sign extends an 8-bit value to a 32-bit value.
Sign extends a 16-bit value to a 32-bit value.
Sign extends two 8-bit values to two 16-bit values.
Zero extends an 8-bit value to a 32-bit value.
Zero extends a 16-bit value to a 32-bit value.
Zero extends two 8-bit values to two 16-bit values.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Specifies the register holding the value to extend.

ROR #n

Is one of:
ROR #8
ROR #16
ROR #24

Value from Rm is rotated right 8 bits.
Value from Rm is rotated right 16 bits.
Value from Rm is rotated right 24 bits.

If ROR #n is omitted, no rotation is performed.
Operation
These instructions do the following:
1.

Rotate the value from Rm right by 0, 8, 16 or 24 bits.

Extract bits from the resulting value:
•
SXTB extracts bits[7:0] and sign extends to 32 bits.
•
UXTB extracts bits[7:0] and zero extends to 32 bits.
•
SXTH extracts bits[15:0] and sign extends to 32 bits.
•
UXTH extracts bits[15:0] and zero extends to 32 bits.
•
SXTB16 extracts bits[7:0] and sign extends to 16 bits, and extracts bits [23:16] and
sign extends to 16 bits.
•
UXTB16 extracts bits[7:0] and zero extends to 16 bits, and extracts bits [23:16] and
zero extends to 16 bits.

Restrictions
Do not use SP and do not use PC.
Condition flags
These instructions do not affect the flags.

ARM DUI 0553A
ID121610

3-110

The Cortex-M4 Instruction Set

Examples

ARM DUI 0553A
ID121610

SXTH

R4, R6, ROR #16

UXTB

R3, R10

;
;
;
;

Rotates R6 right by 16 bits, obtains bottom halfword of
of result, sign extends to 32 bits and writes to R4
Extracts lowest byte of value in R10, zero extends, and
writes to R3.

3-111

The Cortex-M4 Instruction Set

3.8.3

SXTA and UXTA
Signed and Unsigned Extend and Add.
Syntax
op{cond} {Rd,} Rn, Rm {, ROR #n}
op{cond} {Rd,} Rn, Rm {, ROR #n}

where:
Is one of:

SXTAB
SXTAH
SXTAB16
UXTAB
UXTAH
UXTAB16

Sign extends an 8-bit value to a 32-bit value and add.
Sign extends a 16-bit value to a 32-bit value and add.
Sign extends two 8-bit values to two 16-bit values and add.
Zero extends an 8-bit value to a 32-bit value and add.
Zero extends a 16-bit value to a 32-bit value and add.
Zero extends two 8-bit values to two 16-bit values and add.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Specifies the first operand register.

Specifies the register holding the value to rotate and extend.

ROR #n

Is one of:
ROR #8
ROR #16
ROR #24

Value from Rm is rotated right 8 bits.
Value from Rm is rotated right 16 bits.
Value from Rm is rotated right 24 bits.

If ROR #n is omitted, no rotation is performed.
Operation
These instructions do the following:
1.

Rotate the value from Rm right by 0, 8, 16 or 24 bits.

Extract bits from the resulting value:
•
SXTAB extracts bits[7:0] from Rm and sign extends to 32 bits.
•
UXTAB extracts bits[7:0] from Rm and zero extends to 32 bits.
•
SXTAH extracts bits[15:0] from Rm and sign extends to 32 bits.
•
UXTAH extracts bits[15:0] from Rm and zero extends to 32 bits.
•
SXTAB16 extracts bits[7:0] from Rm and sign extends to 16 bits, and extracts bits
[23:16] from Rm and sign extends to 16 bits.
•
UXTAB16 extracts bits[7:0] from Rm and zero extends to 16 bits, and extracts bits
[23:16] from Rm and zero extends to 16 bits.

Adds the signed or zero extended value to the word or corresponding halfword of Rn and
writes the result in Rd.

Restrictions
Do not use SP and do not use PC.

ARM DUI 0553A
ID121610

3-112

The Cortex-M4 Instruction Set

Condition flags
These instructions do not affect the flags.
Examples
SXTAH

UXTAB

ARM DUI 0553A
ID121610

R4, R8, R6, ROR #16 ;
;
;
R3, R4, R10
;
;

Rotates R6 right by 16 bits, obtains bottom
halfword, sign extends to 32 bits, adds R8, and
writes to R4
Extracts bottom byte of R10 and zero extends to 32
bits, adds R4, and writes to R3.

3-113

The Cortex-M4 Instruction Set

3.9

Bitfield instructions
Table 3-12 shows the instructions that operate on adjacent sets of bits in registers or bitfields:
Table 3-12 Packing and unpacking instructions

ARM DUI 0553A
ID121610

Mnemonic

Brief description

See

BFC

Bit Field Clear

BFC and BFI on page 3-115

BFI

Bit Field Insert

BFC and BFI on page 3-115

SBFX

Signed Bit Field Extract

SBFX and UBFX on page 3-116

SXTB

Sign extend a byte

SXT and UXT on page 3-117

SXTH

Sign extend a halfword

SXT and UXT on page 3-117

UBFX

Unsigned Bit Field Extract

SBFX and UBFX on page 3-116

UXTB

Zero extend a byte

SXT and UXT on page 3-117

UXTH

Zero extend a halfword

SXT and UXT on page 3-117

3-114

The Cortex-M4 Instruction Set

3.9.1

BFC and BFI
Bit Field Clear and Bit Field Insert.
Syntax
BFC{cond} Rd, #lsb, #width
BFI{cond} Rd, Rn, #lsb, #width

where:
cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Specifies the source register.

lsb

Specifies the position of the least significant bit of the bitfield. lsb must be in the
range 0 to 31.

width

Specifies the width of the bitfield and must be in the range 1 to 32−lsb.

Operation
BFC clears a bitfield in a register. It clears width bits in Rd, starting at the low bit position lsb.
Other bits in Rd are unchanged.
BFI copies a bitfield into one register from another register. It replaces width bits in Rd starting
at the low bit position lsb, with width bits from Rn starting at bit[0]. Other bits in Rd are

unchanged.
Restrictions
Do not use SP and do not use PC.
Condition flags
These instructions do not affect the flags.
Examples
BFC
BFI

ARM DUI 0553A
ID121610

R4, #8, #12
R9, R2, #8, #12

; Clear bit 8 to bit 19 (12 bits) of R4 to 0
; Replace bit 8 to bit 19 (12 bits) of R9 with
; bit 0 to bit 11 from R2.

3-115

The Cortex-M4 Instruction Set

3.9.2

SBFX and UBFX
Signed Bit Field Extract and Unsigned Bit Field Extract.
Syntax
SBFX{cond} Rd, Rn, #lsb, #width
UBFX{cond} Rd, Rn, #lsb, #width

where:
cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Specifies the source register.

lsb

Specifies the position of the least significant bit of the bitfield. lsb must be in the
range 0 to 31.

width

Specifies the width of the bitfield and must be in the range 1 to 32−lsb.

Operation
SBFX extracts a bitfield from one register, sign extends it to 32 bits, and writes the result to the

destination register.
UBFX extracts a bitfield from one register, zero extends it to 32 bits, and writes the result to the

destination register.
Restrictions
Do not use SP and do not use PC.
Condition flags
These instructions do not affect the flags.
Examples
SBFX
UBFX

ARM DUI 0553A
ID121610

R0, R1, #20, #4

;
;
R8, R11, #9, #10 ;
;

Extract bit 20 to bit 23 (4 bits) from R1 and sign
extend to 32 bits and then write the result to R0.
Extract bit 9 to bit 18 (10 bits) from R11 and zero
extend to 32 bits and then write the result to R8.

3-116

The Cortex-M4 Instruction Set

3.9.3

SXT and UXT
Sign extend and Zero extend.
Syntax
SXTextend{cond} {Rd,} Rm {, ROR #n}
UXTextend{cond} {Rd}, Rm {, ROR #n}

where:
Is one of:

extend

B
H

Extends an 8-bit value to a 32-bit value.
Extends a 16-bit value to a 32-bit value.

cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

Specifies the register holding the value to extend.

ROR #n

Is one of:
ROR #8
ROR #16
ROR #24

Value from Rm is rotated right 8 bits.
Value from Rm is rotated right 16 bits.
Value from Rm is rotated right 24 bits.

If ROR #n is omitted, no rotation is performed.
Operation
These instructions do the following:
1.

Rotate the value from Rm right by 0, 8, 16 or 24 bits.

Restrictions
Do not use SP and do not use PC.
Condition flags
These instructions do not affect the flags.
Examples

ARM DUI 0553A
ID121610

SXTH

R4, R6, ROR #16

UXTB

R3, R10

;
;
;
;
;

Rotate R6 right by 16 bits, then obtain the lower
halfword of the result and then sign extend to
32 bits and write the result to R4.
Extract lowest byte of the value in R10 and zero
extend it, and write the result to R3.

3-117

The Cortex-M4 Instruction Set

3.10

Branch and control instructions
Table 3-13 shows the branch and control instructions:
Table 3-13 Branch and control instructions

ARM DUI 0553A
ID121610

Mnemonic

Brief description

See

Branch

B, BL, BX, and BLX on page 3-119

Branch with Link

B, BL, BX, and BLX on page 3-119

BLX

Branch indirect with Link

B, BL, BX, and BLX on page 3-119

Branch indirect

B, BL, BX, and BLX on page 3-119

CBNZ

Compare and Branch if Non Zero

CBZ and CBNZ on page 3-121

CBZ

Compare and Branch if Zero

CBZ and CBNZ on page 3-121

If-Then

IT on page 3-122

TBB

Table Branch Byte

TBB and TBH on page 3-124

TBH

Table Branch Halfword

TBB and TBH on page 3-124

3-118

The Cortex-M4 Instruction Set

3.10.1

B, BL, BX, and BLX
Branch instructions.
Syntax
B{cond} label
BL{cond} label
BX{cond} Rm
BLX{cond} Rm

where:
B

Is branch (immediate).

Is branch with link (immediate).

Is branch indirect (register).

BLX

Is branch indirect with link (register).

cond

Is an optional condition code, see Conditional execution on page 3-18.

label

Is a PC-relative expression. See PC-relative expressions on page 3-17.

Is a register that indicates an address to branch to. Bit[0] of the value in Rm must
be 1, but the address to branch to is created by changing bit[0] to 0.

Operation
All these instructions cause a branch to label, or to the address indicated in Rm. In addition:
•

The BL and BLX instructions write the address of the next instruction to LR (the link
register, R14).

•

The BX and BLX instructions result in a UsageFault exception if bit[0] of Rm is 0.

Bcond label is the only conditional instruction that can be either inside or outside an IT block.
All other branch instructions must be conditional inside an IT block, and must be unconditional
outside the IT block, see IT on page 3-122.

Table 3-14 shows the ranges for the various branch instructions.
Table 3-14 Branch ranges

ARM DUI 0553A
ID121610

Instruction

Branch range

B label

−16 MB to +16 MB

Bcond label (outside IT block)

−1 MB to +1 MB

Bcond label (inside IT block)

−16 MB to +16 MB

BL{cond} label

−16 MB to +16 MB

BX{cond} Rm

Any value in register

BLX{cond} Rm

Any value in register

3-119

The Cortex-M4 Instruction Set

Note
You might have to use the .W suffix to get the maximum branch range. See Instruction width
selection on page 3-21.

Restrictions
The restrictions are:
•

do not use PC in the BLX instruction

•

for BX and BLX, bit[0] of Rm must be 1 for correct execution but a branch occurs to the target
address created by changing bit[0] to 0

•

when any of these instructions is inside an IT block, it must be the last instruction of the
IT block.

Note
Bcond is the only conditional instruction that is not required to be inside an IT block. However,
it has a longer branch range when it is inside an IT block.

Condition flags
These instructions do not change the flags.
Examples

ARM DUI 0553A
ID121610

B
BLE
B.W
BEQ
BEQ.W
BL

loopA
ng
target
target
target
funC

BX
BXNE
BLX

LR
R0
R0

;
;
;
;
;
;
;
;
;
;
;

Branch to loopA
Conditionally branch to label ng
Branch to target within 16MB range
Conditionally branch to target
Conditionally branch to target within 1MB
Branch with link (Call) to function funC, return address
stored in LR
Return from function call
Conditionally branch to address stored in R0
Branch with link and exchange (Call) to a address stored
in R0.

3-120

The Cortex-M4 Instruction Set

3.10.2

CBZ and CBNZ
Compare and Branch on Zero, Compare and Branch on Non-Zero.
Syntax
CBZ Rn, label
CBNZ Rn, label

where:
Rn
label

Specifies the register holding the operand.
Specifies the branch destination.

Operation
Use the CBZ or CBNZ instructions to avoid changing the condition code flags and to reduce the
number of instructions.
CBZ Rn, label does not change condition flags but is otherwise equivalent to:
CMP
BEQ

Rn, #0
label

CBNZ Rn, label does not change condition flags but is otherwise equivalent to:
CMP
BNE

Rn, #0
label

Restrictions
The restrictions are:
•
Rn must be in the range of R0 to R7
•
the branch destination must be within 4 to 130 bytes after the instruction
•
these instructions must not be used inside an IT block.
Condition flags
These instructions do not change the flags.
Examples
CBZ
CBNZ

ARM DUI 0553A
ID121610

R5, target ; Forward branch if R5 is zero
R0, target ; Forward branch if R0 is not zero

3-121

The Cortex-M4 Instruction Set

3.10.3

IT
If-Then condition instruction.
Syntax
IT{x{y{z}}} cond

where:
x
y
z
cond

Specifies the condition switch for the second instruction in the IT block.
Specifies the condition switch for the third instruction in the IT block.
Specifies the condition switch for the fourth instruction in the IT block.
Specifies the condition for the first instruction in the IT block.

The condition switch for the second, third and fourth instruction in the IT block can be either:
T
Then. Applies the condition cond to the instruction.
E
Else. Applies the inverse condition of cond to the instruction.
Note
It is possible to use AL (the always condition) for cond in an IT instruction. If this is done, all of
the instructions in the IT block must be unconditional, and each of x, y, and z must be T or
omitted but not E.

Operation
The IT instruction makes up to four following instructions conditional. The conditions can be
all the same, or some of them can be the logical inverse of the others. The conditional
instructions following the IT instruction form the IT block.
The instructions in the IT block, including any branches, must specify the condition in the
{cond} part of their syntax.

Note
Your assembler might be able to generate the required IT instructions for conditional
instructions automatically, so that you do not have to write them yourself. See your assembler
documentation for details.
A BKPT instruction in an IT block is always executed, even if its condition fails.
Exceptions can be taken between an IT instruction and the corresponding IT block, or within an
IT block. Such an exception results in entry to the appropriate exception handler, with suitable
return information in LR and stacked PSR.
Instructions designed for use for exception returns can be used as normal to return from the
exception, and execution of the IT block resumes correctly. This is the only way that a
PC-modifying instruction is permitted to branch to an instruction in an IT block.
Restrictions
The following instructions are not permitted in an IT block:
•
IT
•
CBZ and CBNZ
•
CPSID and CPSIE.

ARM DUI 0553A
ID121610

3-122

The Cortex-M4 Instruction Set

Other restrictions when using an IT block are:
•

a branch or any instruction that modifies the PC must either be outside an IT block or must
be the last instruction inside the IT block. These are:
— ADD PC, PC, Rm
— MOV PC, Rm
— B, BL, BX, BLX
— any LDM, LDR, or POP instruction that writes to the PC
— TBB and TBH

•

do not branch to any instruction inside an IT block, except when returning from an
exception handler

•

all conditional instructions except Bcond must be inside an IT block. Bcond can be either
outside or inside an IT block but has a larger branch range if it is inside one

•

each instruction inside the IT block must specify a condition code suffix that is either the
same or logical inverse as for the other instructions in the block.

Note
Your assembler might place extra restrictions on the use of IT blocks, such as prohibiting the
use of assembler directives within them.

Condition flags
This instruction does not change the flags.
Example

ARM DUI 0553A
ID121610

ITTE
ANDNE
ADDSNE
MOVEQ

NE
R0, R0, R1
R2, R2, #1
R2, R3

;
;
;
;

Next 3 instructions are conditional
ANDNE does not update condition flags
ADDSNE updates condition flags
Conditional move

CMP

R0, #9

ITE
ADDGT
ADDLE

GT
R1, R0, #55
R1, R0, #48

;
;
;
;
;

Convert R0 hex value (0 to 15) into ASCII
('0'-'9', 'A'-'F')
Next 2 instructions are conditional
Convert 0xA -> 'A'
Convert 0x0 -> '0'

IT
ADDGT

GT
R1, R1, #1

; IT block with only one conditional instruction
; Increment R1 conditionally

ITTEE
MOVEQ
ADDEQ
ANDNE
BNE.W

EQ
R0, R1
R2, R2, #10
R3, R3, #1
dloop

;
;
;
;
;
;

IT
ADD

NE
R0, R0, R1

; Next instruction is conditional
; Syntax error: no condition code used in IT block

Next 4 instructions are conditional
Conditional move
Conditional add
Conditional AND
Branch instruction can only be used in the last
instruction of an IT block

3-123

The Cortex-M4 Instruction Set

3.10.4

TBB and TBH
Table Branch Byte and Table Branch Halfword.
Syntax
TBB [Rn, Rm]
TBH [Rn, Rm, LSL #1]

where:
Rn

Specifies the register containing the address of the table of branch lengths.
If Rn is PC, then the address of the table is the address of the byte immediately
following the TBB or TBH instruction.

Specifies the index register. This contains an index into the table. For halfword
tables, LSL #1 doubles the value in Rm to form the right offset into the table.

Operation
These instructions cause a PC-relative forward branch using a table of single byte offsets for TBB,
or halfword offsets for TBH. Rn provides a pointer to the table, and Rm supplies an index into the
table. For TBB the branch offset is twice the unsigned value of the byte returned from the table.
and for TBH the branch offset is twice the unsigned value of the halfword returned from the table.
The branch occurs to the address at that offset from the address of the byte immediately after
the TBB or TBH instruction.
Restrictions
The restrictions are:
Rn must not be SP
•
•
Rm must not be SP and must not be PC
•
when any of these instructions is used inside an IT block, it must be the last instruction of
the IT block.
Condition flags
These instructions do not change the flags.

ARM DUI 0553A
ID121610

3-124

The Cortex-M4 Instruction Set

Examples
ADR.W R0, BranchTable_Byte
TBB
[R0, R1]
; R1 is the index, R0 is the base address of the
; branch table
Case1
; an instruction sequence follows
Case2
; an instruction sequence follows
Case3
; an instruction sequence follows
BranchTable_Byte
DCB
0
; Case1 offset calculation
DCB
((Case2-Case1)/2) ; Case2 offset calculation
DCB
((Case3-Case1)/2) ; Case3 offset calculation

TBH

[PC, R1, LSL #1]

; R1 is the index, PC is used as base of the
; branch table

BranchTable_H
DCI
((CaseA - BranchTable_H)/2)
DCI
((CaseB - BranchTable_H)/2)
DCI
((CaseC - BranchTable_H)/2)

; CaseA offset calculation
; CaseB offset calculation
; CaseC offset calculation

CaseA
; an instruction sequence follows
CaseB
; an instruction sequence follows
CaseC
; an instruction sequence follows

ARM DUI 0553A
ID121610

3-125

The Cortex-M4 Instruction Set

3.11

Floating-point instructions
Table 3-15 shows the floating-point instructions.
Note
These instructions are only available if the FPU is included, and enabled, in the system. See
Enabling the FPU on page 4-52 for information about enabling the floating-point unit.

