NXP CodeWarrior Development Studio Reference manual
Reklama
Reklama
CodeWarrior Development Studio for
Advanced Packet Processing Build
Tools Reference Manual
Document Number: CWAPPBTR
Rev. 10.2, 01/2016
2
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Section number
Contents
Title Page
Chapter 1
Introduction
Chapter 2
Using Build Tools on the Command Line
Chapter 3
Command-Line Options for Standard C Conformance
Chapter 4
Command-Line Options for Standard C++ Conformance
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
3
Section number Title Page
Chapter 5
Command-Line Options for Language Translation
Chapter 6
Command-Line Options for Diagnostic Messages
4
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Section number Title Page
Chapter 7
Command-Line Options for Preprocessing
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
5
Section number Title Page
Chapter 8
Command-Line Options for Library and Linking
Chapter 9
Command-Line Options for Object Code
Chapter 10
Command-Line Options for Optimization
Chapter 11
Command-Line
6
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Section number Title Page
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
7
Section number Title Page
8
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Section number Title Page
Chapter 12
Assembler
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
9
Section number Title Page
10
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Section number Title Page
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
11
Section number Title Page
12
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Section number Title Page
Chapter 13
Linker for AIOP
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
13
Section number Title Page
14
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Section number Title Page
Chapter 14
Linker
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
15
Section number Title Page
16
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Section number Title Page
Chapter 15
C Compiler
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
17
Section number Title Page
Chapter 16
C++ Compiler
Chapter 17
18
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Section number Title Page
Precompiling
Chapter 18
Intermediate Optimizations
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
19
Section number Title Page
Chapter 19
AIOP Optimizations
Chapter 20
Inline-Assembly for AIOP Build Tools
20
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Section number Title Page
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
21
Section number Title Page
22
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Section number Title Page
Chapter 21
AIOP Code Generation
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
23
Section number Title Page
24
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Section number Title Page
Chapter 22
Libraries and Support Code
Chapter 23
Declaration Specifications
Chapter 24
Declaration Specifications for AIOP Code
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
25
Section number Title Page
Chapter 25
Predefined Macros
26
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Section number Title Page
Chapter 26
Predefined Macros for AIOP Compiler
Chapter 27
Using Pragmas
Chapter 28
Pragmas for Standard C Conformance
Chapter 29
Pragmas for C++
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
27
Section number Title Page
28
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Section number Title Page
Chapter 30
Pragmas for Language Translation
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
29
Section number Title Page
Chapter 31
Pragmas for Diagnostic Messages
30
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Section number Title Page
Chapter 32
Pragmas for Preprocessing
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
31
Section number Title Page
Chapter 33
Pragmas for Library and Linking
Chapter 34
Pragmas for Code Generation
32
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Section number Title Page
Chapter 35
Pragmas for Optimization
Chapter 36
Pragmas for AIOP Compiler
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
33
Section number Title Page
34
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Section number Title Page
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
35
36
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 1
Introduction
This reference explains how to use CodeWarrior tools to build programs. CodeWarrior build tools translate source code into object code then organize that object code to create a program that is ready to execute. CodeWarrior build tools run on the host system to generate software that runs on the target system. Sometimes the host and target are the same system. Usually, these systems are different.
This reference covers the CodeWarrior compiler and its linker, versions 4.0 and higher.
This chapter explains the processes that CodeWarrior build tools use to create software:
•
•
1.1 Compiler Architecture
From a programmer's point of view, the CodeWarrior compiler translates source code into object code. Internally, however, the CodeWarrior compiler organizes its work between its front-end and back-end, each end taking several steps
steps the compiler takes.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
37
Linker Architecture
Figure 1-1. CodeWarrior Compiler Steps
Front-end steps:
• read settings: retrieves your settings from the host's integrated development environment (IDE) or the command line to configure how to perform subsequent steps
• read and preprocess source code: reads your program's source code files and applies preprocessor directives
• translate to intermediate representation: translates your program's preprocessed source code into a platform-independent intermediate representation
• optimize intermediate representation: rearranges the intermediate representation to reduce your program's size, improve its performance, or both
Back-end steps:
• translate to processor object code: converts the optimized intermediate representation into native object code, containing data and instructions, for the target processor
• optimize object code: rearranges the native object code to reduce its size, improve performance, or both
• output object code and diagnostic data: writes output files on the host system, ready for the linker and diagnostic tools such as a debugger or profiler
38
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 1 Introduction
1.2 Linker Architecture
The linker combines and arranges data and instructions from one or more object code files into a single file, or image. This image is ready to execute on the target platform.