Table 3-15 Floating-point instructions
Mnemonic

Brief description

See

VABS

Floating-point Absolute

VABS on page 3-128

VADD

Floating-point Add

VADD on page 3-129

VCMP

Compare two floating-point registers, or one floating-point
register and zero

VCMP, VCMPE on page 3-130

VCMPE

Compare two floating-point registers, or one floating-point
register and zero with Invalid Operation check

VCMP, VCMPE on page 3-130

VCVT

Convert between floating-point and integer

VCVT, VCVTR between floating-point and integer on
page 3-131

VCVT

Convert between floating-point and fixed point

VCVT between floating-point and fixed-point on
page 3-132

VCVTR

Convert between floating-point and integer with rounding

VCVT, VCVTR between floating-point and integer on
page 3-131

VCVTB

Converts half-precision value to single-precision

VCVTB, VCVTT on page 3-133

VCVTT

Converts single-precision register to half-precision

VCVTB, VCVTT on page 3-133

VDIV

Floating-point Divide

VDIV on page 3-134

VFMA

Floating-point Fused Multiply Accumulate

VFMA, VFMS on page 3-135

VFNMA

Floating-point Fused Negate Multiply Accumulate

VFNMA, VFNMS on page 3-136

VFMS

Floating-point Fused Multiply Subtract

VFMA, VFMS on page 3-135

VFNMS

Floating-point Fused Negate Multiply Subtract

VFNMA, VFNMS on page 3-136

VLDM

Load Multiple extension registers

VLDM on page 3-137

VLDR

Loads an extension register from memory

VLDR on page 3-138

VLMA

Floating-point Multiply Accumulate

VLMA, VLMS on page 3-139

VLMS

Floating-point Multiply Subtract

VLMA, VLMS on page 3-139

VMOV

Floating-point Move Immediate

VMOV Immediate on page 3-140

VMOV

Floating-point Move Register

VMOV Register on page 3-141

VMOV

Copy ARM core register to single precision

VMOV ARM Core register to single precision on
page 3-143

VMOV

Copy 2 ARM core registers to 2 single precision

VMOV Two ARM Core registers to two single precision
on page 3-144

ARM DUI 0553A
ID121610

3-126

The Cortex-M4 Instruction Set

Table 3-15 Floating-point instructions (continued)
Mnemonic

Brief description

See

VMOV

Copies between ARM core register to scalar

VMOV ARM Core register to scalar on page 3-145

VMOV

Copies between Scalar to ARM core register

VMOV Scalar to ARM Core register on page 3-142

VMRS

Move to ARM core register from floating-point System
Register

VMRS on page 3-146

VMSR

Move to floating-point System Register from ARM Core
register

VMSR on page 3-147

VMUL

Multiply floating-point

VMUL on page 3-148

VNEG

Floating-point negate

VNEG on page 3-149

VNMLA

Floating-point multiply and add

VNMLA, VNMLS, VNMUL on page 3-150

VNMLS

Floating-point multiply and subtract

VNMLA, VNMLS, VNMUL on page 3-150

VNMUL

Floating-point multiply

VNMLA, VNMLS, VNMUL on page 3-150

VPOP

Pop extension registers

VPOP on page 3-151

VPUSH

Push extension registers

VPUSH on page 3-152

VSQRT

Floating-point square root

VSQRT on page 3-153

VSTM

Store Multiple extension registers

VSTM on page 3-154

VSTR

Stores an extension register to memory

VSTR on page 3-155

VSUB

Floating-point Subtract

VSUB on page 3-156

ARM DUI 0553A
ID121610

3-127

The Cortex-M4 Instruction Set

3.11.1

VABS
Floating-point Absolute.
Syntax
VABS{cond}.F32 Sd, Sm

where:
cond

Is an optional condition code, see Conditional execution on page 3-18.

Sd, Sm

Are the destination floating-point value and the operand floating-point value.

Operation
This instruction:
1.
Takes the absolute value of the operand floating-point register.
2.
Places the results in the destination floating-point register.
Restrictions
There are no restrictions.
Condition flags
The floating-point instruction clears the sign bit.
Examples
VABS.F32 S4, S6

ARM DUI 0553A
ID121610

3-128

The Cortex-M4 Instruction Set

3.11.2

VADD
Floating-point Add.
Syntax
VADD{cond}.F32 {Sd,} Sn, Sm

where:
cond
Sd
Sn, Sm

Is an optional condition code, see Conditional execution on page 3-18.
Specifies the destination floating-point value.
Are the operand floating-point values.

Operation
This instruction:
1.
Adds the values in the two floating-point operand registers.
2.
Places the results in the destination floating-point register.
Restrictions
There are no restrictions.
Condition flags
This instruction does not change the flags.
Examples
VADD.F32 S4, S6, S7

ARM DUI 0553A
ID121610

3-129

The Cortex-M4 Instruction Set

3.11.3

VCMP, VCMPE
Compares two floating-point registers, or one floating-point register and zero.
Syntax
VCMP{E}{cond}.F32 Sd, Sm
VCMP{E}{cond}.F32 Sd, #0.0

where:
Is an optional condition code, see Conditional execution on page 3-18.
If present, any NaN operand causes an Invalid Operation exception. Otherwise,
only a signaling NaN causes the exception.
Specifies the floating-point operand to compare.
Specifies the floating-point operand that is compared with.

cond
E
Sd
Sm

Operation
This instruction:
1.

Compares:
•

Two floating-point registers.

•

One floating-point register and zero.

Writes the result to the FPSCR flags.

Restrictions
This instruction can optionally raise an Invalid Operation exception if either operand is any type
of NaN. It always raises an Invalid Operation exception if either operand is a signaling NaN.
Condition flags
When this instruction writes the result to the FPSCR flags, the values are normally transferred to
the ARM flags by a subsequent VMRS instruction, see VMRS on page 3-146.
Examples
VCMP.F32

ARM DUI 0553A
ID121610

S4, #0.0VCMP.F32

S4, S2

3-130

The Cortex-M4 Instruction Set

3.11.4

VCVT, VCVTR between floating-point and integer
Converts a value in a register from floating-point to a 32-bit integer.
Syntax
VCVT{R}{cond}.Tm.F32 Sd, Sm
VCVT{cond}.F32.Tm Sd, Sm

where:
R
cond
Tm

Sd, Sm

If R is specified, the operation uses the rounding mode specified by the FPSCR. If R
is omitted. the operation uses the Round towards Zero rounding mode.
Is an optional condition code, see Conditional execution on page 3-18.
Specifies the data type for the operand. It must be one of:
•
S32 signed 32-bit value.
•
U32 unsigned 32-bit value.
Are the destination register and the operand register.

Operation
These instructions:
1.
Either
•
Converts a value in a register from floating-point value to a 32-bit integer.
•
Converts from a 32-bit integer to floating-point value.
2.
Places the result in a second register.
The floating-point to integer operation normally uses the Round towards Zero rounding mode,
but can optionally use the rounding mode specified by the FPSCR.
The integer to floating-point operation uses the rounding mode specified by the FPSCR.
Restrictions
There are no restrictions.
Condition flags
These instructions do not change the flags.

ARM DUI 0553A
ID121610

3-131

The Cortex-M4 Instruction Set

3.11.5

VCVT between floating-point and fixed-point
Converts a value in a register from floating-point to and from fixed-point.
Syntax
VCVT{cond}.Td.F32 Sd, Sd, #fbits
VCVT{cond}.F32.Td Sd, Sd, #fbits

where:
cond
Td

Sd
fbits

Is an optional condition code, see Conditional execution on page 3-18.
Specifies the data type for the fixed-point number. It must be one of:
•
S16 signed 16-bit value.
•
U16 unsigned 16-bit value.
•
S32 signed 32-bit value.
•
U32 unsigned 32-bit value.
Specifies the destination register and the operand register.
Specifies the number of fraction bits in the fixed-point number:
•
If Td is S16 or U16, fbits must be in the range 0-16.
•
I f Td is S32 or U32, fbits must be in the range 1-32.

Operation
These instructions:
1.
Either:
•
converts a value in a register from floating-point to fixed-point
•
converts a value in a register from fixed-point to floating-point.
2.
Places the result in a second register.
The floating-point values are single-precision.
The fixed-point value can be 16-bit or 32-bit. Conversions from fixed-point values take their
operand from the low-order bits of the source register and ignore any remaining bits.
Signed conversions to fixed-point values sign-extend the result value to the destination register
width.
Unsigned conversions to fixed-point values zero-extend the result value to the destination
register width.
The floating-point to fixed-point operation uses the Round towards Zero rounding mode. The
fixed-point to floating-point operation uses the Round to Nearest rounding mode.
Restrictions
There are no restrictions.
Condition flags
These instructions do not change the flags.

ARM DUI 0553A
ID121610

3-132

The Cortex-M4 Instruction Set

3.11.6

VCVTB, VCVTT
Converts between a half-precision value and a single-precision value.
Syntax
VCVT{y}{cond}.F32.F16 Sd, Sm
VCVT{y}{cond}.F16.F32 Sd, Sm

where:
y

cond
Sd
Sm

Specifies which half of the operand register Sm or destination register Sd is used
for the operand or destination:
•
If y is B, then the bottom half, bits [15:0], of Sm or Sd is used.
•
If y is T, then the top half, bits [31:16], of Sm or Sd is used.
Is an optional condition code, see Conditional execution on page 3-18.
Specifies the destination register.
Specifies the operand register.

Operation
This instruction with the .F16.32 suffix:
1.
Converts the half-precision value in the top or bottom half of a single-precision. register
to single-precision.
2.
Writes the result to a single-precision register.
This instruction with the .F32.F16 suffix:
1.
Converts the value in a single-precision register to half-precision.
2.
Writes the result into the top or bottom half of a single-precision register, preserving the
other half of the target register.
Restrictions
There are no restrictions.
Condition flags
These instructions do not change the flags.

ARM DUI 0553A
ID121610

3-133

The Cortex-M4 Instruction Set

3.11.7

VDIV
Divides floating-point values.
Syntax
VDIV{cond}.F32 {Sd,} Sn, Sm

where:
cond
Sd
Sn, Sm

Is an optional condition code, see Conditional execution on page 3-18.
Specifies the destination register.
Are the operand registers.

Operation
This instruction:
1.
Divides one floating-point value by another floating-point value.
2.
Writes the result to the floating-point destination register.
Restrictions
There are no restrictions.
Condition flags
These instructions do not change the flags.

ARM DUI 0553A
ID121610

3-134

The Cortex-M4 Instruction Set

3.11.8

VFMA, VFMS
Floating-point Fused Multiply Accumulate and Subtract.
Syntax
VFMA{cond}.F32 {Sd,} Sn, Sm
VFMS{cond}.F32 {Sd,} Sn, Sm

where:
cond
Sd
Sn, Sm

Is an optional condition code, see Conditional execution on page 3-18.
Specifies the destination register.
Are the operand registers.

Operation
The VFMA instruction:
1.
Multiplies the floating-point values in the operand registers.
2.
Accumulates the results into the destination register.
The result of the multiply is not rounded before the accumulation.
The VFMS instruction:
1.
Negates the first operand register.
2.
Multiplies the floating-point values of the first and second operand registers.
3.
Adds the products to the destination register.
4.
Places the results in the destination register.
The result of the multiply is not rounded before the addition.
Restrictions
There are no restrictions.
Condition flags
These instructions do not change the flags.

ARM DUI 0553A
ID121610

3-135

The Cortex-M4 Instruction Set

3.11.9

VFNMA, VFNMS
Floating-point Fused Negate Multiply Accumulate and Subtract.
Syntax
VFNMA{cond}.F32 {Sd,} Sn, Sm
VFNMS{cond}.F32 {Sd,} Sn, Sm

where:
cond
Sd
Sn, Sm

Is an optional condition code, see Conditional execution on page 3-18.
Specifies the destination register.
Are the operand registers.

Operation
The VFNMA instruction:
1.
Negates the first floating-point operand register.
2.
Multiplies the first floating-point operand with second floating-point operand.
3.
Adds the negation of the floating -point destination register to the product
4.
Places the result into the destination register.
The result of the multiply is not rounded before the addition.
The VFNMS instruction:
1.
Multiplies the first floating-point operand with second floating-point operand.
2.
Adds the negation of the floating-point value in the destination register to the product.
3.
Places the result in the destination register.
The result of the multiply is not rounded before the addition.
Restrictions
There are no restrictions.
Condition flags
These instructions do not change the flags.

ARM DUI 0553A
ID121610

3-136

The Cortex-M4 Instruction Set

3.11.10 VLDM
Floating-point Load Multiple.
Syntax
VLDM{mode}{cond}{.size} Rn{!}, list

where:
mode

cond
size
Rn
!
list

Specifies the addressing mode:
•
IAIncrement After. The consecutive addresses start at the address specified
in Rn.
•
DB Decrement Before. The consecutive addresses end just before
the address specified in Rn.
Is an optional condition code, see Conditional execution on page 3-18.
Is an optional data size specifier.
Specifies the base register. The SP can be used.
Is the command to the instruction to write a modified value back to Rn. This is
required if mode == DB, and is optional if mode == IA.
Specifies the list of extension registers to be loaded, as a list of consecutively
numbered doubleword or singleword registers, separated by commas and
surrounded by brackets.

Operation
This instruction loads:
•
Multiple extension registers from consecutive memory locations using an address from an
ARM core register as the base address.
Restrictions
The restrictions are:
•
If size is present, it must be equal to the size in bits, 32 or 64, of the registers in list.
•
For the base address, the SP can be used. In the ARM instruction set, if ! is not specified
the PC can be used.
•
list must contain at least one register. If it contains doubleword registers, it must not
contain more than 16 registers.
•
If using the Decrement Before addressing mode, the write back flag, !, must be appended
to the base register specification.
Condition flags
These instructions do not change the flags.

ARM DUI 0553A
ID121610

3-137

The Cortex-M4 Instruction Set

3.11.11 VLDR
Loads a single extension register from memory.
Syntax
VLDR{cond}{.64} Dd, [Rn{#imm}]
VLDR{cond}{.64} Dd, label
VLDR{cond}{.64} Dd, [PC, #imm}]
VLDR{cond}{.32} Sd, [Rn {, #imm}]
VLDR{cond}{.32} Sd, label
VLDR{cond}{.32} Sd, [PC, #imm]

where:
cond
64, 32
Dd
Sd
Rn
imm
label

Is an optional condition code, see Conditional execution on page 3-18.
Are the optional data size specifiers.
Specifies the destination register for a doubleword load.
Specifies the destination register for a singleword load.
Specifies the base register. The SP can be used.
Is the + or - immediate offset used to form the address. Permitted address values
are multiples of 4 in the range 0 to 1020.
Specifies the label of the literal data item to be loaded.

Operation
This instruction:
•
Loads a single extension register from memory, using a base address from an ARM core
register, with an optional offset.
Restrictions
There are no restrictions.
Condition flags
These instructions do not change the flags.

ARM DUI 0553A
ID121610

3-138

The Cortex-M4 Instruction Set

3.11.12 VLMA, VLMS
Multiplies two floating-point values, and accumulates or subtracts the results.
Syntax
VLMA{cond}.F32 Sd, Sn, Sm
VLMS{cond}.F32 Sd, Sn, Sm

where:
cond
Sd
Sn, Sm

Is an optional condition code, see Conditional execution on page 3-18.
Specifies the destination floating-point value.
Are the operand floating-point values.

Operation
The floating-point Multiply Accumulate instruction:
1.
Multiplies two floating-point values.
2.
Adds the results to the destination floating-point value.
The floating-point Multiply Subtract instruction:
1.
Multiplies two floating-point values.
2.
Subtracts the products from the destination floating-point value.
3.
Places the results in the destination register.
Restrictions
There are no restrictions.
Condition flags
These instructions do not change the flags.

ARM DUI 0553A
ID121610

3-139

The Cortex-M4 Instruction Set

3.11.13 VMOV Immediate
Move floating-point Immediate.
Syntax
VMOV{cond}.F32 Sd, #imm

where:
cond
Sd
imm

Is an optional condition code, see Conditional execution on page 3-18.
Specifies the branch destination.
Is a floating-point constant.

Operation
This instruction copies a constant value to a floating-point register.
Restrictions
There are no restrictions.
Condition flags
These instructions do not change the flags.

ARM DUI 0553A
ID121610

3-140

The Cortex-M4 Instruction Set

3.11.14 VMOV Register
Copies the contents of one register to another.
Syntax
VMOV{cond}.F64 Dd, Dm
VMOV{cond}.F32 Sd, Sm

where:
cond
Dd
Dm
Sd
Sm

Is an optional condition code, see Conditional execution on page 3-18.
Specifies the destination register, for a doubleword operation.
Specifies the source register, for a doubleword operation.
Specifies the destination register, for a singleword operation.
Specifies the source register, for a singleword operation.

Operation
This instruction copies the contents of one floating-point register to another.
Restrictions
There are no restrictions
Condition flags
These instructions do not change the flags.

ARM DUI 0553A
ID121610

3-141

The Cortex-M4 Instruction Set

3.11.15 VMOV Scalar to ARM Core register
Transfers one word of a doubleword floating-point register to an ARM core register.
Syntax
VMOV{cond} Rt, Dn[x]

where:
Is an optional condition code, see Conditional execution on page 3-18.
Specifies the destination ARM core register.
Specifies the 64-bit doubleword register.
Specifies which half of the doubleword register to use:
•
If x is 0, use lower half of doubleword register
•
If x is 1, use upper half of doubleword register.

cond
Rt
Dn
x

Operation
This instruction transfers:
•

one word from the upper or lower half of a doubleword floating-point register to an ARM
core register.

Restrictions
Rt cannot be PC or SP.
Condition flags
These instructions do not change the flags.

ARM DUI 0553A
ID121610

3-142

The Cortex-M4 Instruction Set

3.11.16 VMOV ARM Core register to single precision
Transfers a single-precision register to and from an ARM core register.
Syntax
VMOV{cond} Sn, Rt
VMOV{cond} Rt, Sn

where:
cond
Sn
Rt

Is an optional condition code, see Conditional execution on page 3-18.
Specifies the single-precision floating-point register.
Specifies the ARM core register.

Operation
This instruction transfers:
•
The contents of a single-precision register to an ARM core register.
•
The contents of an ARM core register to a single-precision register.
Restrictions
Rt cannot be PC or SP.
Condition flags
These instructions do not change the flags.

ARM DUI 0553A
ID121610

3-143

The Cortex-M4 Instruction Set

3.11.17 VMOV Two ARM Core registers to two single precision
Transfers two consecutively numbered single-precision registers to and from two ARM core
registers.
Syntax
VMOV{cond} Sm, Sm1, Rt, Rt2
VMOV{cond} Rt, Rt2, Sm, Sm

where:
cond
Sm
Sm1
Rt
Rt2

Is an optional condition code, see Conditional execution on page 3-18.
Specifies the first single-precision register.
Specifies the second single-precision register. This is the next single-precision
register after Sm.
Specifies the ARM core register that Sm is transferred to or from.
Specifies the The ARM core register that Sm1 is transferred to or from.

Operation
This instruction transfers:
•
The contents of two consecutively numbered single-precision registers to two ARM core
registers.
•
The contents of two ARM core registers to a pair of single-precision registers.
Restrictions
The restrictions are:
•
The floating-point registers must be contiguous, one after the other.
•
The ARM core registers do not have to be contiguous.
•
Rt cannot be PC or SP.
Condition flags
These instructions do not change the flags.

ARM DUI 0553A
ID121610

3-144

The Cortex-M4 Instruction Set

3.11.18 VMOV ARM Core register to scalar
Transfers one word to a floating-point register from an ARM core register.
Syntax
VMOV{cond}{.32} Dd[x], Rt

where:
cond
32
Dd[x]

Is an optional condition code, see Conditional execution on page 3-18.
Is an optional data size specifier.
Specifies the destination, where [x] defines which half of the doubleword is
transferred, as follows:
•
If x is 0, the lower half is extracted
•
If x is 1, the upper half is extracted.
Specifies the source ARM core register.

Operation
This instruction transfers one word to the upper or lower half of a doubleword floating-point
register from an ARM core register.
Restrictions
Rt cannot be PC or SP.
Condition flags
These instructions do not change the flags.

ARM DUI 0553A
ID121610

3-145

The Cortex-M4 Instruction Set

3.11.19 VMRS
Move to ARM Core register from floating-point System Register.
Syntax
VMRS{cond} Rt, FPSCR
VMRS{cond} APSR_nzcv, FPSCR

where:
cond
Rt
APSR_nzcv

Is an optional condition code, see Conditional execution on page 3-18.
Specifies the destination ARM core register. This register can be R0-R14.
Transfer floating-point flags to the APSR flags.