The CodeWarrior linker uses settings from the host's integrated development environment (IDE) or command line to determine how to generate the image file.
The linker also optionally reads a linker command file. A linker command file allows you to specify precise details of how data and instructions should be arranged in the image file.
shows the steps the CodeWarrior linker takes to build an executable image.
Figure 1-2. CodeWarrior Linker Steps
• read settings: retrieves your settings from the IDE or the command line to determine how to perform subsequent steps
• read linker command file: retrieves commands to determine how to arrange object code in the final image
• read object code: retrieves data and executable objects that are the result of compilation or assembly
• delete unused objects ("deadstripping"): deletes objects that are not referred to by the rest of the program
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
39
Linker Architecture
• resolve references among objects: arranges objects to compose the image then computes the addresses of the objects
• output link map and image files: writes files on the host system, ready to load onto the target system
40
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 2
Using Build Tools on the Command Line
CodeWarrior build tools may be invoked from the command-line. These command-line tools operate almost identically to their counterparts in an integrated development environment (IDE). CodeWarrior command-line compilers and assemblers translate source code files into object code files. CodeWarrior command-line linkers then combine one or more object code files to produce an executable image file, ready to load and execute on the target platform. Each command-line tool has options that you configure when you invoke the tool.
•
Configuring Command-Line Tools
•
•
•
2.1 Configuring Command-Line Tools
This chapter explains:
•
Setting CodeWarrior Environment Variables
•
Setting the PATH Environment Variable
2.1.1 Setting CodeWarrior Environment Variables
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
41
Configuring Command-Line Tools
Use environment variables on the host system to specify to the CodeWarrior command line tools where to find CodeWarrior files for compiling and linking. The table below describes these environment variables.
Table 2-1. Environment variables for CodeWarrior command-line tools
The environment variable...
CWFolder
MWCEABIPPCIncludes
MWEABIPPCLibraries
MWEABIPPCLibraryFiles
specifies this information
CodeWarrior installation path on the host system.
Directories on the host system for system header files for the
CodeWarrior compiler.
Directories on the host system for system libraries for the
CodeWarrior linker.
List of library names to be linked.
A system header file is a header file that is enclosed with the "<" and ">" characters in include directives. For example
#include <stdlib.h> /* stdlib.h system header. */
Typically, you define the
MWCEABIPPCIncludes
and
MWEABIPPCLibraries
environment variables to refer to the header file and library search paths in the subdirectories of your
CodeWarrior software.
To specify more than one directory for the
MWCEABIPPCIncludes
and
MWEABIPPCLibraries variables, use the conventional separator for your host operating system command-line shell.
CWFolder
NOTE
environment variable specifies the CodeWarrior installation path on a host system. Verify the CodeWarrior installation path on the host system before setting the
CWFolder environment variable.
Listing 2-1. Setting environment variables in Microsoft® Windows® operating systems
rem Use ; to separate directory paths set CWFolder=Replace this text with the CodeWarrior installation path on the host system.
set MWCEABIPPCIncludes=%CWFolder%\AIOP_Support\ewl\EWL_C\include;%CWFolder%\AIOP_Support\ewl
\EWL_C++\include;%CWFolder%\PA_Support\ewl\EWL_Runtime\Runtime_AIOP\Include set MWEABIPPCLibraries=%CWFolder%\AIOP_Support\ewl\lib
2.1.2 Setting the PATH Environment Variable
42
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 2 Using Build Tools on the Command Line
The
PATH
variable should include the paths for your CodeWarrior tools, shown in
. Toolset represents the name of the folder that contains the command line tools for your build target.
Listing 2-2. Example of setting PATH
Set CWFolder=Replace this text with the CodeWarrior installation path on the host system.