Operation
This instruction performs one of the following actions:
•
Copies the value of the FPSCR to a general-purpose register.
•
Copies the value of the FPSCR flag bits to the APSR N, Z, C, and V flags.
Restrictions
Rt cannot be PC or SP.
Condition flags
These instructions optionally change the flags: N, Z, C, V

ARM DUI 0553A
ID121610

3-146

The Cortex-M4 Instruction Set

3.11.20 VMSR
Move to floating-point System Register from ARM Core register.
Syntax
VMSR{cond} FPSCR, Rt

where:
cond
Rt

Is an optional condition code, see Conditional execution on page 3-18.
Specifies the general-purpose register to be transferred to the FPSCR.

Operation
This instruction moves the value of a general-purpose register to the FPSCR. See Floating-point
Status Control Register on page 4-50 for more information.
Restrictions
The restrictions are:
•
Rt cannot be PC or SP.
Condition flags
This instruction updates the FPSCR.

ARM DUI 0553A
ID121610

3-147

The Cortex-M4 Instruction Set

3.11.21 VMUL
Floating-point Multiply.
Syntax
VMUL{cond}.F32 {Sd,} Sn, Sm

where:
cond
Sd
Sn, Sm

Is an optional condition code, see Conditional execution on page 3-18.
Specifies the destination floating-point value.
Are the operand floating-point values.

Operation
This instruction:
1.
Multiplies two floating-point values.
2.
Places the results in the destination register.
Restrictions
There are no restrictions.
Condition flags
These instructions do not change the flags.

ARM DUI 0553A
ID121610

3-148

The Cortex-M4 Instruction Set

3.11.22 VNEG
Floating-point Negate.
Syntax
VNEG{cond}.F32 Sd, Sm

where:
cond
Sd
Sm

Is an optional condition code, see Conditional execution on page 3-18.
Specifies the destination floating-point value.
Specifies the operand floating-point value.

Operation
This instruction:
1.
Negates a floating-point value.
2.
Places the results in a second floating-point register.
The floating-point instruction inverts the sign bit.
Restrictions
There are no restrictions.
Condition flags
These instructions do not change the flags.

ARM DUI 0553A
ID121610

3-149

The Cortex-M4 Instruction Set

3.11.23 VNMLA, VNMLS, VNMUL
Floating-point multiply with negation followed by add or subtract.
Syntax
VNMLA{cond}.F32 Sd, Sn, Sm
VNMLS{cond}.F32 Sd, Sn, Sm
VNMUL{cond}.F32 {Sd,} Sn, Sm

where:
cond
Sd
Sn, Sm

Is an optional condition code, see Conditional execution on page 3-18.
Specifies the destination floating-point register.
Are the operand floating-point registers.

Operation
The VNMLA instruction:
1.
Multiplies two floating-point register values.
2.
Adds the negation of the floating-point value in the destination register to the negation of
the product.
3.
Writes the result back to the destination register.
The VNMLS instruction:
1.
Multiplies two floating-point register values.
2.
Adds the negation of the floating-point value in the destination register to the product.
3.
writes the result back to the destination register.
The VNMUL instruction:
1.
Multiplies together two floating-point register values.
2.
Writes the negation of the result to the destination register.
Restrictions
There are no restrictions.
Condition flags
These instructions do not change the flags.

ARM DUI 0553A
ID121610

3-150

The Cortex-M4 Instruction Set

3.11.24 VPOP
Floating-point extension register Pop.
Syntax
VPOP{cond}{.size} list

where:
cond
size
list

Is an optional condition code, see Conditional execution on page 3-18.
Is an optional data size specifier. If present, it must be equal to the size in bits, 32
or 64, of the registers in list.
Is a list of extension registers to be loaded, as a list of consecutively numbered
doubleword or singleword registers, separated by commas and
surrounded by brackets.

Operation
This instruction loads multiple consecutive extension registers from the stack.
Restrictions
The list must contain at least one register, and not more than sixteen registers.
Condition flags
These instructions do not change the flags.

ARM DUI 0553A
ID121610

3-151

The Cortex-M4 Instruction Set

3.11.25 VPUSH
Floating-point extension register Push.
Syntax
VPUSH{cond}{.size} list

where:
Is an optional condition code, see Conditional execution on page 3-18.
Is an optional data size specifier. If present, it must be equal to the size in bits, 32
or 64, of the registers in list.
Is a list of the extension registers to be stored, as a list of consecutively numbered
doubleword or singleword registers, separated by commas and surrounded by
brackets.

cond
size
list

Operation
This instruction:
•

Stores multiple consecutive extension registers to the stack.

Restrictions
The restrictions are:
•
list must contain at least one register, and not more than sixteen.
Condition flags
These instructions do not change the flags.

ARM DUI 0553A
ID121610

3-152

The Cortex-M4 Instruction Set

3.11.26 VSQRT
Floating-point Square Root.
Syntax
VSQRT{cond}.F32 Sd, Sm

where:
cond
Sd
Sm

Is an optional condition code, see Conditional execution on page 3-18.
Specifies the destination floating-point value.
Specifies the operand floating-point value.

Operation
This instruction:
•
Calculates the square root of the value in a floating-point register.
•
Writes the result to another floating-point register.
Restrictions
There are no restrictions.
Condition flags
These instructions do not change the flags.

ARM DUI 0553A
ID121610

3-153

The Cortex-M4 Instruction Set

3.11.27 VSTM
Floating-point Store Multiple.
Syntax
VSTM{mode}{cond}{.size} Rn{!}, list

where:
mode

cond
size
Rn
!
list

Specifies the addressing mode:
•
IA Increment After. The consecutive addresses start at the address specified
in Rn. This is the default and can be omitted.
•
DB Decrement Before. The consecutive addresses end just before the
address specified in Rn.
Is an optional condition code, see Conditional execution on page 3-18.
Is an optional data size specifier. If present, it must be equal to the size in bits, 32
or 64, of the registers in list.
Specifies the base register. The SP can be used.
is the function that causes the instruction to write a modified value back to Rn.
Required if mode == DB.
Is a list of the extension registers to be stored, as a list of consecutively numbered
doubleword or singleword registers, separated by commas and surrounded by
brackets.

Operation
This instruction:
•
Stores multiple extension registers to consecutive memory locations using a base address
from an ARM core register.
Restrictions
The restrictions are:
•
list must contain at least one register. If it contains doubleword registers it must not
contain more than 16 registers.
•
Use of the PC as Rn is deprecated.
Condition flags
These instructions do not change the flags.

ARM DUI 0553A
ID121610

3-154

The Cortex-M4 Instruction Set

3.11.28 VSTR
Floating-point Store.
Syntax
VSTR{cond}{.32} Sd, [Rn{, #imm}]
VSTR{cond}{.64} Dd, [Rn{, #imm}]

where:
cond
32, 64
Sd
Dd
Rn
imm

Is an optional condition code, see Conditional execution on page 3-18.
Are the optional data size specifiers.
Specifies the source register for a singleword store.
Specifies the source register for a doubleword store.
Specifies the base register. The SP can be used.
Is the + or - immediate offset used to form the address. Values are multiples of 4
in the range 0-1020. imm can be omitted, meaning an offset of +0.

Operation
This instruction:
•
Stores a single extension register to memory, using an address from an ARM core register,
with an optional offset, defined in imm.
Restrictions
The restrictions are:
•
The use of PC for Rn is deprecated.
Condition flags
These instructions do not change the flags.

ARM DUI 0553A
ID121610

3-155

The Cortex-M4 Instruction Set

3.11.29 VSUB
Floating-point Subtract.
Syntax
VSUB{cond}.F32 {Sd,} Sn, Sm

where:
cond
Sd
Sn, Sm

Is an optional condition code, see Conditional execution on page 3-18.
Specifies the destination floating-point value.
Are the operand floating-point value.

Operation
This instruction:
1.
Subtracts one floating-point value from another floating-point value.
2.
Places the results in the destination floating-point register.
Restrictions
There are no restrictions.
Condition flags
These instructions do not change the flags.

ARM DUI 0553A
ID121610

3-156

The Cortex-M4 Instruction Set

3.12

Miscellaneous instructions
Table 3-16 shows the remaining Cortex-M4 instructions:
Table 3-16 Miscellaneous instructions

ARM DUI 0553A
ID121610

Mnemonic

Brief description

See

BKPT

Breakpoint

BKPT on page 3-158

CPSID

Change Processor State, Disable Interrupts

CPS on page 3-159

CPSIE

Change Processor State, Enable Interrupts

CPS on page 3-159

DMB

Data Memory Barrier

DMB on page 3-160

DSB

Data Synchronization Barrier

DSB on page 3-161

ISB

Instruction Synchronization Barrier

ISB on page 3-162

MRS

Move from special register to register

MRS on page 3-163

MSR

Move from register to special register

MSR on page 3-164

NOP

No Operation

NOP on page 3-165

SEV

Send Event

SEV on page 3-166

SVC

Supervisor Call

SVC on page 3-167

WFE

Wait For Event

WFE on page 3-168

WFI

Wait For Interrupt

WFI on page 3-169

3-157

The Cortex-M4 Instruction Set

3.12.1

BKPT
Breakpoint.
Syntax
BKPT #imm

where:
Is an expression evaluating to an integer in the range 0-255 (8-bit value).

imm

Operation
The BKPT instruction causes the processor to enter Debug state. Debug tools can use this to
investigate system state when the instruction at a particular address is reached.
imm is ignored by the processor. If required, a debugger can use it to store additional information

about the breakpoint.
The BKPT instruction can be placed inside an IT block, but it executes unconditionally, unaffected
by the condition specified by the IT instruction.
Condition flags
This instruction does not change the flags.
Examples
BKPT #0x3

; Breakpoint with immediate value set to 0x3 (debugger can
; extract the immediate value by locating it using the PC)

Note
ARM does not recommend the use of the BKPT instruction with an immediate value set to
0xAB for any purpose other than Semi-hosting.

ARM DUI 0553A
ID121610

3-158

The Cortex-M4 Instruction Set

3.12.2

CPS
Change Processor State.
Syntax
CPSeffect iflags

where:
Is one of:

effect

IE
ID

Clears the special purpose register.
Sets the special purpose register.

Is a sequence of one or more flags:
i
Set or clear PRIMASK.
f
Set or clear FAULTMASK.

iflags

Operation
CPS changes the PRIMASK and FAULTMASK special register values. See Exception mask

registers on page 2-7 for more information about these registers.
Restrictions
The restrictions are:
•
use CPS only from privileged software, it has no effect if used in unprivileged software
•
CPS cannot be conditional and so must not be used inside an IT block.
Condition flags
This instruction does not change the condition flags.
Examples
CPSID
CPSID
CPSIE
CPSIE

ARM DUI 0553A
ID121610

i
f
i
f

;
;
;
;

Disable interrupts and configurable fault handlers (set PRIMASK)
Disable interrupts and all fault handlers (set FAULTMASK)
Enable interrupts and configurable fault handlers (clear PRIMASK)
Enable interrupts and fault handlers (clear FAULTMASK)

3-159

The Cortex-M4 Instruction Set

3.12.3

DMB
Data Memory Barrier.
Syntax
DMB{cond}

where:
Is an optional condition code, see Conditional execution on page 3-18.

cond

Operation
DMB acts as a data memory barrier. It ensures that all explicit memory accesses that appear, in
program order, before the DMB instruction are completed before any explicit memory accesses
that appear, in program order, after the DMB instruction. DMB does not affect the ordering or

execution of instructions that do not access memory.
Condition flags
This instruction does not change the flags.
Examples
DMB

ARM DUI 0553A
ID121610

; Data Memory Barrier

3-160

The Cortex-M4 Instruction Set

3.12.4

DSB
Data Synchronization Barrier.
Syntax
DSB{cond}

where:
Is an optional condition code, see Conditional execution on page 3-18.

cond

Operation
DSB acts as a special data synchronization memory barrier. Instructions that come after the DSB,
in program order, do not execute until the DSB instruction completes. The DSB instruction
completes when all explicit memory accesses before it complete.

Condition flags
This instruction does not change the flags.
Examples
DSB ; Data Synchronisation Barrier

ARM DUI 0553A
ID121610

3-161

The Cortex-M4 Instruction Set

3.12.5

ISB
Instruction Synchronization Barrier.
Syntax
ISB{cond}

where:
Is an optional condition code, see Conditional execution on page 3-18.

cond

Operation
ISB acts as an instruction synchronization barrier. It flushes the pipeline of the processor, so that
all instructions following the ISB are fetched from cache or memory again, after the ISB

instruction has been completed.
Condition flags
This instruction does not change the flags.
Examples
ISB

ARM DUI 0553A
ID121610

; Instruction Synchronisation Barrier

3-162

The Cortex-M4 Instruction Set

3.12.6

MRS
Move the contents of a special register to a general-purpose register.
Syntax
MRS{cond} Rd, spec_reg

where:
cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the destination register.

spec_reg

Can be any of: APSR, IPSR, EPSR, IEPSR, IAPSR, EAPSR, PSR, MSP, PSP, PRIMASK, BASEPRI,
BASEPRI_MAX, FAULTMASK, or CONTROL.
Note
All the EPSR and IPSR fields are zero when read by the MRS instruction.

Operation
Use MRS in combination with MSR as part of a read-modify-write sequence for updating a PSR,
for example to clear the Q flag.
In process swap code, the programmers model state of the process being swapped out must be
saved, including relevant PSR contents. Similarly, the state of the process being swapped in
must also be restored. These operations use MRS in the state-saving instruction sequence and MSR
in the state-restoring instruction sequence.
Note
BASEPRI_MAX is an alias of BASEPRI when used with the MRS instruction.
See MSR on page 3-164.
Restrictions
Rd must not be SP and must not be PC.

Condition flags
This instruction does not change the flags.
Examples
MRS

ARM DUI 0553A
ID121610

R0, PRIMASK ; Read PRIMASK value and write it to R0

3-163

The Cortex-M4 Instruction Set

3.12.7

MSR
Move the contents of a general-purpose register into the specified special register.
Syntax
MSR{cond} spec_reg, Rn

where:
cond

Is an optional condition code, see Conditional execution on page 3-18.

Specifies the source register.

spec_reg

Can be any of: APSR_nzcvq, APSR_g, APSR_nzcvqg, MSP, PSP, PRIMASK, BASEPRI,
BASEPRI_MAX, FAULTMASK, or CONTROL.
Note
You can use APSR to refer to APSR_nzcvq.

Operation
The register access operation in MSR depends on the privilege level. Unprivileged software can
only access the APSR, see Table 2-4 on page 2-5. Privileged software can access all special
registers.
In unprivileged software writes to unallocated or execution state bits in the PSR are ignored.
Note
When you write to BASEPRI_MAX, the instruction writes to BASEPRI only if either:
•
Rn is non-zero and the current BASEPRI value is 0
•
Rn is non-zero and less than the current BASEPRI value.
See MRS on page 3-163.
Restrictions
Rn must not be SP and must not be PC.

Condition flags
This instruction updates the flags explicitly based on the value in Rn.
Examples
MSR

ARM DUI 0553A
ID121610

CONTROL, R1 ; Read R1 value and write it to the CONTROL register

3-164

The Cortex-M4 Instruction Set

3.12.8

NOP
No Operation.
Syntax
NOP{cond}

where:
Is an optional condition code, see Conditional execution on page 3-18.

cond

Operation
NOP does nothing. NOP is not necessarily a time-consuming NOP. The processor might remove it
from the pipeline before it reaches the execution stage.

Use NOP for padding, for example to place the following instruction on a 64-bit boundary.
Condition flags
This instruction does not change the flags.
Examples
NOP

ARM DUI 0553A
ID121610

; No operation

3-165

The Cortex-M4 Instruction Set

3.12.9

SEV
Send Event.
Syntax
SEV{cond}

where:
Is an optional condition code, see Conditional execution on page 3-18.

cond

Operation
SEV is a hint instruction that causes an event to be signaled to all processors within a

multiprocessor system. It also sets the local event register to 1, see Power management on
page 2-32.
Condition flags
This instruction does not change the flags.
Examples
SEV ; Send Event

ARM DUI 0553A
ID121610

3-166

The Cortex-M4 Instruction Set

3.12.10 SVC
Supervisor Call.
Syntax
SVC{cond} #imm

where:
cond

Is an optional condition code, see Conditional execution on page 3-18.

imm

Is an expression evaluating to an integer in the range 0-255 (8-bit value).

Operation
The SVC instruction causes the SVC exception.
imm is ignored by the processor. If required, it can be retrieved by the exception handler to

determine what service is being requested.
Condition flags
This instruction does not change the flags.
Examples
SVC

ARM DUI 0553A
ID121610

#0x32 ; Supervisor Call (SVCall handler can extract the immediate value
; by locating it through the stacked PC)

3-167

The Cortex-M4 Instruction Set

3.12.11 WFE
Wait For Event.
Syntax
WFE{cond}

where:
Is an optional condition code, see Conditional execution on page 3-18.

cond

Operation
WFE is a hint instruction.

If the event register is 0, WFE suspends execution until one of the following events occurs:
•

an exception, unless masked by the exception mask registers or the current priority level

•

an exception enters the Pending state, if SEVONPEND in the System Control Register is set

•

a Debug Entry request, if Debug is enabled

•

an event signaled by a peripheral or another processor in a multiprocessor system using
the SEV instruction.

If the event register is 1, WFE clears it to 0 and returns immediately.
For more information see Power management on page 2-32.
Condition flags
This instruction does not change the flags.
Examples
WFE

ARM DUI 0553A
ID121610

; Wait for event

3-168

The Cortex-M4 Instruction Set

3.12.12 WFI
Wait for Interrupt.
Syntax
WFI{cond}

where:
Is an optional condition code, see Conditional execution on page 3-18.

cond

Operation
WFI is a hint instruction that suspends execution until one of the following events occurs:

•
•
•

a non-masked interrupt occurs and is taken
an interrupt masked by PRIMASK becomes pending
a Debug Entry request.

Condition flags
This instruction does not change the flags.
Examples
WFI ; Wait for interrupt

ARM DUI 0553A
ID121610

3-169

Chapter 4
Cortex-M4 Peripherals

This chapter describes the ARM Cortex-M4 core peripherals. It contains the following sections:
•
About the Cortex-M4 peripherals on page 4-2
•
Nested Vectored Interrupt Controller on page 4-3
•
System control block on page 4-11
•
System timer, SysTick on page 4-33.
•
Optional Memory Protection Unit on page 4-37
•
Floating Point Unit (FPU) on page 4-48.

ARM DUI 0553A
ID121610

4-1

Cortex-M4 Peripherals

4.1

About the Cortex-M4 peripherals
The address map of the Private Peripheral Bus (PPB) is:
Table 4-1 Core peripheral register regions
Address

Core peripheral

Description

0xE000E008-0xE000E00F

SyStem Control Block

Table 4-12 on page 4-11

0xE000E010-0xE000E01F

System timer

Table 4-32 on page 4-33

0xE000E100-0xE000E4EF

Nested Vectored Interrupt Controller

Table 4-2 on page 4-3

0xE000ED00-0xE000ED3F

System Control Block

Table 4-12 on page 4-11

0xE000ED90-0xE000ED93

MPU Type Register

Reads as zero, indicating MPU is not implemented a

0xE000ED90-0xE000EDB8

Memory Protection Unit

Table 4-38 on page 4-38

0xE000EF00-0xE000EF03

Nested Vectored Interrupt Controller

Table 4-2 on page 4-3

0xE000EF30-0xE000EF44

Floating Point Unit

Table 4-49 on page 4-48

a. Software can read the MPU Type Register at 0xE000ED90 to test for the presence of a Memory Protection Unit (MPU)

In register descriptions:
•
the register type is described as follows:
RW
Read and write.
RO
Read-only.
WO
Write-only.
•
the required privilege gives the privilege level required to access the register, as follows:
Privileged
Only privileged software can access the register.
Unprivileged
Both unprivileged and privileged software can access the register.