Set PATH=%PATH%;%CWFolder%\CW_APP\LS\bin;%CWFolder%\CW_APP\LS\AIOP_Tools\Command_Line_Tools
2.2 Invoking Command-Line Tools
To compile, assemble, link, or perform some other programming task with the
CodeWarrior command-line tools, you type a command at a command line's prompt. This command specifies the tool you want to run, what options to use while the tool runs, and what files the tool should operate on. For example,
tool options files
Where tool is the name of the CodeWarrior command-line tool to invoke, options is a list of zero or more options that specify to the tool what operation it should perform and how it should be performed, and files is a list of files zero or more files that the tool should operate on.
The tool performs the operation on the files you specify. If the tool is successful it simply finishes its operation and a new prompt appears at the command line. If the tool encounters problems it reports these problems as text messages on the command-line before a new prompt appears.
Scripts that automate the process to build a piece of software contain commands to invoke command-line tools. For example, make
, a common software development tool, uses scripts to manage dependencies among source code files and invoke command-line compilers, assemblers and linkers as needed.
Response files encapsulate complex lists of options and filenames into a convenient file.
tool [@response_file] OR tool [@@response_file]
Where response_file is used to insert command-line arguments from a file.
The response file is parsed such that arguments are separated by whitespace except where surrounded by quote marks. Anything followed by a pound character
'#'
is used to indicate the rest of the line is a comment. Use
'\#'
in a response file if an argument actually starts with
'#'
.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
43
Getting Help
The
@
response file prefix will make sure the response file content is used "as is", whereas a '
@@
' prefix will ensure that environment variables in the response file contents are expanded. Note that only one level of environment variable expansion is performed.
Environment variables that expand to another variable-like syntax remain unchanged. For example, in
FOO=$BAR
,
$FOO
expands to
$BAR
, whereas
$BAR
is not expanded.
Any of the following syntax can be used to recognize environment variables:
•
$[A-Za-z_][A-Za-z0-9_]*
•
$([A-Za-z_][A-Za-z0-9_]*)
•
${[A-Za-z_][A-Za-z0-9_]*}
Points to remember:
• The end of the environment variable is terminated when any char that is not a recognized environment variable token, is encountered.
• The
$
is escaped by a preceding
$
character in an '
@@
' prefixed response file.
• Lines in the response file should not exceed 64 KB in length. It is recommended to put each option specified on a separate line, if many options are to be specified.
• A response file cannot have another response file, that is,
@
and
@@
cannot be used inside a response file.
• Example usage: tool @argu.txt //Here all arguments and input files are passed in the file argu.txt
tool @argu.txt main.c // Here only arguments are passed in the file argu.txt
tool @@argu.txt
2.3 Getting Help
To show short descriptions of a tool's options, type this command at the command line:
tool -help
Where tool is the name of the CodeWarrior build tool.
To show only a few lines of help information at a time, pipe the tool's output to a pager program. For example,
tool -help | more
Will use the more
pager program to display the help information.
44
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 2 Using Build Tools on the Command Line
Enter the following command in a Command Prompt window to see a list of specifications that describe how options are formatted:
tool -help usage
Where tool is the name of the CodeWarrior build tool.
2.3.1 Parameter Formats
Parameters in an option are formatted as follows:
• A parameter included in brackets "
[]
" is optional.
• Use of the ellipsis "
...
" character indicates that the previous type of parameter may be repeated as a list.
2.3.2 Option Formats
Options are formatted as follows:
• For most options, the option and the parameters are separated by a space as in "
-xxx param
".
When the option's name is "
-xxx+
", however, the parameter must directly follow the option, without the "
+
" character (as in "
-xxx45
") and with no space separator.
• An option given as "
-[no]xxx
" may be issued as "
-xxx
" or "
-noxxx
".
The use of "
-noxxx
" reverses the meaning of the option.
• When an option is specified as "
-xxx | yy[y] | zzz
", then either "
-xxx
", "
-yy
", "
-yyy
", or "
-zzz
" matches the option.
• The symbols "
,
" and "
=
" separate options and parameters unconditionally; to include one of these symbols in a parameter or filename, escape it (for example, "
\,
" in ccaiop file.c\,v)
.
2.3.3 Common Terms
These common terms appear in many option descriptions:
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
45
File Name Extensions
• A "cased" option is considered case-sensitive. By default, no options are casesensitive.
• "compatibility" indicates that the option is borrowed from another vendor's tool and its behavior may only approximate its counterpart.
• A "global" option has an effect over the entire command line and is parsed before any other options. When several global options are specified, they are interpreted in order.
• A "deprecated" option will be eliminated in the future and should no longer be used.
An alternative form is supplied.
• An "ignored" option is accepted by the tool but has no effect.
• A "meaningless" option is accepted by the tool but probably has no meaning for the target operating system.
• An "obsolete" option indicates a deprecated option that is no longer available.
• A "substituted" option has the same effect as another option. This points out a preferred form and prevents confusion when similar options appear in the help.
• Use of "default" in the help text indicates that the given value or variation of an option is used unless otherwise overridden.
This tool calls the linker (unless a compiler option such as
-c
prevents it) and understands linker options
-
use "
-help tool=other
" to see them. Options marked "passed to linker" are used by the compiler and the linker; options marked "for linker" are used only by the linker. When using the compiler and linker separately, you must pass the common options to both.
2.4 File Name Extensions
Files specified on the command line are identified by contents and file extension. The command-line version of the CodeWarrior C/C++ compiler accepts non-standard file extensions as source code but also emits a warning message. By default, the compiler assumes that a file with any extensions besides
.c
,
.h
,
.pch
is C++ source code. The linker ignores all the files that it can not identify as object code, libraries, or command files.
Linker command files must end in
.lcf
. They may be simply added to the link line, for
example ( Listing 2-3 on page 46 ).
Listing 2-3. Example of using linker command files
ldaiop file.o lib.a commandfile.lcf
46
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 3
Command-Line Options for Standard C
Conformance
This chapter explains the following command-line options for standard C conformance:
•
•
•
3.1 -ansi
Controls the ISO/IEC 9899-1990 ("C90") conformance options, overriding the given settings.
Syntax
-ansi keyword
The arguments for keyword
are: off
Turns ISO conformance off. Same as
-stdkeywords off -enum min -strict off. on | relaxed
Turns ISO conformance on in relaxed mode. Same as
-stdkeywords on -enum min -strict on strict
Turns ISO conformance on in strict mode. Same as
-stdkeywords on -enum int -strict on
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
47
-stdkeywords
3.2 -stdkeywords
Controls the use of ISO/IEC 9899-1990 ("C90") keywords.
Syntax
-stdkeywords on | off
Remarks
Default setting is off
.
3.3 -strict
Controls the use of non-standard ISO/IEC 9899-1990 ("C90") language features.
Syntax
-strict on | off
Remarks
If this option is on
, the compiler generates an error message if it encounters some
CodeWarrior extensions to the C language defined by the ISO/IEC 9899-1990 ("C90") standard:
• C++-style comments
• unnamed arguments in function definitions
• non-standard keywords
The default setting is off
.
48
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 4
Command-Line Options for Standard C++
Conformance
This chapter explains the following command-line options for standard C++ conformance:
•
•
•
•
•
•
•
•
4.1 -bool
Controls the use of true
and false
keywords for the C++ bool
data type.
Syntax
-bool on | off
Remarks
When on
, the compiler recognizes the true
and false
keywords in expressions of type bool
.
When off, the compiler does recognizes the keywords, forcing the source code to provide definitions for these names. The default is on
.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
49
-dialect
4.2 -cpp_exceptions
Controls the use of C++ exceptions.
Syntax
-cpp_exceptions on | off
Remarks
When on
, the compiler recognizes the try
, catch
, and throw
keywords and generates extra executable code and data to handle exception throwing and catching. The default is on
.
4.3 -dialect
Specifies the source language.
Syntax
-dialect keyword
-lang keyword
The arguments for keyword are: c
Expect source code to conform to the language specified by the ISO/IEC 9899-1990
("C90") standard.
c99
Expect source code to conform to the language specified by the ISO/IEC 9899-1999
("C99") standard.
c++ | cplus
Always treat source as the C++ language.
ec++
Generate error messages for use of C++ features outside the Embedded C++ subset.
Implies
-dialect cplus
.
objc
Always treat source as the Objective-C language.
50
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 4 Command-Line Options for Standard C++ Conformance
4.4 -for_scoping
Controls legacy scope behavior in for loops.
Syntax
-for_scoping
Remarks
When enabled, variables declared in for
loops are visible to the enclosing scope; when disabled, such variables are scoped to the loop only. The default is off
.