ARM DUI 0553A
ID121610

4-2

Cortex-M4 Peripherals

4.2

Nested Vectored Interrupt Controller
This section describes the NVIC and the registers it uses. The NVIC supports:
•

An implementation-defined number of interrupts, in the range 1-240 interrupts.

•

A programmable priority level of 0-255 for each interrupt. A higher level corresponds to
a lower priority, so level 0 is the highest interrupt priority.

•

Level and pulse detection of interrupt signals.

•

Dynamic reprioritization of interrupts.

•

Grouping of priority values into group priority and subpriority fields.

•

Interrupt tail-chaining.

•

An external Non Maskable Interrupt (NMI)

•

Optional WIC, providing ultra-low power sleep mode support.

The processor automatically stacks its state on exception entry and unstacks this state on
exception exit, with no instruction overhead. This provides low latency exception handling. The
hardware implementation of the NVIC registers is:
Table 4-2 NVIC register summary
Address

Name

Type

Required
privilege

Reset value

Description

0xE000E1000xE000E11C

NVIC_ISER0NVIC_ISER7

Privileged

0x00000000

Interrupt Set-enable Registers on page 4-4

0XE000E1800xE000E19C

NVIC_ICER0NVIC_ICER7

Privileged

0x00000000

Interrupt Clear-enable Registers on page 4-5

0XE000E2000xE000E21C

NVIC_ISPR0NVIC_ISPR7

Privileged

0x00000000

Interrupt Set-pending Registers on page 4-5

0XE000E2800xE000E29C

NVIC_ICPR0NVIC_ICPR7

Privileged

0x00000000

Interrupt Clear-pending Registers on page 4-6

0xE000E3000xE000E31C

NVIC_IABR0NVIC_IABR7

Privileged

0x00000000

Interrupt Active Bit Registers on page 4-7

0xE000E4000xE000E4EF

NVIC_IPR0NVIC_IPR59

Privileged

0x00000000

Interrupt Priority Registers on page 4-7

0xE000EF00

STIR

Configurable a

0x00000000

Software Trigger Interrupt Register on page 4-8

a. See the register description for more information.

ARM DUI 0553A
ID121610

4-3

Cortex-M4 Peripherals

4.2.1

Accessing the Cortex-M4 NVIC registers using CMSIS
CMSIS functions enable software portability between different Cortex-M profile processors. To
access the NVIC registers when using CMSIS, use the following functions:
Table 4-3 CMSIS access NVIC functions
CMSIS function

Description

void NVIC_EnableIRQ(IRQn_Type IRQn)a

Enables an interrupt or exception.

void NVIC_DisableIRQ(IRQn_Type IRQn)a

Disables an interrupt or exception.

void NVIC_SetPendingIRQ(IRQn_Type IRQn)a

Sets the pending status of interrupt or exception to 1.

void NVIC_ClearPendingIRQ(IRQn_Type IRQn)a

Clears the pending status of interrupt or exception to 0.

uint32_t NVIC_GetPendingIRQ(IRQn_Type IRQn)a

Reads the pending status of interrupt or exception. This
function returns non-zero value if the pending status is
set to 1.

void NVIC_SetPriority(IRQn_Type IRQn, uint32_t priority)a

Sets the priority of an interrupt or exception with
configurable priority level to 1.

uint32_t NVIC_GetPriority(IRQn_Type IRQn)a

Reads the priority of an interrupt or exception with
configurable priority level. This function return the
current priority level.

a. The input parameter IRQn is the IRQ number, see Table 2-16 on page 2-22 for more information.

4.2.2

Interrupt Set-enable Registers
The NVIC_ISER0-NVIC_ISER7 registers enable interrupts, and show which interrupts are
enabled. See the register summary in Table 4-2 on page 4-3 for the register attributes.
The bit assignments are:
31

0
SETENA bits

Table 4-4 ISER bit assignments
Bits

Name

Function

[31:0]

SETENA

Interrupt set-enable bits.
Write:
0 = no effect
1 = enable interrupt.
Read:
0 = interrupt disabled
1 = interrupt enabled.

If a pending interrupt is enabled, the NVIC activates the interrupt based on its priority. If an
interrupt is not enabled, asserting its interrupt signal changes the interrupt state to pending, but
the NVIC never activates the interrupt, regardless of its priority.

ARM DUI 0553A
ID121610

4-4

Cortex-M4 Peripherals

4.2.3

Interrupt Clear-enable Registers
The NVIC_ICER0-NVIC_ICER7 registers disable interrupts, and show which interrupts are
enabled. See the register summary in Table 4-2 on page 4-3 for the register attributes.
The bit assignments are:
31

0
CLRENA bits

Table 4-5 ICER bit assignments

4.2.4

Bits

Name

Function

[31:0]

CLRENA

Interrupt clear-enable bits.
Write:
0 = no effect
1 = disable interrupt.
Read:
0 = interrupt disabled
1 = interrupt enabled.

Interrupt Set-pending Registers
The NVIC_ISPR0-NVIC_ISPR7 registers force interrupts into the pending state, and show
which interrupts are pending. See the register summary in Table 4-2 on page 4-3 for the register
attributes.
The bit assignments are:
31

0
SETPEND bits

Table 4-6 ISPR bit assignments
Bits

Name

Function

[31:0]

SETPEND

Interrupt set-pending bits.
Write:
0 = no effect
1 = changes interrupt state to pending.
Read:
0 = interrupt is not pending
1 = interrupt is pending.

Note
Writing 1 to the ISPR bit corresponding to:
•
an interrupt that is pending has no effect
•
a disabled interrupt sets the state of that interrupt to pending.

ARM DUI 0553A
ID121610

4-5

Cortex-M4 Peripherals

4.2.5

Interrupt Clear-pending Registers
The NVIC_ICPR0-NCVIC_ICPR7 registers remove the pending state from interrupts, and
show which interrupts are pending. See the register summary in Table 4-2 on page 4-3 for the
register attributes.
The bit assignments are:
31

0
CLRPEND bits

Table 4-7 ICPR bit assignments
Bits

Name

Function

[31:0]

CLRPEND

Interrupt clear-pending bits.
Write:
0 = no effect
1 = removes pending state an interrupt.
Read:
0 = interrupt is not pending
1 = interrupt is pending.

Note
Writing 1 to an ICPR bit does not affect the active state of the corresponding interrupt.

ARM DUI 0553A
ID121610

4-6

Cortex-M4 Peripherals

4.2.6

Interrupt Active Bit Registers
The NVIC_IABR0-NVIC_IABR7 registers indicate which interrupts are active. See the register
summary in Table 4-2 on page 4-3 for the register attributes.
The bit assignments are:
31

0
ACTIVE bits

Table 4-8 IABR bit assignments
Bits

Name

Function

[31:0]

ACTIVE

Interrupt active flags:
0 = interrupt not active
1 = interrupt active.

A bit reads as one if the status of the corresponding interrupt is active or active and pending.
4.2.7

Interrupt Priority Registers
The NVIC_IPR0-NVIC_IPR59 registers provide an 8-bit priority field for each interrupt and
each register holds four priority fields. These registers are byte-accessible. See the register
summary in Table 4-2 on page 4-3 for their attributes. Each register holds four priority fields as
shown:
31

16 15
PRI_238

PRI_4n+3

PRI_4n+2

PRI_3

PRI_2

IPR0

PRI_237

PRI_236

PRI_4n+1

PRI_4n

PRI_1

PRI_0

...

IPRn

8 7

...

PRI_239

...

IPR59

24 23

Table 4-9 IPR bit assignments
Bits

Name

Function

[31:24]

Priority, byte offset 3

[23:16]

Priority, byte offset 2

[15:8]

Priority, byte offset 1

Each implementation-defined priority field can hold a priority value, 0-255. The
lower the value, the greater the priority of the corresponding interrupt. Register
priority value fields are eight bits wide, and non-implemented low-order bits read as
zero and ignore writes.

[7:0]

Priority, byte offset 0

See Accessing the Cortex-M4 NVIC registers using CMSIS on page 4-4 for more information
about the access to the interrupt priority array, which provides the software view of the interrupt
priorities.

ARM DUI 0553A
ID121610

4-7

Cortex-M4 Peripherals

Find the IPR number and byte offset for interrupt m as follows:
•
the corresponding IPR number, see Table 4-8 on page 4-7 n is given by n = m DIV 4
•
the byte offset of the required Priority field in this register is m MOD 4, where:
— byte offset 0 refers to register bits[7:0]
— byte offset 1 refers to register bits[15:8]
— byte offset 2 refers to register bits[23:16]
— byte offset 3 refers to register bits[31:24].
4.2.8

Software Trigger Interrupt Register
Write to the STIR to generate an interrupt from software. See the register summary in Table 4-2
on page 4-3 for the STIR attributes.
When the USERSETMPEND bit in the SCR is set to 1, unprivileged software can access the
STIR, see System Control Register on page 4-19.
Note
Only privileged software can enable unprivileged access to the STIR.
The bit assignments are:
31

9 8
Reserved

0
INTID

Table 4-10 STIR bit assignments

4.2.9

Bits

Field

Function

[31:9]

Reserved.

[8:0]

INTID

Interrupt ID of the interrupt to trigger, in
the range 0-239. For example, a value of
0x03 specifies interrupt IRQ3.

Level-sensitive and pulse interrupts
A Cortex-M4 device can support both level-sensitive and pulse interrupts. Pulse interrupts are
also described as edge-triggered interrupts.
A level-sensitive interrupt is held asserted until the peripheral deasserts the interrupt signal.
Typically this happens because the ISR accesses the peripheral, causing it to clear the interrupt
request. A pulse interrupt is an interrupt signal sampled synchronously on the rising edge of the
processor clock. To ensure the NVIC detects the interrupt, the peripheral must assert the
interrupt signal for at least one clock cycle, during which the NVIC detects the pulse and latches
the interrupt.
When the processor enters the ISR, it automatically removes the pending state from the
interrupt, see Hardware and software control of interrupts on page 4-9. For a level-sensitive
interrupt, if the signal is not deasserted before the processor returns from the ISR, the interrupt
becomes pending again, and the processor must execute its ISR again. This means that the
peripheral can hold the interrupt signal asserted until it no longer requires servicing.
See the documentation supplied by your device vendor for details of which interrupts are
level-based and which are pulsed.

ARM DUI 0553A
ID121610

4-8

Cortex-M4 Peripherals

Hardware and software control of interrupts
The Cortex-M4 latches all interrupts. A peripheral interrupt becomes pending for one of the
following reasons:
•

the NVIC detects that the interrupt signal is HIGH and the interrupt is not active

•

the NVIC detects a rising edge on the interrupt signal

•

software writes to the corresponding interrupt set-pending register bit, see Interrupt
Set-pending Registers on page 4-5, or to the STIR to make an interrupt pending, see
Software Trigger Interrupt Register on page 4-8.

A pending interrupt remains pending until one of the following:
•

•

The processor enters the ISR for the interrupt. This changes the state of the interrupt from
pending to active. Then:
—

For a level-sensitive interrupt, when the processor returns from the ISR, the NVIC
samples the interrupt signal. If the signal is asserted, the state of the interrupt
changes to pending, which might cause the processor to immediately re-enter the
ISR. Otherwise, the state of the interrupt changes to inactive.

—

For a pulse interrupt, the NVIC continues to monitor the interrupt signal, and if this
is pulsed the state of the interrupt changes to pending and active. In this case, when
the processor returns from the ISR the state of the interrupt changes to pending,
which might cause the processor to immediately re-enter the ISR.
If the interrupt signal is not pulsed while the processor is in the ISR, when the
processor returns from the ISR the state of the interrupt changes to inactive.

Software writes to the corresponding interrupt clear-pending register bit.
For a level-sensitive interrupt, if the interrupt signal is still asserted, the state of the
interrupt does not change. Otherwise, the state of the interrupt changes to inactive.
For a pulse interrupt, state of the interrupt changes to:
— inactive, if the state was pending
— active, if the state was active and pending.

4.2.10

NVIC usage hints and tips
Ensure software uses correctly aligned register accesses. The processor does not support
unaligned accesses to NVIC registers. See the individual register descriptions for the supported
access sizes.
A interrupt can enter pending state even if it is disabled. Disabling an interrupt only prevents the
processor from taking that interrupt.
Before programming VTOR to relocate the vector table, ensure the vector table entries of the
new vector table are setup for fault handlers, NMI and all enabled exception like interrupts. For
more information see Vector Table Offset Register on page 4-16.
NVIC programming hints
Software uses the CPSIE I and CPSID I instructions to enable and disable interrupts. The
CMSIS provides the following intrinsic functions for these instructions:
void __disable_irq(void) // Disable Interrupts
void __enable_irq(void) // Enable Interrupts

ARM DUI 0553A
ID121610

4-9

Cortex-M4 Peripherals

In addition, the CMSIS provides a number of functions for NVIC control, including:
Table 4-11 CMSIS functions for NVIC control
CMSIS interrupt control function

Description

void NVIC_SetPriorityGrouping(uint32_t priority_grouping)

Set the priority grouping

void NVIC_EnableIRQ(IRQn_t IRQn)

Enable IRQn

void NVIC_DisableIRQ(IRQn_t IRQn)

Disable IRQn

uint32_t NVIC_GetPendingIRQ (IRQn_t IRQn)

Return true (IRQ-Number) if IRQn is pending

void NVIC_SetPendingIRQ (IRQn_t IRQn)

Set IRQn pending

void NVIC_ClearPendingIRQ (IRQn_t IRQn)

Clear IRQn pending status

uint32_t NVIC_GetActive (IRQn_t IRQn)

Return the IRQ number of the active interrupt

void NVIC_SetPriority (IRQn_t IRQn, uint32_t priority)

Set priority for IRQn

uint32_t NVIC_GetPriority (IRQn_t IRQn)

Read priority of IRQn

void NVIC_SystemReset (void)

Reset the system

The input parameter IRQn is the IRQ number, see Table 2-16 on page 2-22. For more
information about these functions see the CMSIS documentation.

ARM DUI 0553A
ID121610

4-10

Cortex-M4 Peripherals

4.3

System control block
The System Control Block (SCB) provides system implementation information, and system
control. This includes configuration, control, and reporting of the system exceptions. The
system control block registers are:
Table 4-12 Summary of the system control block registers

Address

Name

Type

Required
privilege

Reset value

Description

0xE000E008

ACTLR

Privileged

0x00000000

Auxiliary Control Register

0xE000ED00

CPUID

Privileged

0x410FC240

CPUID Base Register on page 4-13

0xE000ED04

ICSR

RWa

Privileged

0x00000000

Interrupt Control and State Register on page 4-13

0xE000ED08

VTOR

Privileged

0x00000000

Vector Table Offset Register on page 4-16

0xE000ED0C

AIRCR

RW a

Privileged

0xFA050000

Application Interrupt and Reset Control Register on page 4-16

0xE000ED10

SCR

Privileged

0x00000000

System Control Register on page 4-19

0xE000ED14

CCR

Privileged

0x00000200

Configuration and Control Register on page 4-19

0xE000ED18

SHPR1

Privileged

0x00000000

System Handler Priority Register 1 on page 4-21

0xE000ED1C

SHPR2

Privileged

0x00000000

System Handler Priority Register 2 on page 4-22

0xE000ED20

SHPR3

Privileged

0x00000000

System Handler Priority Register 3 on page 4-22

0xE000ED24

SHCRS

Privileged

0x00000000

System Handler Control and State Register on page 4-23

0xE000ED28

CFSR

Privileged

0x00000000

Configurable Fault Status Register on page 4-24

0xE000ED28

MMSR b

Privileged

0x00

MemManage Fault Status Register on page 4-25

0xE000ED29

BFSR b

Privileged

0x00

BusFault Status Register on page 4-26

0xE000ED2A

UFSR b

Privileged

0x0000

UsageFault Status Register on page 4-28

0xE000ED2C

HFSR

Privileged

0x00000000

HardFault Status Register on page 4-30

0xE000ED34

MMAR

Privileged

Unknown

MemManage Fault Address Register on page 4-30

0xE000ED38

BFAR

Privileged

Unknown

BusFault Address Register on page 4-31

0xE000ED3C

AFSR

Privileged

0x00000000

Auxiliary Fault Status Register on page 4-31

a. See the register description for more information.
b. A subregister of the CFSR.

4.3.1

Auxiliary Control Register
The ACTLR provides disable bits for the following processor functions:
•
IT folding
•
write buffer use for accesses to the default memory map
•
interruption of multi-cycle instructions.
By default this register is set to provide optimum performance from the Cortex-M4 processor,
and does not normally require modification.

ARM DUI 0553A
ID121610

4-11

Cortex-M4 Peripherals

See the register summary in Table 4-12 on page 4-11 for the ACTLR attributes. The bit
assignments are:
10 9 8 7

31
Reserved

3 2 1 0
Reserved

DISOOFP
DISFPCA

DISFOLD
DISDEFWBUF
DISMCYCINT

Table 4-13 ACTLR bit assignments
Bits

Name

Function

[31:10]

Reserved.

[9]

DISOOFPa

Disables floating point instructions completing out of order with respect to integer instructions.

[8]

DISFPCAa

Disables automatic update of CONTROL.FPCA.

[7:3]

Reserved.

[2]

DISFOLD

When set to 1, disables IT folding. see About IT folding for more information.

[1]

DISDEFWBUF

When set to 1, disables write buffer use during default memory map accesses. This causes all
BusFaults to be precise BusFaults but decreases performance because any store to memory must
complete before the processor can execute the next instruction.

Note
This bit only affects write buffers implemented in the Cortex-M4 processor.
[0]

DISMCYCINT

When set to 1, disables interruption of load multiple and store multiple instructions. This increases
the interrupt latency of the processor because any LDM or STM must complete before the
processor can stack the current state and enter the interrupt handler.

a. Only implemented in a Cortex-M4F device

About IT folding
In some situations, the processor can start executing the first instruction in an IT block while it
is still executing the IT instruction. This behavior is called IT folding, and improves
performance, However, IT folding can cause jitter in looping. If a task must avoid jitter, set the
DISFOLD bit to 1 before executing the task, to disable IT folding.

ARM DUI 0553A
ID121610

4-12

Cortex-M4 Peripherals

4.3.2

CPUID Base Register
The CPUID register contains the processor part number, version, and implementation
information. See the register summary in Table 4-12 on page 4-11 for its attributes. The bit
assignments are:
31

24 23
Implementer

20 19

Variant

16 15

Constant

4 3
PartNo

Revision

Table 4-14 CPUID register bit assignments
Bits

Name

[31:24]

Implementer

Function
Implementer code:
0x41 = ARM

4.3.3

[23:20]

Variant

Variant number, the r value in the rnpn product revision identifier:
0x0 = Revision 0

[19:16]

Constant

Reads as 0xF

[15:4]

PartNo

Part number of the processor:
0xC24 = Cortex-M4

[3:0]

Revision

Revision number, the p value in the rnpn product revision identifier:
0x0 = Patch 0

Interrupt Control and State Register
The ICSR:
•
provides:
— a set-pending bit for the Non-Maskable Interrupt (NMI) exception
— set-pending and clear-pending bits for the PendSV and SysTick exceptions
•
indicates:
— the exception number of the exception being processed
— whether there are preempted active exceptions
— the exception number of the highest priority pending exception
— whether any interrupts are pending.

ARM DUI 0553A
ID121610

4-13

Cortex-M4 Peripherals

See the register summary in Table 4-12 on page 4-11, and the Type descriptions in Table 4-15,
for the ICSR attributes. The bit assignments are:
31 30 29 28 27 26 25 24 23 22 21

12 11 10 9 8
VECTPENDING

0
VECTACTIVE

ISRPENDING
Reserved for Debug
Reserved
PENDSTCLR
PENDSTSET
PENDSVCLR
PENDSVSET
Reserved
NMIPENDSET

Reserved
RETTOBASE

Table 4-15 ICSR bit assignments
Bits

Name

Type

Function

[31]

NMIPENDSET

NMI set-pending bit.
Write:
0 = no effect
1 = changes NMI exception state to pending.
Read:
0 = NMI exception is not pending
1 = NMI exception is pending.
Because NMI is the highest-priority exception, normally the processor enter the NMI
exception handler as soon as it registers a write of 1 to this bit, and entering the handler clears
this bit to 0. A read of this bit by the NMI exception handler returns 1 only if the NMI signal
is reasserted while the processor is executing that handler.