4.5 -instmgr
Controls whether the instance manager for templates is active.
Syntax
-inst[mgr] keyword [,...]
The options for keyword are: off
Turn off the C++ instance manager. This is the default.
on
Turn on the C++ instance manager.
file= path
Specify the path to the database used for the C++ instance manager. Unless specified the default database is
cwinst.db
.
Remarks
This command is global. The default setting is off
.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
51
-iso_templates
4.6 -iso_templates
Controls whether the ISO/IEC 14882:2003 standard C++ template parser is active.
Syntax
-iso_templates on | off
Remarks
Default setting is on
.
4.7 -RTTI
Controls the availability of runtime type information (RTTI).
Syntax
-RTTI on | off
Remarks
Default setting is on
.
4.8 -wchar_t
Controls the use of the wchar_t
data type in C++ source code.
Syntax
-wchar_t on | off
Remarks
The
-wchar on
option tells the C++ compiler to recognize the wchar_t
type as a built-in type for wide characters. The
-wchar off
option tells the compiler not to allow this built-in type, forcing the user to provide a definition for this type. Default setting is on
.
52
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 5
Command-Line Options for Language Translation
This chapter explains the following command-line options for language translation:
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
5.1 -char
Controls the default sign of the char
data type.
Syntax
-char keyword
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
53
-defaults
The arguments for keyword are: signed char
data items are signed.
unsigned char
data items are unsigned.
Remarks
The default is signed
.
5.2 -defaults
Controls whether the compiler uses environment variables to search for include and library paths.
Syntax
-defaults
-nodefaults
Remarks
defaults
is the default setting, that is environment variables are used to find include and library paths.
–nodefaults
disables the environment variable search and all paths must be explicitly provided.
5.3 -encoding
Specifies the default source encoding used by the compiler.
Syntax
-enc[oding] keyword
The options for keyword are: ascii
American Standard Code for Information Interchange (ASCII) format. This is the default.
54
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 5 Command-Line Options for Language Translation
autodetect | multibyte | mb
Scan file for multibyet encoding.
system
Uses local system format.
UTF[8 | -8]
Unicode Transformation Format (UTF).
SJIS | Shift-JIS | ShiftJIS
Shift Japanese Industrial Standard (Shift-JIS) format.f
EUC[JP | -JP]
Japanese Extended UNIX Code (EUCJP) format.
ISO[2022JP | -2022-JP]
International Organization of Standards (ISO) Japanese format.
Remarks
The compiler automatically detects UTF-8 (Unicode Transformation Format) header or
UCS-2/UCS-4 (Uniform Communications Standard) encodings regardless of setting. The default setting is ascii
.
5.4 -flag
Specifies compiler
#pragma
as either on
or off
.
Syntax
-fl[ag] [no-]pragma
Remarks
For example, this option setting
-flag require_prototypes is equivalent to
#pragma require_prototypes on
This option setting
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
55
-gccext
-flag no-require_prototypes is the same as
#pragma require_prototypes off
5.5 -gccext
Enables GCC (Gnu Compiler Collection) C language extensions.
Syntax
-gcc[ext] on | off
Remarks
See for a list of language extensions that the compiler recognizes when this option is on
.
The default setting is off
.
5.6 -gcc_extensions
Equivalent to the
-gccext
option.
Syntax
-gcc[_extensions] on | off
5.7 -M
Scans source files for dependencies and emit a Makefile, without generating object code.
Syntax
-M
Remarks
This command is global and case-sensitive.
56
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 5 Command-Line Options for Language Translation
5.8 -make
Scans source files for dependencies and emit a Makefile, without generating object code.
Syntax
-make
Remarks
This command is global.
5.9 -mapcr
Swaps the values of the
\n
and
\r
escape characters.
Syntax
-mapcr
-nomapcr
Remarks
The
-mapcr
option tells the compiler to treat the
'\n'
character as ASCII 13 and the
'\r' character as ASCII 10. The
-nomapcr
option tells the compiler to treat these characters as
ASCII 10 and 13, respectively.
5.10 -MM
Scans source files for dependencies and emit a Makefile, without generating object code or listing system
#include
files.
Syntax
-MM
Remarks
This command is global and case-sensitive.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
57
-MD
5.11 -MD
Scans source files for dependencies and emit a Makefile, generate object code, and write a dependency map.