[30:29]

Reserved.

[28]

PENDSVSET

PendSV set-pending bit.
Write:
0 = no effect
1 = changes PendSV exception state to pending.
Read:
0 = PendSV exception is not pending
1 = PendSV exception is pending.
Writing 1 to this bit is the only way to set the PendSV exception state to pending.

[27]

PENDSVCLR

PendSV clear-pending bit.
Write:
0 = no effect
1 = removes the pending state from the PendSV exception.

[26]

PENDSTSET

SysTick exception set-pending bit.
Write:
0 = no effect
1 = changes SysTick exception state to pending.
Read:
0 = SysTick exception is not pending
1 = SysTick exception is pending.

ARM DUI 0553A
ID121610

4-14

Cortex-M4 Peripherals

Table 4-15 ICSR bit assignments (continued)
Bits

Name

Type

Function

[25]

PENDSTCLR

SysTick exception clear-pending bit.
Write:
0 = no effect
1 = removes the pending state from the SysTick exception.
This bit is WO. On a register read its value is Unknown.

[24]

Reserved.

[23]

Reserved for
Debug use

This bit is reserved for Debug use and reads-as-zero when the processor is not in Debug.

[22]

ISRPENDING

Interrupt pending flag, excluding NMI and Faults:
0 = interrupt not pending
1 = interrupt pending.

[21:18]

Reserved.

[17:12]

VECTPENDING

Indicates the exception number of the highest priority pending enabled exception:
0 = no pending exceptions
Nonzero = the exception number of the highest priority pending enabled exception.
The value indicated by this field includes the effect of the BASEPRI and FAULTMASK
registers, but not any effect of the PRIMASK register.

[11]

RETTOBASE

Indicates whether there are preempted active exceptions:
0 = there are preempted active exceptions to execute
1 = there are no active exceptions, or the currently-executing exception is the only active
exception.

[10:9]

Reserved.

[8:0]

VECTACTIVEa

Contains the active exception number:
0 = Thread mode
Nonzero = The exception numbera of the currently active exception.

Note
Subtract 16 from this value to obtain the CMSIS IRQ number required to index into the
Interrupt Clear-Enable, Set-Enable, Clear-Pending, Set-Pending, or Priority Registers, see
Table 2-5 on page 2-6.
a. This is the same value as IPSR bits[8:0], see Interrupt Program Status Register on page 2-6.

When you write to the ICSR, the effect is Unpredictable if you:
•
write 1 to the PENDSVSET bit and write 1 to the PENDSVCLR bit
•
write 1 to the PENDSTSET bit and write 1 to the PENDSTCLR bit.

ARM DUI 0553A
ID121610

4-15

Cortex-M4 Peripherals

4.3.4

Vector Table Offset Register
The VTOR indicates the offset of the vector table base address from memory address
0x00000000. See the register summary in Table 4-12 on page 4-11 for its attributes. The bit
assignments are:
31

7 6
TBLOFF

0
Reserved

Table 4-16 VTOR bit assignments
Bits

Name

Function

[31:7]

TBLOFF

Vector table base offset field. It contains bits[29:7] of the offset of the table base from the bottom
of the memory map.

Note
Bit[29] determines whether the vector table is in the code or SRAM memory region:
•
0 = code
•
1 = SRAM.
In implementations bit[29] is sometimes called the TBLBASE bit.
[6:0]

Reserved.

When setting TBLOFF, you must align the offset to the number of exception entries in the vector
table. The minimum alignment is 32 words, enough for up to 16 interrupts. For more interrupts,
adjust the alignment by rounding up to the next power of two. For example, if you require 21
interrupts, the alignment must be on a 64-word boundary because the required table size is 37
words, and the next power of two is 64. See your vendor documentation for the alignment details
of your device.
Note
Table alignment requirements mean that bits[6:0] of the table offset are always zero.

4.3.5

Application Interrupt and Reset Control Register
The AIRCR provides priority grouping control for the exception model, endian status for data
accesses, and reset control of the system. See the register summary in Table 4-12 on page 4-11
and Table 4-17 on page 4-17 for its attributes.
To write to this register, you must write 0x5FA to the VECTKEY field, otherwise the processor
ignores the write.

ARM DUI 0553A
ID121610

4-16

Cortex-M4 Peripherals

The bit assignments are:
31

16 15 14
On read: VECTKEYSTAT
On write: VECTKEY

11 10

8 7

Reserved

ENDIANNESS

PRIGROUP

Reserved for Debug use

3 2 1 0
Reserved

SYSRESETREQ
VECTCLRACTIVE
VECTRESET

Table 4-17 AIRCR bit assignments
Bits

Name

Type

Function

[31:16]

Write: VECTKEYSTAT
Read: VECTKEY

[15]

ENDIANNESS

Data endianness bit is implementation defined:
0 = Little-endian
1 = Big-endian.

[14:11]

Reserved.

[10:8]

PRIGROUP

R/W

Interrupt priority grouping field is implementation defined. This field
determines the split of group priority from subpriority, see Binary point on
page 4-18.

[7:3]

Reserved.

[2]

SYSRESETREQ

System reset request bit is implementation defined:
0 = no system reset request
1 = asserts a signal to the outer system that requests a reset.
This is intended to force a large system reset of all major components
except for debug.
This bit reads as 0.
See you vendor documentation for more information about the use of this
signal in your implementation.

[1]

VECTCLRACTIVE

Reserved for Debug use. This bit reads as 0. When writing to the register
you must write 0 to this bit, otherwise behavior is Unpredictable.

[0]

VECTRESET

Reserved for Debug use. This bit reads as 0. When writing to the register
you must write 0 to this bit, otherwise behavior is Unpredictable.

ARM DUI 0553A
ID121610

4-17

Cortex-M4 Peripherals

Binary point
The PRIGROUP field indicates the position of the binary point that splits the PRI_n fields in
the Interrupt Priority Registers into separate group priority and subpriority fields. Table 4-18
shows how the PRIGROUP value controls this split. Implementations having fewer than 8-bits
of interrupt priority treat the least significant bits as zero
Table 4-18 Priority grouping
Interrupt priority level value, PRI_N[7:0]

Number of

PRIGROUP

Binary point a

Group priority bits

Subpriority bits

Group priorities

Subpriorities

0b000

bxxxxxxx.y

[7:1]

[0]

128

0b001

bxxxxxx.yy

[7:2]

[1:0]

0b010

bxxxxx.yyy

[7:3]

[2:0]

0b011

bxxxx.yyyy

[7:4]

[3:0]

0b100

bxxx.yyyyy

[7:5]

[4:0]

0b101

bxx.yyyyyy

[7:6]

[5:0]

0b110

bx.yyyyyyy

[7]

[6:0]

128

0b111

b.yyyyyyyy

None

[7:0]

256

a. PRI_n[7:0] field showing the binary point. x denotes a group priority field bit, and y denotes a subpriority field bit.

Note
Determining preemption of an exception uses only the group priority field, see Interrupt priority
grouping on page 2-25.

ARM DUI 0553A
ID121610

4-18

Cortex-M4 Peripherals

4.3.6

System Control Register
The SCR controls features of entry to and exit from low power state. See the register summary
in Table 4-12 on page 4-11 for its attributes. The bit assignments are:
31

5 4 3 2 1 0
Reserved

SEVONPEND
Reserved
SLEEPDEEP
SLEEPONEXIT
Reserved

Table 4-19 SCR bit assignments

4.3.7

Bits

Name

Function

[31:5]

Reserved.

[4]

SEVONPEND

Send Event on Pending bit:
0 = only enabled interrupts or events can wakeup the processor, disabled interrupts are
excluded
1 = enabled events and all interrupts, including disabled interrupts, can wakeup the
processor.
When an event or interrupt enters pending state, the event signal wakes up the processor
from WFE. If the processor is not waiting for an event, the event is registered and affects the
next WFE.
The processor also wakes up on execution of an SEV instruction or an external event.

[3]

Reserved.

[2]

SLEEPDEEP

Controls whether the processor uses sleep or deep sleep as its low power mode:
0 = sleep
1 = deep sleep.

[1]

SLEEPONEXIT

Indicates sleep-on-exit when returning from Handler mode to Thread mode:
0 = do not sleep when returning to Thread mode.
1 = enter sleep, or deep sleep, on return from an ISR.
Setting this bit to 1 enables an interrupt driven application to avoid returning to an empty
main application.

[0]

Reserved.

Configuration and Control Register
The CCR controls entry to Thread mode and enables:
•

the handlers for NMI, hard fault and faults escalated by FAULTMASK to ignore
BusFaults

•

trapping of divide by zero and unaligned accesses

•

access to the STIR by unprivileged software, see Software Trigger Interrupt Register on
page 4-8.

See the register summary in Table 4-12 on page 4-11 for the CCR attributes.

ARM DUI 0553A
ID121610

4-19

Cortex-M4 Peripherals

The bit assignments are:
31

10 9 8 7

5 4 3 2 1 0

Reserved

STKALIGN
BFHFNMIGN
Reserved
DIV_0_TRP
UNALIGN_TRP
Reserved
USERSETMPEND
NONBASETHRDENA

Table 4-20 CCR bit assignments
Bits

Name

Function

[31:10]

Reserved.

[9]

STKALIGN

Indicates stack alignment on exception entry:
0 = 4-byte aligned
1 = 8-byte aligned.
On exception entry, the processor uses bit[9] of the stacked PSR to indicate the
stack alignment. On return from the exception it uses this stacked bit to restore
the correct stack alignment.

[8]

BFHFNMIGN

Enables handlers with priority -1 or -2 to ignore data BusFaults caused by load
and store instructions. This applies to the hard fault, NMI, and FAULTMASK
escalated handlers:
0 = data bus faults caused by load and store instructions cause a lock-up
1 = handlers running at priority -1 and -2 ignore data bus faults caused by load
and store instructions.
Set this bit to 1 only when the handler and its data are in absolutely safe memory.
The normal use of this bit is to probe system devices and bridges to detect control
path problems and fix them.

[7:5]

Reserved.

[4]

DIV_0_TRP

Enables faulting or halting when the processor executes an SDIV or UDIV
instruction with a divisor of 0:
0 = do not trap divide by 0
1 = trap divide by 0.
When this bit is set to 0, a divide by zero returns a quotient of 0.

[3]

UNALIGN_TRP

Enables unaligned access traps:
0 = do not trap unaligned halfword and word accesses
1 = trap unaligned halfword and word accesses.
If this bit is set to 1, an unaligned access generates a UsageFault.
Unaligned LDM, STM, LDRD, and STRD instructions always fault irrespective of
whether UNALIGN_TRP is set to 1.

ARM DUI 0553A
ID121610

4-20

Cortex-M4 Peripherals

Table 4-20 CCR bit assignments (continued)

4.3.8

Bits

Name

Function

[2]

Reserved.

[1]

USERSETMPEND

Enables unprivileged software access to the STIR, see Software Trigger Interrupt
Register on page 4-8:
0 = disable
1 = enable.

[0]

NONBASETHRDENA

Indicates how the processor enters Thread mode:
0 = processor can enter Thread mode only when no exception is active.
1 = processor can enter Thread mode from any level under the control of an
EXC_RETURN value, see Exception return on page 2-28.

System Handler Priority Registers
The SHPR1-SHPR3 registers set the priority level, 0 to 255, of the exception handlers that have
configurable priority.
SHPR1-SHPR3 are byte accessible. See the register summary in Table 4-12 on page 4-11 for
their attributes.
To access to the system exception priority level using CMSIS, use the following CMSIS
functions:
•
uint32_t NVIC_GetPriority(IRQn_Type IRQn)
•
void NVIC_SetPriority(IRQn_Type IRQn, uint32_t priority)
The input parameter IRQn is the IRQ number, see Table 2-16 on page 2-22 for more
information.
System Handler Priority Register 1
The bit assignments are:
31

24 23
Reserved

16 15
PRI_6

8 7
PRI_5

0
PRI_4

Table 4-21 SHPR1 register bit assignments

ARM DUI 0553A
ID121610

Bits

Name

Function

[31:24]

PRI_7

Reserved.

[23:16]

PRI_6

Priority of system handler 6, UsageFault

[15:8]

PRI_5

Priority of system handler 5, BusFault

[7:0]

PRI_4

Priority of system handler 4, MemManage

4-21

Cortex-M4 Peripherals

System Handler Priority Register 2
The bit assignments are:
31

24 23

PRI_11

Reserved

Table 4-22 SHPR2 register bit assignments
Bits

Name

Function

[31:24]

PRI_11

Priority of system handler 11, SVCall

[23:0]

Reserved.

System Handler Priority Register 3
The bit assignments are:
31

24 23
PRI_15

16 15
PRI_14

0
Reserved

Table 4-23 SHPR3 register bit assignments

ARM DUI 0553A
ID121610

Bits

Name

Function

[31:24]

PRI_15

Priority of system handler 15, SysTick exception

[23:16]

PRI_14

Priority of system handler 14, PendSV

[15:0]

Reserved.

4-22

Cortex-M4 Peripherals

4.3.9

System Handler Control and State Register
The SHCSR enables the system handlers, and indicates:
•
the pending status of the BusFault, MemManage fault, and SVC exceptions
•
the active status of the system handlers.
See the register summary in Table 4-12 on page 4-11 for the SHCSR attributes. The bit
assignments are:
31

19 18 17 16 15 14 13 12 11 10 9 8 7 6

4 3 2 1 0

Reserved

MEMFAULTACT
BUSFAULTACT
Reserved
USGFAULTACT
Reserved

USGFAULTENA
BUSFAULTENA
MEMFAULTENA
SVCALLPENDED
BUSFAULTPENDED
MEMFAULTPENDED
USGFAULTPENDED
SYSTICKACT
PENDSVACT
Reserved
MONITORACT
SVCALLACT

Table 4-24 SHCSR bit assignments
Bits

Name

Function

[31:19]

Reserved.

[18]

USGFAULTENA

UsageFault enable bit, set to 1 to enable a

[17]

BUSFAULTENA

BusFault enable bit, set to 1 to enable a

[16]

MEMFAULTENA

MemManage enable bit, set to 1 to enable a

[15]

SVCALLPENDED

SVCall pending bit, reads as 1 if exception is pending b

[14]

BUSFAULTPENDED

BusFault exception pending bit, reads as 1 if exception is pending b

[13]

MEMFAULTPENDED

MemManage exception pending bit, reads as 1 if exception is pending b

[12]

USGFAULTPENDED

UsageFault exception pending bit, reads as 1 if exception is pending b

[11]

SYSTICKACT

SysTick exception active bit, reads as 1 if exception is active c

[10]

PENDSVACT

PendSV exception active bit, reads as 1 if exception is active

[9]

Reserved.

[8]

MONITORACT

Debug monitor active bit, reads as 1 if Debug monitor is active

[7]

SVCALLACT

SVCall active bit, reads as 1 if SVC call is active

[6:4]

Reserved.

[3]

USGFAULTACT

UsageFault exception active bit, reads as 1 if exception is active

ARM DUI 0553A
ID121610

4-23

Cortex-M4 Peripherals

Table 4-24 SHCSR bit assignments (continued)
Bits

Name

Function

[2]

Reserved.

[1]

BUSFAULTACT

BusFault exception active bit, reads as 1 if exception is active

[0]

MEMFAULTACT

MemManage exception active bit, reads as 1 if exception is active

a. Enable bits, set to 1 to enable the exception, or set to 0 to disable the exception.
b. Pending bits, read as 1 if the exception is pending, or as 0 if it is not pending. You can write to these bits to change the pending
status of the exceptions.
c. Active bits, read as 1 if the exception is active, or as 0 if it is not active. You can write to these bits to change the active status
of the exceptions, but see the Caution in this section.

If you disable a system handler and the corresponding fault occurs, the processor treats the fault
as a hard fault.
You can write to this register to change the pending or active status of system exceptions. An
OS kernel can write to the active bits to perform a context switch that changes the current
exception type.
Caution
Software that changes the value of an active bit in this register without correct adjustment
to the stacked content can cause the processor to generate a fault exception. Ensure
software that writes to this register retains and subsequently restores the current active
status.

•

4.3.10

After you have enabled the system handlers, if you have to change the value of a bit in this
register you must use a read-modify-write procedure to ensure that you change only the
required bit.

Configurable Fault Status Register
The CFSR indicates the cause of a MemManage fault, BusFault, or UsageFault. See the register
summary in Table 4-12 on page 4-11 for its attributes. The bit assignments are:
31

16 15

8 7

Usage Fault Status Register

Bus Fault Status
Register

Memory Management
Fault Status Register

UFSR

BFSR

MMFSR

The following subsections describe the subregisters that make up the CFSR:
•
MemManage Fault Status Register on page 4-25
•
BusFault Status Register on page 4-26
•
UsageFault Status Register on page 4-28.
The CFSR is byte accessible. You can access the CFSR or its subregisters as follows:
•
access the complete CFSR with a word access to 0xE000ED28
•
access the MMFSR with a byte access to 0xE000ED28
•
access the MMFSR and BFSR with a halfword access to 0xE000ED28
•
access the BFSR with a byte access to 0xE000ED29
•
access the UFSR with a halfword access to 0xE000ED2A.

ARM DUI 0553A
ID121610

4-24

Cortex-M4 Peripherals

MemManage Fault Status Register
The flags in the MMFSR indicate the cause of memory access faults. The bit assignments are:
7 6 5 4 3 2 1 0

MMARVALID
Reserved
MLSPERR
MSTKERR

IACCVIOL
DACCVIOL
Reserved
MUNSTKERR

Table 4-25 MMFSR bit assignments
Bits

Name

Function

[7]

MMARVALID

MemManage Fault Address Register (MMFAR) valid flag:
0 = value in MMAR is not a valid fault address
1 = MMAR holds a valid fault address.
If a MemManage fault occurs and is escalated to a HardFault because of priority, the HardFault
handler must set this bit to 0. This prevents problems on return to a stacked active MemManage
fault handler whose MMAR value has been overwritten.

[6]

Reserved.

[5]

MLSPERRa

0 = no MemManage fault occurred during floating-point lazy state preservation
1 = a MemManage fault occurred during floating-point lazy state preservation.

[4]

MSTKERR

MemManage fault on stacking for exception entry:
0 = no stacking fault
1 = stacking for an exception entry has caused one or more access violations.
When this bit is 1, the SP is still adjusted but the values in the context area on the stack might
be incorrect. The processor has not written a fault address to the MMAR.

[3]

MUNSTKERR

MemManage fault on unstacking for a return from exception:
0 = no unstacking fault
1 = unstack for an exception return has caused one or more access violations.
This fault is chained to the handler. This means that when this bit is 1, the original return stack
is still present. The processor has not adjusted the SP from the failing return, and has not
performed a new save. The processor has not written a fault address to the MMAR.

[2]

Reserved.

[1]

DACCVIOL

Data access violation flag:
0 = no data access violation fault
1 = the processor attempted a load or store at a location that does not permit the operation.
When this bit is 1, the PC value stacked for the exception return points to the faulting
instruction. The processor has loaded the MMAR with the address of the attempted access.

[0]

IACCVIOL

Instruction access violation flag:
0 = no instruction access violation fault
1 = the processor attempted an instruction fetch from a location that does not permit execution.
This fault occurs on any access to an XN region, even when the MPU is disabled or not present.
When this bit is 1, the PC value stacked for the exception return points to the faulting
instruction. The processor has not written a fault address to the MMAR.

a. Only present in a Cortex-M4F device.

ARM DUI 0553A
ID121610

4-25

Cortex-M4 Peripherals

BusFault Status Register
The flags in the BFSR indicate the cause of a bus access fault. The bit assignments are:
7 6 5 4 3 2 1 0

BFARVALID
Reserved
LSPERR
STKERR

IBUSERR
PRECISERR
IMPRECISERR
UNSTKERR

Table 4-26 BFSR bit assignments
Bits

Name

Function

[7]

BFARVALID

BusFault Address Register (BFAR) valid flag:
0 = value in BFAR is not a valid fault address
1 = BFAR holds a valid fault address.
The processor sets this bit to 1 after a BusFault where the address is known. Other faults can set this bit to
0, such as a MemManage fault occurring later.
If a BusFault occurs and is escalated to a hard fault because of priority, the hard fault handler must set this
bit to 0. This prevents problems if returning to a stacked active BusFault handler whose BFAR value has
been overwritten.

[6]

Reserved.

[5]

LSPERRa

0 = no bus fault occurred during floating-point lazy state preservation
1 = a bus fault occurred during floating-point lazy state preservation.

[4]

STKERR

BusFault on stacking for exception entry:
0 = no stacking fault
1 = stacking for an exception entry has caused one or more BusFaults.
When the processor sets this bit to 1, the SP is still adjusted but the values in the context area on the stack
might be incorrect. The processor does not write a fault address to the BFAR.

[3]

UNSTKERR

BusFault on unstacking for a return from exception:
0 = no unstacking fault
1 = unstack for an exception return has caused one or more BusFaults.
This fault is chained to the handler. This means that when the processor sets this bit to 1, the original return
stack is still present. The processor does not adjust the SP from the failing return, does not performed a
new save, and does not write a fault address to the BFAR.

ARM DUI 0553A
ID121610

4-26

Cortex-M4 Peripherals

Table 4-26 BFSR bit assignments (continued)
Bits

Name

Function

[2]

IMPRECISERR

Imprecise data bus error:
0 = no imprecise data bus error
1 = a data bus error has occurred, but the return address in the stack frame is not related to the instruction
that caused the error.
When the processor sets this bit to 1, it does not write a fault address to the BFAR.
This is an asynchronous fault. Therefore, if it is detected when the priority of the current process is higher
than the BusFault priority, the BusFault becomes pending and becomes active only when the processor
returns from all higher priority processes. If a precise fault occurs before the processor enters the handler
for the imprecise BusFault, the handler detects both IMPRECISERR set to 1 and one of the precise fault
status bits set to 1.

[1]

PRECISERR

Precise data bus error:
0 = no precise data bus error
1 = a data bus error has occurred, and the PC value stacked for the exception return points to the instruction
that caused the fault.
When the processor sets this bit is 1, it writes the faulting address to the BFAR.

[0]

IBUSERR

Instruction bus error:
0 = no instruction bus error
1 = instruction bus error.
The processor detects the instruction bus error on prefetching an instruction, but it sets the IBUSERR flag
to 1 only if it attempts to issue the faulting instruction.
When the processor sets this bit is 1, it does not write a fault address to the BFAR.

a. Only present in a Cortex-M4F device.

ARM DUI 0553A
ID121610

4-27

Cortex-M4 Peripherals

UsageFault Status Register
The UFSR indicates the cause of a UsageFault. The bit assignments are:
15

10 9 8 7
Reserved

DIVBYZERO
UNALIGNED

4 3 2 1 0

Reserved

NOCP
INVPC
INVSTATE
UNDEFINSTR

Table 4-27 UFSR bit assignments
Bits

Name

Function

[15:10]

Reserved.

[9]

DIVBYZERO

Divide by zero UsageFault:
0 = no divide by zero fault, or divide by zero trapping not enabled
1 = the processor has executed an SDIV or UDIV instruction with a divisor of 0.
When the processor sets this bit to 1, the PC value stacked for the exception return points to the instruction
that performed the divide by zero.
Enable trapping of divide by zero by setting the DIV_0_TRP bit in the CCR to 1, see Configuration and
Control Register on page 4-19.

[8]

UNALIGNED

Unaligned access UsageFault:
0 = no unaligned access fault, or unaligned access trapping not enabled
1 = the processor has made an unaligned memory access.
Enable trapping of unaligned accesses by setting the UNALIGN_TRP bit in the CCR to 1, see
Configuration and Control Register on page 4-19.
Unaligned LDM, STM, LDRD, and STRD instructions always fault irrespective of the setting of UNALIGN_TRP.

[7:4]

Reserved.

[3]

NOCP

No coprocessor UsageFault. The processor does not support coprocessor instructions:
0 = no UsageFault caused by attempting to access a coprocessor
1 = the processor has attempted to access a coprocessor.

ARM DUI 0553A
ID121610

4-28

Cortex-M4 Peripherals

Table 4-27 UFSR bit assignments (continued)
Bits

Name

Function

[2]

INVPC

Invalid PC load UsageFault, caused by an invalid PC load by EXC_RETURN:
0 = no invalid PC load UsageFault
1 = the processor has attempted an illegal load of EXC_RETURN to the PC, as a result of an invalid
context, or an invalid EXC_RETURN value.
When this bit is set to 1, the PC value stacked for the exception return points to the instruction that tried
to perform the illegal load of the PC.

[1]

INVSTATE

Invalid state UsageFault:
0 = no invalid state UsageFault
1 = the processor has attempted to execute an instruction that makes illegal use of the EPSR.
When this bit is set to 1, the PC value stacked for the exception return points to the instruction that
attempted the illegal use of the EPSR.
This bit is not set to 1 if an undefined instruction uses the EPSR.

[0]

UNDEFINSTR

Undefined instruction UsageFault:
0 = no undefined instruction UsageFault
1 = the processor has attempted to execute an undefined instruction.
When this bit is set to 1, the PC value stacked for the exception return points to the undefined instruction.
An undefined instruction is an instruction that the processor cannot decode.

Note
The UFSR bits are sticky. This means as one or more fault occurs, the associated bits are set to
1. A bit that is set to 1 is cleared to 0 only by writing 1 to that bit, or by a reset.

ARM DUI 0553A
ID121610

4-29

Cortex-M4 Peripherals

4.3.11

HardFault Status Register
The HFSR gives information about events that activate the HardFault handler. See the register
summary in Table 4-12 on page 4-11 for its attributes.
This register is read, write to clear. This means that bits in the register read normally, but writing
1 to any bit clears that bit to 0. The bit assignments are:
31 30 29

2 1 0
Reserved

FORCED
DEBUGEVT

VECTTBL
Reserved

Table 4-28 HFSR bit assignments
Bits

Name

Function

[31]

DEBUGEVT

Reserved for Debug use. When writing to the register you must write 0 to this bit, otherwise
behavior is Unpredictable.

[30]

FORCED

Indicates a forced hard fault, generated by escalation of a fault with configurable priority that
cannot be handles, either because of priority or because it is disabled:
0 = no forced HardFault
1 = forced HardFault.
When this bit is set to 1, the HardFault handler must read the other fault status registers to find
the cause of the fault.

[29:2]

Reserved.

[1]

VECTTBL

Indicates a BusFault on a vector table read during exception processing:
0 = no BusFault on vector table read
1 = BusFault on vector table read.
This error is always handled by the hard fault handler.
When this bit is set to 1, the PC value stacked for the exception return points to the instruction
that was preempted by the exception.

[0]

Reserved.

Note
The HFSR bits are sticky. This means as one or more fault occurs, the associated bits are set to
1. A bit that is set to 1 is cleared to 0 only by writing 1 to that bit, or by a reset.

4.3.12

MemManage Fault Address Register
The MMFAR contains the address of the location that generated a MemManage fault. See the
register summary in Table 4-12 on page 4-11 for its attributes. The bit assignments are:
Table 4-29 MMFAR bit assignments
Bits

Name

Function

[31:0]

ADDRESS

When the MMARVALID bit of the MMFSR is set to 1, this field holds the address of the location
that generated the MemManage fault

ARM DUI 0553A
ID121610

4-30

Cortex-M4 Peripherals

When an unaligned access faults, the address is the actual address that faulted. Because a single
read or write instruction can be split into multiple aligned accesses, the fault address can be any
address in the range of the requested access size.
Flags in the MMFSR indicate the cause of the fault, and whether the value in the MMFAR is
valid. See MemManage Fault Status Register on page 4-25.
4.3.13

BusFault Address Register
The BFAR contains the address of the location that generated a BusFault. See the register
summary in Table 4-12 on page 4-11 for its attributes. The bit assignments are:
Table 4-30 BFAR bit assignments
Bits

Name

Function

[31:0]

ADDRESS

When the BFARVALID bit of the BFSR is set to 1, this field holds the address of the location that
generated the BusFault

When an unaligned access faults the address in the BFAR is the one requested by the instruction,
even if it is not the address of the fault.
Flags in the BFSR indicate the cause of the fault, and whether the value in the BFAR is valid.
See BusFault Status Register on page 4-26.
4.3.14

Auxiliary Fault Status Register
The AFSR contains additional system fault information. See the register summary in Table 4-12
on page 4-11 for its attributes.
This register is read, write to clear. This means that bits in the register read normally, but writing
1 to any bit clears that bit to 0.
The bit assignments are:
Table 4-31 AFSR bit assignments
Bits

Name

Function

[31:0]

IMPDEF

Implementation defined. The bits map to the AUXFAULT input signals.

Each AFSR bit maps directly to an AUXFAULT input of the processor, and a single-cycle
HIGH signal on the input sets the corresponding AFSR bit to one. It remains set to 1 until you
write 1 to the bit to clear it to zero. See your vendor documentation for more information.
When an AFSR bit is latched as one, an exception does not occur. Use an interrupt if an
exception is required.
4.3.15

System control block usage hints and tips
Ensure software uses aligned accesses of the correct size to access the system control block
registers:
•
except for the CFSR and SHPR1-SHPR3, it must use aligned word accesses
•
for the CFSR and SHPR1-SHPR3 it can use byte or aligned halfword or word accesses.
The processor does not support unaligned accesses to system control block registers.

ARM DUI 0553A
ID121610

4-31

Cortex-M4 Peripherals

In a fault handler. to determine the true faulting address:
1.

Read and save the MMFAR or BFAR value.

Read the MMARVALID bit in the MMFSR, or the BFARVALID bit in the BFSR. The
MMFAR or BFAR address is valid only if this bit is 1.

Software must follow this sequence because another higher priority exception might change the
MMFAR or BFAR value. For example, if a higher priority handler preempts the current fault
handler, the other fault might change the MMFAR or BFAR value.

ARM DUI 0553A
ID121610

4-32

Cortex-M4 Peripherals

4.4

System timer, SysTick
The processor has a 24-bit system timer, SysTick, that counts down from the reload value to
zero, reloads, that is wraps to, the value in the SYST_RVR register on the next clock edge, then
counts down on subsequent clocks.
Note
When the processor is halted for debugging the counter does not decrement.
The system timer registers are:
Table 4-32 System timer registers summary
Address

Name

Type

Required
privilege

Reset value

Description

0xE000E010

SYST_CSR

Privileged

SysTick Control and Status Register

0xE000E014

SYST_RVR

Privileged

Unknown

SysTick Reload Value Register on page 4-34

0xE000E018

SYST_CVR

Privileged

Unknown

SysTick Current Value Register on page 4-35

0xE000E01C

SYST_CALIB

Privileged

-a

SysTick Calibration Value Register on page 4-35

a. See the register description for more information.
4.4.1

SysTick Control and Status Register
The SysTick SYST_CSR register enables the SysTick features. The register resets to
0x00000000, or to 0x00000004 if your device does not implement a reference clock. See the
register summary in Table 4-32 for its attributes. The bit assignments are:
31

17 16 15
Reserved

3 2 1 0
Reserved

COUNTFLAG

0 0 0

CLKSOURCE
TICKINT
ENABLE

Table 4-33 SysTick SYST_CSR register bit assignments
Bits

Name

Function

[31:17]

Reserved.

[16]

COUNTFLAG

Returns 1 if timer counted to 0 since last time this was read.

[15:3]

Reserved.

ARM DUI 0553A
ID121610

4-33

Cortex-M4 Peripherals

Table 4-33 SysTick SYST_CSR register bit assignments (continued)
Bits

Name

Function

[2]

CLKSOURCE

Indicates the clock source:
0 = external clock
1 = processor clock.

[1]

TICKINT

Enables SysTick exception request:
0 = counting down to zero does not assert the SysTick exception request
1 = counting down to zero asserts the SysTick exception request.
Software can use COUNTFLAG to determine if SysTick has ever counted to zero.

[0]

ENABLE

Enables the counter:
0 = counter disabled
1 = counter enabled.

When ENABLE is set to 1, the counter loads the RELOAD value from the SYST_RVR register
and then counts down. On reaching 0, it sets the COUNTFLAG to 1 and optionally asserts the
SysTick depending on the value of TICKINT. It then loads the RELOAD value again, and
begins counting.
4.4.2

SysTick Reload Value Register
The SYST_RVR register specifies the start value to load into the SYST_CVR register. See the
register summary in Table 4-32 on page 4-33 for its attributes. The bit assignments are:
31

24 23
Reserved

RELOAD

Table 4-34 SYST_RVR register bit assignments
Bits

Name

Function

[31:24]

Reserved.

[23:0]

RELOAD

Value to load into the SYST_CVR register when the counter is enabled and when it reaches 0, see
Calculating the RELOAD value.

Calculating the RELOAD value
The RELOAD value can be any value in the range 0x00000001-0x00FFFFFF. A start value of 0 is
possible, but has no effect because the SysTick exception request and COUNTFLAG are
activated when counting from 1 to 0.
The RELOAD value is calculated according to its use. For example, to generate a multi-shot
timer with a period of N processor clock cycles, use a RELOAD value of N-1. If the SysTick
interrupt is required every 100 clock pulses, set RELOAD to 99.

ARM DUI 0553A
ID121610

4-34

Cortex-M4 Peripherals

4.4.3

SysTick Current Value Register
The SYST_CVR register contains the current value of the SysTick counter. See the register
summary in Table 4-32 on page 4-33 for its attributes. The bit assignments are:
31

24 23
Reserved

CURRENT

Table 4-35 SYST_CVR register bit assignments

4.4.4

Bits

Name

Function

[31:24]

Reserved.

[23:0]

CURRENT

Reads return the current value of the SysTick counter.
A write of any value clears the field to 0, and also clears the SYST_CSR COUNTFLAG bit to 0.

SysTick Calibration Value Register
The SYST_CALIB register indicates the SysTick calibration properties. See the register
summary in Table 4-32 on page 4-33 for its attributes. The reset value of this register is
implementation-defined. See the documentation supplied by your device vendor for more
information about the meaning of the SYST_CALIB field values. The bit assignments are:
31 30 29

24 23
Reserved

TENMS

SKEW
NOREF

Table 4-36 SYST_CALIB register bit assignments
Bits

Name

Function

[31]

NOREF

Indicates whether the device provides a reference clock to the processor:
0 = reference clock provided
1 = no reference clock provided.
If your device does not provide a reference clock, the SYST_CSR.CLKSOURCE bit reads-as-one
and ignores writes.

[30]

SKEW

Indicates whether the TENMS value is exact:
0 = TENMS value is exact
1 = TENMS value is inexact, or not given.
An inexact TENMS value can affect the suitability of SysTick as a software real time clock.

[29:24]

Reserved.

[23:0]

TENMS

Reload value for 10ms (100Hz) timing, subject to system clock skew errors. If the value reads as
zero, the calibration value is not known.

If calibration information is not known, calculate the calibration value required from the
frequency of the processor clock or external clock.

ARM DUI 0553A
ID121610

4-35

Cortex-M4 Peripherals

4.4.5

SysTick usage hints and tips
Some implementations stop all the processor clock signals during deep sleep mode. If this
happens, the SysTick counter stops.
Ensure software uses aligned word accesses to access the SysTick registers.
The SysTick counter reload and current value are not initialized by hardware. This means the
correct initialization sequence for the SysTick counter is:
1.
Program reload value.
2.
Clear current value.
3.
Program Control and Status register.

ARM DUI 0553A
ID121610

4-36

Cortex-M4 Peripherals

4.5

Optional Memory Protection Unit
This section describes the optional Memory Protection Unit (MPU).
The MPU divides the memory map into a number of regions, and defines the location, size,
access permissions, and memory attributes of each region. It supports:
•
independent attribute settings for each region
•
overlapping regions
•
export of memory attributes to the system.
The memory attributes affect the behavior of memory accesses to the region. The Cortex-M4
MPU defines:
•
eight separate memory regions, 0-7
•
a background region.
When memory regions overlap, a memory access is affected by the attributes of the region with
the highest number. For example, the attributes for region 7 take precedence over the attributes
of any region that overlaps region 7.
The background region has the same memory access attributes as the default memory map, but
is accessible from privileged software only.
The Cortex-M4 MPU memory map is unified. This means instruction accesses and data
accesses have same region settings.
If a program accesses a memory location that is prohibited by the MPU, the processor generates
a MemManage fault. This causes a fault exception, and might cause termination of the process
in an OS environment. In an OS environment, the kernel can update the MPU region setting
dynamically based on the process to be executed. Typically, an embedded OS uses the MPU for
memory protection.
Configuration of MPU regions is based on memory types, see Memory regions, types and
attributes on page 2-12.
Table 4-37 shows the possible MPU region attributes. These include Shareability and cache
behavior attributes are not relevant to most microcontroller implementations. See MPU
configuration for a microcontroller on page 4-47 and your vendor documentation for
programming guidelines if implemented.
Table 4-37 Memory attributes summary
Memory type

Shareability

Other attributes

Description

Strongly- ordered

All accesses to Strongly-ordered memory occur in program
order. All Strongly-ordered regions are assumed to be shared.

Device

Shared

Memory-mapped peripherals that several processors share.

Non-shared

Memory-mapped peripherals that only a single processor uses.

Shared

Non-cacheable
Write-through Cacheable
Write-back Cacheable

Normal memory that is shared between several processors.

Non-shared

Non-cacheable
Write-through Cacheable
Write-back Cacheable

Normal memory that only a single processor uses.

Normal

ARM DUI 0553A
ID121610

4-37

Cortex-M4 Peripherals

Use the MPU registers to define the MPU regions and their attributes. The MPU registers are:
Table 4-38 MPU registers summary
Address

Name

Type

Required
privilege

Reset
value

Description

0xE000ED90

MPU_TYPE

Privileged

0x00000800

MPU Type Register

0xE000ED94

MPU_CTRL

Privileged

0x00000000

MPU Control Register on page 4-39

0xE000ED98

MPU_RNR

Privileged

0x00000000

MPU Region Number Register on page 4-40

0xE000ED9C

MPU_RBAR

Privileged

0x00000000

MPU Region Base Address Register on page 4-40

0xE000EDA0

MPU_RASR

Privileged

0x00000000

MPU Region Attribute and Size Register on page 4-41

0xE000EDA4

MPU_RBAR_A1

Privileged

0x00000000

Alias of RBAR, see MPU Region Base Address Register on
page 4-40

0xE000EDA8

MPU_RASR_A1

Privileged

0x00000000

Alias of RASR, see MPU Region Attribute and Size Register
on page 4-41

0xE000EDAC

MPU_RBAR_A2

Privileged

0x00000000

Alias of RBAR, see MPU Region Base Address Register on
page 4-40

0xE000EDB0

MPU_RASR_A2

Privileged

0x00000000

Alias of RASR, see MPU Region Attribute and Size Register
on page 4-41

0xE000EDB4

MPU_RBAR_A3

Privileged

0x00000000

Alias of RBAR, see MPU Region Base Address Register on
page 4-40

0xE000EDB8

MPU_RASR_A3

Privileged

0x00000000

Alias of RASR, see MPU Region Attribute and Size Register
on page 4-41

4.5.1

MPU Type Register
The MPU_TYPE register indicates whether the MPU is present, and if so, how many regions it
supports. See the register summary in Table 4-38 for its attributes. The bit assignments are:
31

24 23
Reserved

16 15
IREGION

8 7
DREGION

1 0
Reserved

SEPARATE

Table 4-39 TYPE register bit assignments
Bits

Name

Function

[31:24]

Reserved.

[23:16]

IREGION

Indicates the number of supported MPU instruction regions.
Always contains 0x00. The MPU memory map is unified and is described by the DREGION field.

[15:8]

DREGION

Indicates the number of supported MPU data regions:
0x08 = Eight MPU regions.

[7:1]

Reserved.

[0]

SEPARATE

Indicates support for unified or separate instruction and date memory maps:
0 = unified.

ARM DUI 0553A
ID121610

4-38

Cortex-M4 Peripherals

4.5.2

MPU Control Register
The MPU_CTRL register:
•

enables the MPU

•

enables the default memory map background region

•

enables use of the MPU when in the hard fault, Non-maskable Interrupt (NMI), and
FAULTMASK escalated handlers.

See the register summary in Table 4-38 on page 4-38 for the MPU_CTRL attributes. The bit
assignments are:
31

3 2 1 0
Reserved

PRIVDEFENA
HFNMIENA
ENABLE

Table 4-40 MPU_CTRL register bit assignments
Bits

Name

Function

[31:3]

Reserved.

[2]

PRIVDEFENA

Enables privileged software access to the default memory map:
0 = If the MPU is enabled, disables use of the default memory map. Any memory access to a
location not covered by any enabled region causes a fault.
1 = If the MPU is enabled, enables use of the default memory map as a background region for
privileged software accesses.
When enabled, the background region acts as if it is region number -1. Any region that is
defined and enabled has priority over this default map.
If the MPU is disabled, the processor ignores this bit.

[1]

HFNMIENA

Enables the operation of MPU during hard fault, NMI, and FAULTMASK handlers.
When the MPU is enabled:
0 = MPU is disabled during hard fault, NMI, and FAULTMASK handlers, regardless of the
value of the ENABLE bit
1 = the MPU is enabled during hard fault, NMI, and FAULTMASK handlers.
When the MPU is disabled, if this bit is set to 1 the behavior is Unpredictable.

[0]

ENABLE

Enables the MPU:
0 = MPU disabled
1 = MPU enabled.

When ENABLE and PRIVDEFENA are both set to 1:
•

For privileged accesses, the default memory map is as described in Memory model on
page 2-12. Any access by privileged software that does not address an enabled memory
region behaves as defined by the default memory map.

•

Any access by unprivileged software that does not address an enabled memory region
causes a MemManage fault.

XN and Strongly-ordered rules always apply to the System Control Space regardless of the
value of the ENABLE bit.

ARM DUI 0553A
ID121610

4-39

Cortex-M4 Peripherals

When the ENABLE bit is set to 1, at least one region of the memory map must be enabled for
the system to function unless the PRIVDEFENA bit is set to 1. If the PRIVDEFENA bit is set
to 1 and no regions are enabled, then only privileged software can operate.
When the ENABLE bit is set to 0, the system uses the default memory map. This has the same
memory attributes as if the MPU is not implemented, see Table 2-11 on page 2-14. The default
memory map applies to accesses from both privileged and unprivileged software.
When the MPU is enabled, accesses to the System Control Space and vector table are always
permitted. Other areas are accessible based on regions and whether PRIVDEFENA is set to 1.
Unless HFNMIENA is set to 1, the MPU is not enabled when the processor is executing the
handler for an exception with priority –1 or –2. These priorities are only possible when handling
a hard fault or NMI exception, or when FAULTMASK is enabled. Setting the HFNMIENA bit
to 1 enables the MPU when operating with these two priorities.
4.5.3

MPU Region Number Register
The MPU_RNR selects which memory region is referenced by the MPU_RBAR and
MPU_RASR registers. See the register summary in Table 4-38 on page 4-38 for its attributes.
The bit assignments are:
31

8 7
Reserved

0
REGION

Table 4-41 MPU_RNR bit assignments
Bits

Name

Function

[31:8]

Reserved.

[7:0]

REGION

Indicates the MPU region referenced by the MPU_RBAR and MPU_RASR registers.
The MPU supports 8 memory regions, so the permitted values of this field are 0-7.

Normally, you write the required region number to this register before accessing the
MPU_RBAR or MPU_RASR. However you can change the region number by writing to the
MPU RBAR with the VALID bit set to 1, see MPU Region Base Address Register. This write
updates the value of the REGION field.
4.5.4

MPU Region Base Address Register
The MPU_RBAR defines the base address of the MPU region selected by the MPU_RNR, and
can update the value of the MPU_RNR. See the register summary in Table 4-38 on page 4-38
for its attributes.

ARM DUI 0553A
ID121610

4-40

Cortex-M4 Peripherals

Write MPU_RBAR with the VALID bit set to 1 to change the current region number and update
the MPU_RNR. The bit assignments are:
31

N N-1
ADDR

Reserved

If the region size is 32B, the ADDR field is bits [31:5] and there is no Reserved field

5 4 3

REGION

VALID

Table 4-42 MPU_RBAR bit assignments
Bits

Name

Function

[31:N]

ADDR

Region base address field. The value of N depends on the region size. For more information see
The ADDR field.

[(N-1):5]

Reserved.

[4]

VALID

MPU Region Number valid bit:
Write:
0 = MPU_RNR not changed, and the processor:
•
updates the base address for the region specified in the MPU_RNR
•
ignores the value of the REGION field
1 = the processor:
•
updates the value of the MPU_RNR to the value of the REGION field
•
updates the base address for the region specified in the REGION field.
Always reads as zero.

[3:0]

REGION

MPU region field:
For the behavior on writes, see the description of the VALID field.
On reads, returns the current region number, as specified by the RNR.

The ADDR field
The ADDR field is bits[31:N] of the MPU_RBAR. The region size, as specified by the SIZE
field in the MPU_RASR, defines the value of N:
N = Log2(Region size in bytes),
If the region size is configured to 4GB, in the MPU_RASR, there is no valid ADDR field. In
this case, the region occupies the complete memory map, and the base address is 0x00000000.
The base address is aligned to the size of the region. For example, a 64KB region must be
aligned on a multiple of 64KB, for example, at 0x00010000 or 0x00020000.
4.5.5

MPU Region Attribute and Size Register
The MPU_RASR defines the region size and memory attributes of the MPU region specified by
the MPU_RNR, and enables that region and any subregions. See the register summary in
Table 4-38 on page 4-38 for its attributes.
MPU_RASR is accessible using word or halfword accesses:

ARM DUI 0553A
ID121610

•

the most significant halfword holds the region attributes

•

the least significant halfword holds the region size and the region and subregion enable
bits.

4-41

Cortex-M4 Peripherals

The bit assignments are:
31

29 28 27 26

24 23 22 21
AP

Reserved
XN
Reserved

19 18 17 16 15

TEX

S C B

Reserved

8 7 6 5
SRD

1 0
SIZE

Reserved
ENABLE

Table 4-43 MPU_RASR bit assignments
Bits

Name

Function

[31:29]

Reserved.

[28]

Instruction access disable bit:
0 = instruction fetches enabled
1 = instruction fetches disabled.

[27]

Reserved.

[26:24]

Access permission field, see Table 4-47 on page 4-44.

[23:22]

Reserved.

[21:19, 17, 16]

TEX, C, B

Memory access attributes, see Table 4-45 on page 4-43.

[18]

Shareable bit, see Table 4-45 on page 4-43.

[15:8]

SRD

Subregion disable bits. For each bit in this field:
0 = corresponding sub-region is enabled
1 = corresponding sub-region is disabled
See Subregions on page 4-46 for more information.
Region sizes of 128 bytes and less do not support subregions. When writing the attributes
for such a region, write the SRD field as 0x00.

[7:6]

Reserved.

[5:1]

SIZE

Specifies the size of the MPU protection region. The minimum permitted value is 3
(0b00010). See SIZE field values for more information.

[0]

ENABLE

Region enable bit.

For information about access permission, see MPU access permission attributes on page 4-43.
SIZE field values
The SIZE field defines the size of the MPU memory region specified by the RNR. as follows:
(Region size in bytes) = 2(SIZE+1)

ARM DUI 0553A
ID121610

4-42

Cortex-M4 Peripherals

The smallest permitted region size is 32B, corresponding to a SIZE value of 4. Table 4-44 gives
example SIZE values, with the corresponding region size and value of N in the MPU_RBAR.
Table 4-44 Example SIZE field values
SIZE value

Region size

Value of N a

Note

0b00100 (4)

32B

Minimum permitted size

0b01001 (9)

1KB

0b10011 (19)

1MB

0b11101 (29)

1GB

0b11111 (31)

4GB

Maximum possible size

a. In the MPU_RBAR, see MPU Region Base Address Register on page 4-40.

4.5.6

MPU access permission attributes
This section describes the MPU access permission attributes. The access permission bits, TEX,
C, B, S, AP, and XN, of the RASR, control access to the corresponding memory region. If an
access is made to an area of memory without the required permissions, then the MPU generates
a permission fault. Table 4-45 shows encodings for the TEX, C, B, and S access permission bits.
Table 4-45 TEX, C, B, and S encoding
TEX

Memory type

Shareability

Other attributes

0b000

Strongly-ordered

Shareable

Device

Shareable

Normal

Not shareable

Outer and inner write-through. No write allocate.

1
1

Shareable
Normal

1
0b001

Normal

0b1BB

Not shareable

Outer and inner noncacheable.

Shareable

Reserved encoding

Implementation defined attributes.

Normal

Outer and inner write-back. Write and read allocate.

1
0b010

Outer and inner write-back. No write allocate.

Shareable

Not shareable

Not shareable
Shareable

Device

Reserved encoding

Normal

Not shareable

Not shareable
Shareable

Nonshared Device.

Cached memory, BB = outer policy, AA = inner policy.
See Table 4-46 on page 4-44 for the encoding of the AA
and BB bits.

a. The MPU ignores the value of this bit.

ARM DUI 0553A
ID121610

4-43

Cortex-M4 Peripherals

Table 4-46 shows the cache policy for memory attribute encodings with a TEX value is in the
range 4-7.
Table 4-46 Cache policy for memory attribute encoding
Encoding, AA or BB

Corresponding cache policy

Non-cacheable

Write back, write and read allocate

Write through, no write allocate

Write back, no write allocate

Table 4-47 shows the AP encodings that define the access permissions for privileged and
unprivileged software.
Table 4-47 AP encoding

4.5.7

AP[2:0]

Privileged
permissions

Unprivileged
permissions

Description

000

No access

All accesses generate a permission fault

001

No access

Access from privileged software only

010

Writes by unprivileged software generate a permission fault

011

Full access

100

Unpredictable

Reserved

101

No access

Reads by privileged software only

110

Read only, by privileged or unprivileged software

111

Read only, by privileged or unprivileged software

MPU mismatch
When an access violates the MPU permissions, the processor generates a MemManage fault, see
Exceptions and interrupts on page 2-10. The MMFSR indicates the cause of the fault. See
MemManage Fault Status Register on page 4-25 for more information.

4.5.8

Updating an MPU region
To update the attributes for an MPU region, update the MPU_RNR, MPU_RBAR and
MPU_RASR registers. You can program each register separately, or use a multiple-word write
to program all of these registers. You can use the MPU_RBAR and MPU_RASR aliases to
program up to four regions simultaneously using an STM instruction.
Updating an MPU region using separate words
Simple code to configure one region:
; R1 = region number
; R2 = size/enable
; R3 = attributes
; R4 = address
LDR R0,=MPU_RNR

ARM DUI 0553A
ID121610

; 0xE000ED98, MPU region number register

4-44

Cortex-M4 Peripherals

STR R1, [R0, #0x0]
STR R4, [R0, #0x4]
STRH R2, [R0, #0x8]
STRH R3, [R0, #0xA]

;
;
;
;

Region
Region
Region
Region

Number
Base Address
Size and Enable
Attribute

Disable a region before writing new region settings to the MPU if you have previously enabled
the region being changed. For example:
; R1 = region number
; R2 = size/enable
; R3 = attributes
; R4 = address
LDR R0,=MPU_RNR
STR R1, [R0, #0x0]
BIC R2, R2, #1
STRH R2, [R0, #0x8]
STR R4, [R0, #0x4]
STRH R3, [R0, #0xA]
ORR R2, #1
STRH R2, [R0, #0x8]

;
;
;
;
;
;
;
;

0xE000ED98, MPU region number register
Region Number
Disable
Region Size and Enable
Region Base Address
Region Attribute
Enable
Region Size and Enable

Software must use memory barrier instructions:
•

before MPU setup if there might be outstanding memory transfers, such as buffered
writes, that might be affected by the change in MPU settings

•

after MPU setup if it includes memory transfers that must use the new MPU settings.

However, memory barrier instructions are not required if the MPU setup process starts by
entering an exception handler, or is followed by an exception return, because the exception entry
and exception return mechanism cause memory barrier behavior.
Software does not require any memory barrier instructions during MPU setup, because it
accesses the MPU through the PPB, which is a Strongly-Ordered memory region.
For example, if you want all of the memory access behavior to take effect immediately after the
programming sequence, use a DSB instruction and an ISB instruction. A DSB is required after
changing MPU settings, such as at the end of context switch. An ISB is required if the code that
programs the MPU region or regions is entered using a branch or call. If the programming
sequence is entered using a return from exception, or by taking an exception, then you do not
require an ISB.
Updating an MPU region using multi-word writes
You can program directly using multi-word writes, depending on how the information is
divided. Consider the following reprogramming:
; R1 = region number
; R2 = address
; R3 = size, attributes in one
LDR R0, =MPU_RNR
; 0xE000ED98, MPU region number register
STR R1, [R0, #0x0] ; Region Number
STR R2, [R0, #0x4] ; Region Base Address
STR R3, [R0, #0x8] ; Region Attribute, Size and Enable

Use an STM instruction to optimize this:
; R1 = region number
; R2 = address
; R3 = size, attributes in one
LDR R0, =MPU_RNR
; 0xE000ED98, MPU region number register
STM R0, {R1-R3}
; Region Number, address, attribute, size and enable

ARM DUI 0553A
ID121610

4-45

Cortex-M4 Peripherals

You can do this in two words for pre-packed information. This means that the MPU_RBAR
contains the required region number and had the VALID bit set to 1, see MPU Region Base
Address Register on page 4-40. Use this when the data is statically packed, for example in a boot
loader:
; R1 = address and region number in one
; R2 = size and attributes in one
LDR R0, =MPU_RBAR
; 0xE000ED9C, MPU Region Base register
STR R1, [R0, #0x0] ; Region base address and
; region number combined with VALID (bit 4) set to 1
STR R2, [R0, #0x4] ; Region Attribute, Size and Enable

Subregions
Regions of 256 bytes or more are divided into eight equal-sized subregions. Set the
corresponding bit in the SRD field of the MPU_RASR to disable a subregion, see MPU Region
Attribute and Size Register on page 4-41. The least significant bit of SRD controls the first
subregion, and the most significant bit controls the last subregion. Disabling a subregion means
another region overlapping the disabled range matches instead. If no other enabled region
overlaps the disabled subregion the MPU issues a fault.
Regions of 32, 64, and 128 bytes do not support subregions, With regions of these sizes, you
must set the SRD field to 0x00, otherwise the MPU behavior is Unpredictable.
Example of SRD use

Two regions with the same base address overlap. Region one is 128KB, and region two is
512KB. To ensure the attributes from region one apply to the first 128KB region, set the SRD
field for region two to 0b00000011 to disable the first two subregions, as the figure shows.
Region 2, with
subregions

Region 1

Base address of both regions

4.5.9

Offset from
base address
512KB
448KB
384KB
320KB
256KB
192KB
128KB
Disabled subregion
64KB
Disabled subregion
0

MPU usage hints and tips
To avoid unexpected behavior, disable the interrupts before updating the attributes of a region
that the interrupt handlers might access.
Ensure software uses aligned accesses of the correct size to access MPU registers:
•
except for the MPU_RASR, it must use aligned word accesses
•
for the MPU_RASR it can use byte or aligned halfword or word accesses.
The processor does not support unaligned accesses to MPU registers.
When setting up the MPU, and if the MPU has previously been programmed, disable unused
regions to prevent any previous region settings from affecting the new MPU setup.

ARM DUI 0553A
ID121610

4-46

Cortex-M4 Peripherals

MPU configuration for a microcontroller
Usually, a microcontroller system has only a single processor and no caches. In such a system,
program the MPU as follows:
Table 4-48 Memory region attributes for a microcontroller
Memory region

TEX

Memory type and attributes

Flash memory

0b000

Normal memory, Non-shareable, write-through

Internal SRAM

0b000

Normal memory, Shareable, write-through

External SRAM

0b000

Normal memory, Shareable, write-back, write-allocate

Peripherals

0b000

Device memory, Shareable

In most microcontroller implementations, the shareability and cache policy attributes do not
affect the system behavior. However, using these settings for the MPU regions can make the
application code more portable. The values given are for typical situations. In special systems,
such as multiprocessor designs or designs with a separate DMA engine, the shareability attribute
might be important. In these cases see the recommendations of the memory device
manufacturer.

ARM DUI 0553A
ID121610

4-47

Cortex-M4 Peripherals

4.6

Floating Point Unit (FPU)
This section describes the optional Floating Point Unit (FPU) in a Cortex-M4F device. The FPU
implements the FPv4-SP floating-point extension.
The FPU fully supports single-precision add, subtract, multiply, divide, multiply and
accumulate, and square root operations. It also provides conversions between fixed-point and
floating-point data formats, and floating-point constant instructions.
The FPU provides floating-point computation functionality that is compliant with the
ANSI/IEEE Std 754-2008, IEEE Standard for Binary Floating-Point Arithmetic, referred to as
the IEEE 754 standard.
The FPU contains 32 single-precision extension registers, which you can also access as 16
doubleword registers for load, store, and move operations.
Table 4-49 shows the floating-point system registers in the Cortex-M4F FPU.
Table 4-49 Cortex-M4F floating-point system registers
Address

Name

Type

Reset

Description

0xE000ED88

CPACR

0x00000000

Coprocessor Access Control Register

0xE000EF34

FPCCR

0xC0000000

Floating-point Context Control Register on page 4-49

0xE000EF38

FPCAR

Floating-point Context Address Register on page 4-50

FPSCR

Floating-point Status Control Register on page 4-50

0xE000EF3C

FPDSCR

0x00000000

Floating-point Default Status Control Register on page 4-52

The following sections describe the floating-point system registers whose implementation is
specific to this processor.
4.6.1

Coprocessor Access Control Register
The CPACR register specifies the access privileges for coprocessors. See the register summary
in Cortex-M4F floating-point system registers for its attributes. The bit assignments are:
31 30 29 28 27 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 0
Reserved

CP11 CP10

Reserved

Table 4-50 CPACR register bit assignments
Bits

Name

Function

[31:24]

Reserved. Read as Zero, Write Ignore.

[2n+1:2n] for
n values10
and 11

CPn

Access privileges for coprocessor n. The possible values of each field are:
0b00 = Access denied. Any attempted access generates a NOCP

UsageFault.
0b01 = Privileged access only. An unprivileged access generates a NOCP

fault.
0b10 = Reserved. The result of any access is Unpredictable.
0b11 = Full access.
[19:0]

ARM DUI 0553A
ID121610

Reserved. Read as Zero, Write Ignore.

4-48

Cortex-M4 Peripherals

4.6.2

Floating-point Context Control Register
The FPCCR register sets or returns FPU control data. See the register summary in Cortex-M4F
floating-point system registers on page 4-48 for its attributes. The bit assignments are:
31 30 29

9 8 7 6 5 4 3 2 1 0
Reserved

LSPEN
ASPEN

MONRDY
Reserved
BFRDY
MMRDY
HFRDY
THREAD
Reserved
USER
LSPACT

Table 4-51 FPCCR register bit assignments
Bits

Name

Function

[31]

ASPEN

Enables CONTROL<2> setting on execution of a floating-point instruction. This results in automatic
hardware state preservation and restoration, for floating-point context, on exception entry and
exit.
0 = Disable CONTROL<2> setting on execution of a floating-point instruction.
1 = Enable CONTROL<2> setting on execution of a floating-point instruction.

[30]

LSPEN

0 = Disable automatic lazy state preservation for floating-point context.
1 = Enable automatic lazy state preservation for floating-point context.

[29:9]

Reserved.

[8]

MONRDY

0 = DebugMonitor is disabled or priority did not permit setting MON_PEND when the floating-point
stack frame was allocated.
1 = DebugMonitor is enabled and priority permits setting MON_PEND when the floating-point stack
frame was allocated.

[7]

Reserved.

[6]

BFRDY

0 = BusFault is disabled or priority did not permit setting the BusFault handler to the pending
state when the floating-point stack frame was allocated.
1 = BusFault is enabled and priority permitted setting the BusFault handler to the pending state
when the floating-point stack frame was allocated.

[5]

MMRDY

0 = MemManage is disabled or priority did not permit setting the MemManage handler to the
pending state when the floating-point stack frame was allocated.
1 = MemManage is enabled and priority permitted setting the MemManage handler to the
pending state when the floating-point stack frame was allocated.

[4]

HFRDY

0 = Priority did not permit setting the HardFault handler to the pending state when the
floating-point stack frame was allocated.
1 = Priority permitted setting the HardFault handler to the pending state when the floating-point
stack frame was allocated.

[3]

THREAD

0 = Mode was not Thread Mode when the floating-point stack frame was allocated.
1 = Mode was Thread Mode when the floating-point stack frame was allocated.

ARM DUI 0553A
ID121610

4-49

Cortex-M4 Peripherals

Table 4-51 FPCCR register bit assignments (continued)

4.6.3

Bits

Name

Function

[2]

Reserved.

[1]

USER

0 = Privilege level was not user when the floating-point stack frame was allocated.
1 = Privilege level was user when the floating-point stack frame was allocated.

[0]

LSPACT

0 = Lazy state preservation is not active.
1 = Lazy state preservation is active. floating-point stack frame has been allocated but saving
state to it has been deferred.

Floating-point Context Address Register
The FPCAR register holds the location of the unpopulated floating-point register space
allocated on an exception stack frame. See the register summary in Cortex-M4F floating-point
system registers on page 4-48 for its attributes. The bit assignments are:
31

3 2

ADDRESS
Reserved

Table 4-52 FPCAR register bit assignments

4.6.4

Bits

Name

Function

[31:3]

ADDRESS

The location of the unpopulated floating-point register
space allocated on an exception stack frame.

[2:0]

Reserved. Read as Zero, Writes Ignored.

Floating-point Status Control Register
The FPSCR register provides all necessary User level control of the floating-point system. The
bit assignments are:
31 30 29 28 27 26 25 24 23 22 21
N Z C V

8 7 6 5 4 3 2 1 0
Reserved

Reserved
AHP

RMode
FZ
DN

IDC
Reserved
IXC

IOC
DZC
OFC
UFC

Table 4-53 FPSCR bit assignments

ARM DUI 0553A
ID121610

Bits

Name

Function

[31]

[30]

[29]

[28]

Condition code flags. Floating-point comparison operations update these flags.
N
Negative condition code flag.
Z
Zero condition code flag.
C
Carry condition code flag.
V
Overflow condition code flag.

[27]

Reserved.

4-50

Cortex-M4 Peripherals

Table 4-53 FPSCR bit assignments (continued)

ARM DUI 0553A
ID121610

Bits

Name

Function

[26]

AHP

Alternative half-precision control bit:
0
IEEE half-precision format selected.
1
Alternative half-precision format selected.

[25]

Default NaN mode control bit:
0
NaN operands propagate through to the output of a floating-point operation.
1
Any operation involving one or more NaNs returns the Default NaN.

[24]

Flush-to-zero mode control bit:
0
Flush-to-zero mode disabled. Behavior of the floating-point system is fully
compliant with the IEEE 754 standard.
1
Flush-to-zero mode enabled.

[23:22]

RMode

Rounding Mode control field. The encoding of this field is:
0b00
Round to Nearest (RN) mode
0b01
Round towards Plus Infinity (RP) mode
0b10
Round towards Minus Infinity (RM) mode
0b11
Round towards Zero (RZ) mode.
The specified rounding mode is used by almost all floating-point instructions.

[21:8]

Reserved.

[7]

IDC

Input Denormal cumulative exception bit, see bits [4:0].

[6:5]

Reserved.

[4]

IXC

[3]

UFC

[2]

OFC

[1]

DZC

[0]

IOC

Cumulative exception bits for floating-point exceptions, see also bit [7]. Each of these bits is
set to 1 to indicate that the corresponding exception has occurred since 0 was last written to it.
IDC, bit[7]
Input Denormal cumulative exception bit.
IXC
Inexact cumulative exception bit.
UFC
Underflow cumulative exception bit.
OFC
Overflow cumulative exception bit.
DZC
Division by Zero cumulative exception bit.
IOC
Invalid Operation cumulative exception bit.

4-51

Cortex-M4 Peripherals

4.6.5

Floating-point Default Status Control Register
The FPDSCR register holds the default values for the floating-point status control data. See the
register summary in Cortex-M4F floating-point system registers on page 4-48 for its attributes.
The bit assignments are:
31

27 26 25 24 23 22 21
Reserved
AHP
DN

0 0 0 0 0

Reserved

RMode
FZ

Table 4-54 FPDSCR register bit assignments

4.6.6

Bits

Name

Function

[31:27]

Reserved

[26]

AHP

Default value for FPSCR.AHP

[25]

Default value for FPSCR.DN

[24]

Default value for FPSCR.FZ

[23:22]

RMode

Default value for FPSCR.RMode

[21:0]

Reserved

Enabling the FPU
The FPU is disabled from reset. You must enable it before you can use any floating-point
instructions. Example 4-1 shows an example code sequence for enabling the FPU in both
privileged and user modes. The processor must be in privileged mode to read from and write to
the CPACR.
Example 4-1 Enabling the FPU

; CPACR is located at address 0xE000ED88
LDR.W
R0, =0xE000ED88
; Read CPACR
LDR
R1, [R0]
; Set bits 20-23 to enable CP10 and CP11 coprocessors
ORR
R1, R1, #(0xF << 20)
; Write back the modified value to the CPACR
STR
R1, [R0]; wait for store to complete
DSB
;reset pipeline now the FPU is enabled
ISB

ARM DUI 0553A
ID121610

4-52

Appendix A
Cortex-M4 Options

This appendix describes the configuration options for a Cortex-M4 processor implementation. It
shows what features of a Cortex-M4 implementation are determined by the device manufacturer. It
contains the following section:
•
Cortex-M4 implementation options on page A-2.

ARM DUI 0553A
ID121610

A-1

Cortex-M4 Options

A.1

Cortex-M4 implementation options
Table A-1 shows the Cortex-M4 implementation options:
Table A-1 Effects of the Cortex-M4 implementation options
Option

Description, and affected documentation

Inclusion of
MPU

The implementer decides whether to include the Memory Protection Unit (MPU). See the Optional
Memory Protection Unit on page 4-37.

Inclusion of FPU

Only the Cortex-M4F includes the Floating Point Unit (FPU). See:
•
Floating-point instructions on page 3-126
•
Interruptible-continuable instructions in Core registers on page 2-3
•
The FPACTV bit in the CONTROL register
•
Table 2-17 on page 2-28
•
the MLSPERR bit in the MemManage Fault Status Register (MMFSR)
•
the LSPERR bit in the BusFault Status Register (BFSR).

Number of
interrupts

The implementer decides how many interrupts the Cortex-M4 implementation supports Cortex-M4
implementation supports, in the range 1-240. This affects:
The range of IRQ values in Table 2-5 on page 2-6
Entries in the last row of Table 2-16 on page 2-22, particularly if only one interrupt is implemented.
The maximum interrupt number, and associated information where appropriate, in:
•
Exception handlers on page 2-23
•
Figure 2-2 on page 2-24
•
Nested Vectored Interrupt Controller on page 4-3.
The number of implemented Nested Vectored Interrupt Controller (NVIC) registers in:
•
Table 4-2 on page 4-3
•
The appropriate register descriptions in sections Interrupt Set-enable Registers on page 4-4
to Interrupt Priority Registers on page 4-7.
Vector Table Offset Register on page 4-16, including the figure and Table 4-16 on page 4-16. See the
configuration information in the section for guidance on the required configuration.

Number of
priority bits

The implementer decides how many priority bits are implemented in priority value fields, in the range
3-8. This affects The maximum priority level value in Nested Vectored Interrupt Controller on
page 4-3.

Inclusion of the
WIC

The implementer decides whether to include the Wakeup interrupt Controller (WIC), see The optional
Wakeup Interrupt Controller on page 2-33.

Sleep mode
power-saving

The implementer decides what sleep modes to implement, and the power-saving measures associated
with any implemented mode, See Power management on page 2-32.
Sleep mode power saving might also affect the SysTick behavior, see SysTick usage hints and tips on
page 4-36.

The implementer decides whether all registers in the register bank can be reset. This affects the reset
values, see Table 2-2 on page 2-3.

Endianness

The implementer decides whether the memory system is little-endian or big-endian, see on
page 2-10Data types on page 2-10 and Memory endianness on page 2-18.

Memory features

Some features of the memory system are implementation-specific. This means that the Memory model
on page 2-12 cannot completely describe the memory map for a specific Cortex-M4 implementation.

ARM DUI 0553A
ID121610

A-2

Cortex-M4 Options

Table A-1 Effects of the Cortex-M4 implementation options (continued)
Option

Description, and affected documentation

Bit-banding

The implementer decides whether bit-banding is implemented., see Optional bit-banding on
page 2-16 and Memory model on page 2-12.

SysTick timer

The SYST_CALIB register is implementation- defined. This can affect:
•
•

ARM DUI 0553A
ID121610

SysTick Calibration Value Register on page 4-35
The entry for SYST_CALIB in Table 4-32 on page 4-33.

A-3

Glossary

This glossary describes some of the terms used in technical documents from ARM.
Abort

A mechanism that indicates to a processor that the value associated with a memory access is
invalid. An abort can be caused by the external or internal memory system as a result of attempting
to access invalid instruction or data memory.

Aligned

A data item stored at an address that is divisible by the number of bytes that defines the data size
is said to be aligned. Aligned words and halfwords have addresses that are divisible by four and
two respectively. The terms word-aligned and halfword-aligned therefore stipulate addresses that
are divisible by four and two respectively.

Banked register

A register that has multiple physical copies, where the state of the processor determines which copy
is used. The Stack Pointer, SP (R13) is a banked register.

Base register

In instruction descriptions, a register specified by a load or store instruction that is used to hold the
base value for the address calculation for the instruction. Depending on the instruction and its
addressing mode, an offset can be added to or subtracted from the base register value to form the
address that is sent to memory.
See also Index register.

Big-endian (BE)

Byte ordering scheme in which bytes of decreasing significance in a data word are stored at
increasing addresses in memory.
See also Byte-invariant, Endianness, Little-endian (LE).

Big-endian memory

Memory in which:
•

ARM DUI 0553A
ID121610

a byte or halfword at a word-aligned address is the most significant byte or halfword within
the word at that address

Glossary-1

Glossary

•

a byte at a halfword-aligned address is the most significant byte within the halfword at
that address.

See also Little-endian memory.
Breakpoint

A breakpoint is a mechanism provided by debuggers to identify an instruction at which program
execution is to be halted. Breakpoints are inserted by the programmer to enable inspection of
register contents, memory locations, variable values at fixed points in the program execution to
test that the program is operating correctly. Breakpoints are removed after the program is
successfully tested.

Byte-invariant

In a byte-invariant system, the address of each byte of memory remains unchanged when
switching between little-endian and big-endian operation. When a data item larger than a byte
is loaded from or stored to memory, the bytes making up that data item are arranged into the
correct order depending on the endianness of the memory access. An ARM byte-invariant
implementation also supports unaligned halfword and word memory accesses. It expects
multi-word accesses to be word-aligned.

Cache

A block of on-chip or off-chip fast access memory locations, situated between the processor and
main memory, used for storing and retrieving copies of often used instructions, data, or
instructions and data. This is done to greatly increase the average speed of memory accesses and
so improve processor performance.

Condition field

A four-bit field in an instruction that specifies a condition under which the instruction can
execute.

Conditional execution

If the condition code flags indicate that the corresponding condition is true when the instruction
starts executing, it executes normally. Otherwise, the instruction does nothing.
Context

The environment that each process operates in for a multitasking operating system. In ARM
processors, this is limited to mean the physical address range that it can access in memory and
the associated memory access permissions.

Coprocessor

A processor that supplements the main processor. The Cortex-M4 processor does not support
any coprocessors.

Debugger

A debugging system that includes a program, used to detect, locate, and correct software faults,
together with custom hardware that supports software debugging.

Direct Memory Access (DMA)

An operation that accesses main memory directly, without the processor performing any
accesses to the data concerned.
Doubleword

A 64-bit data item. The contents are taken as being an unsigned integer unless otherwise stated.

Doubleword-aligned

A data item having a memory address that is divisible by eight.
Endianness

Byte ordering. The scheme that determines the order that successive bytes of a data word are
stored in memory. An aspect of the systems memory mapping.
See also Little-endian and Big-endian

Exception

ARM DUI 0553A
ID121610

An event that interrupts program execution. When an exception occurs, the processor suspends
the normal program flow and starts execution at the address indicated by the corresponding
exception vector. The indicated address contains the first instruction of the handler for the
exception.

Glossary-2

Glossary

An exception can be an interrupt request, a fault, or a software-generated system exception.
Faults include attempting an invalid memory access, attempting to execute an instruction in an
invalid processor state, and attempting to execute an undefined instruction.
Exception service routine

See Interrupt handler.
Exception vector

See Interrupt vector.

Flat address mapping

A system of organizing memory in which each physical address in the memory space is the same
as the corresponding virtual address.
Halfword

A 16-bit data item.

Illegal instruction

An instruction that is architecturally Undefined.

Implementation-defined

The behavior is not architecturally defined, but is defined and documented by individual
implementations.
Implementation-specific

The behavior is not architecturally defined, and does not have to be documented by individual
implementations. Used when there are a number of implementation options available and the
option chosen does not affect software compatibility.
Index register

In some load and store instruction descriptions, the value of this register is used as an offset to
be added to or subtracted from the base register value to form the address that is sent to memory.
Some addressing modes optionally enable the index register value to be shifted prior to the
addition or subtraction.
See also Base register.

Instruction cycle count

The number of cycles that an instruction occupies the Execute stage of the pipeline.
Interrupt handler

A program that control of the processor is passed to when an interrupt occurs.

Interrupt vector

One of a number of fixed addresses in low memory, or in high memory if high vectors are
configured, that contains the first instruction of the corresponding interrupt handler.

Little-endian (LE)

Byte ordering scheme in which bytes of increasing significance in a data word are stored at
increasing addresses in memory.
See also Big-endian (BE), Byte-invariant, Endianness.

Little-endian memory

Memory in which:
•

a byte or halfword at a word-aligned address is the least significant byte or halfword
within the word at that address

•

a byte at a halfword-aligned address is the least significant byte within the halfword at that
address.

See also Big-endian memory.
Load/store architecture

A processor architecture where data-processing operations only operate on register contents, not
directly on memory contents.

ARM DUI 0553A
ID121610

Glossary-3

Glossary

Memory Protection Unit (MPU)

Hardware that controls access permissions to blocks of memory. An MPU does not perform any
address translation.
Prefetching

In pipelined processors, the process of fetching instructions from memory to fill up the pipeline
before the preceding instructions have finished executing. Prefetching an instruction does not
mean that the instruction has to be executed.

Read

Reads are defined as memory operations that have the semantics of a load. Reads include the
Thumb instructions LDM, LDR, LDRSH, LDRH, LDRSB, LDRB, and POP.

Region

A partition of memory space.

Reserved

A field in a control register or instruction format is reserved if the field is to be defined by the
implementation, or produces Unpredictable results if the contents of the field are not zero. These
fields are reserved for use in future extensions of the architecture or are
implementation-specific. All reserved bits not used by the implementation must be written as 0
and read as 0.

Should Be One (SBO)

Write as 1, or all 1s for bit fields, by software. Writing as 0 produces Unpredictable results.
Should Be Zero (SBZ)

Write as 0, or all 0s for bit fields, by software. Writing as 1 produces Unpredictable results.
Should Be Zero or Preserved (SBZP)

Write as 0, or all 0s for bit fields, by software, or preserved by writing the same value back that
has been previously read from the same field on the same processor.
Thread-safe

In a multi-tasking environment, thread-safe functions use safeguard mechanisms when
accessing shared resources, to ensure correct operation without the risk of shared access
conflicts.

Thumb instruction

One or two halfwords that specify an operation for a processor to perform. Thumb instructions
must be halfword-aligned.

Unaligned

A data item stored at an address that is not divisible by the number of bytes that defines the data
size is said to be unaligned. For example, a word stored at an address that is not divisible by four.

Undefined

Indicates an instruction that generates an Undefined instruction exception.

Unpredictable (UNP)

You cannot rely on the behavior. Unpredictable behavior must not represent security holes.
Unpredictable behavior must not halt or hang the processor, or any parts of the system.
Warm reset

Also known as a core reset. Initializes the majority of the processor excluding the debug
controller and debug logic. This type of reset is useful if you are using the debugging features
of a processor.

See Write-allocate (WA).

See Write-back (WB).

Word

A 32-bit data item.

Write

Writes are defined as operations that have the semantics of a store. Writes include the Thumb
instructions STM, STR, STRH, STRB, and PUSH.

Write-allocate (WA)

In a write-allocate cache, a cache miss on storing data causes a cache line to be allocated into
the cache.

ARM DUI 0553A
ID121610

Glossary-4

Glossary

Write-back (WB)

In a write-back cache, data is only written to main memory when it is forced out of the cache on
line replacement following a cache miss. Otherwise, writes by the processor only update the
cache. This is also known as copyback.

Write buffer

A block of high-speed memory, arranged as a FIFO buffer, between the data cache and main
memory, whose purpose is to optimize stores to main memory.

Write-through (WT)

In a write-through cache, data is written to main memory at the same time as the cache is
updated.

ARM DUI 0553A
ID121610

Glossary-5

Source Exif Data:

File Type                       : PDF
File Type Extension             : pdf
MIME Type                       : application/pdf
PDF Version                     : 1.7
Linearized                      : Yes
Page Mode                       : UseOutlines
XMP Toolkit                     : Adobe XMP Core 4.0-c321 44.398116, Tue Aug 04 2009 14:24:39
Creator Tool                    : FrameMaker 8.0
Modify Date                     : 2010:12:16 19:04:18Z
Create Date                     : 2010:12:16 19:04:18Z
Copyright                       : Copyright © 2010 ARM. All rights reserved.
Producer                        : Acrobat Distiller 8.2.0 (Windows)
Keywords                        : Cortex-M, Cortex-M4
Format                          : application/pdf
Title                           : Cortex-M4 Devices Generic User Guide
Creator                         : ARM Limited
Description                     : A generic User Guide for devices that implement the ARM Cortex-M4 processor, intended for end-users. The document includes 4GB unified memory map, the exception model and Nested Vectored Interrupt Controller (NVIC), Optional bit-banding, and the implemented Thumb and floating-point instruction set descriptions.
Document ID                     : uuid:338a98b7-89a4-4fdb-996a-35113a3fa29e
Instance ID                     : uuid:8d5e6f14-b9a6-4c4a-bb41-377634d08b4f
Page Count                      : 276
Subject                         : A generic User Guide for devices that implement the ARM Cortex-M4 processor, intended for end-users. The document includes 4GB unified memory map, the exception model and Nested Vectored Interrupt Controller (NVIC), Optional bit-banding, and the implemented Thumb and floating-point instruction set descriptions.
Author                          : ARM Limited

EXIF Metadata provided by EXIF.tools

Cortex M4 Devices Generic User Guide

Cortex-M4%20Devices%20Generic%20User%20Guide

Cortex_m4%20Generic%20User%20Guide

DUI0553A_cortex_m4_Generic_User_Guide

cortex_m4_User_Guide

CortexM4_Generic_Users_Guide_DUI0553A_cortex_m4_dgug

Navigation menu

Versions of this User Manual:

Views

Navigation