Syntax
-MD
Remarks
This command is global and case-sensitive.
5.12 -MMD
Scans source files for dependencies and emit a Makefile, generate object code, write a dependency map, without listing system
#include
files.
Syntax
-MMD
Remarks
This command is global and case-sensitive.
5.13 -msext
Allows Microsoft® Visual C++ extensions.
Syntax
-msext on | off
Remarks
Turn on this option to allow Microsoft Visual C++ extensions:
• Redefinition of macros
• Allows
XXX::yyy
syntax when declaring method yyy
of class
XXX
• Allows extra commas
58
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 5 Command-Line Options for Language Translation
• Ignores casts to the same type
• Treats function types with equivalent parameter lists but different return types as equal
• Allows pointer-to-integer conversions, and various syntactical differences
5.14 -once
Prevents header files from being processed more than once.
Syntax
-once
Remarks
You can also add #pragma once on in a prefix file.
5.15 -pragma
Defines a pragma for the compiler.
Syntax
-pragma "name [
setting]"
The arguments are: name
Same of the pragma.
setting
Arguments to give to the pragma
Remarks
For example, this command-line option
-pragma "c99 on"
Is equivalent to inserting this directive in source code
#pragma c99 on
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
59
-relax_pointers
5.16 -relax_pointers
Relaxes the pointer type-checking rules in C.
Syntax
-relax_pointers
Remarks
This option is equivalent to
#pragma mpwc_relax on
5.17 -requireprotos
Controls whether or not the compiler should expect function prototypes.
Syntax
-r[equireprotos]
5.18 -search
Globally searches across paths for source files, object code, and libraries specified in the command line.
Syntax
-search
5.19 -trigraphs
Controls the use of trigraph sequences specified by the ISO/IEC standards for C and C++.
Syntax
-trigraphs on | off
60
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Remarks
Default setting is off
.
Chapter 5 Command-Line Options for Language Translation
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
61
-trigraphs
62
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 6
Command-Line Options for Diagnostic Messages
This chapter explains the following command-line options for diagnostic messages:
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
6.1 -disassemble
Instructs the command-line tool to disassemble files and send result to stdout
.
Syntax
-dis[assemble]
Remarks
This option is global.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
63
-warningerror
6.2 -warningerror
Controls whether or not warnings are treated as errors.
Syntax
-warningerror [on | off]
Remarks
If you enable this option, the compiler treats all warning messages as though they were errors and does not translate your file until you resolve them.
6.3 -help
Lists descriptions of the CodeWarrior tool's command-line options.
Syntax
-help [keyword [,...]]
The options for keyword are: all
Show all standard options group= keyword
Show help for groups whose names contain keyword (case-sensitive).
[no]compatible
Use compatible
to show options compatible with this compiler. Use nocompatible
to show options that do not work with this compiler.
[no]deprecated
Shows deprecated options
[no]ignored
Shows ignored options
[no]meaningless
64
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 6 Command-Line Options for Diagnostic Messages
Shows options meaningless for this target
[no]normal
Shows only standard options
[no]obsolete
Shows obsolete options
[no]spaces
Inserts blank lines between options in printout.
opt[ion]= name
Shows help for a given option; for name, maximum length 63 chars search= keyword
Shows help for an option whose name or help contains keyword (case-sensitive), maximum length 63 chars tool=keyword[ all | this | other | skipped | both ]
Categorizes groups of options by tool; default.
• all
-show all options available in this tool
• this
-show options executed by this tool; default
• other | skipped
-show options passed to another tool
• both
-show options used in all tools usage
Displays usage information.
6.4 -maxerrors
Specifies the maximum number of errors messages to show.
Syntax
-maxerrors max max
Use max
to specify the number of error messages. Common values are:
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
65
-maxwarnings
•
0
(zero) - disable maximum count, show all error messages (default).
• n
- Maximum number of errors to show, such as
-maxwarnings
.
6.5 -maxwarnings
Specifies the maximum number of warning messages to show.
Syntax
-maxwarnings max
max
Specifies the number of warning messages. Common values are:
•
0
(zero) - Disable maximum count (default).
• n - Maximum number of warnings to show.
6.6 -msgstyle
Controls the style used to show error and warning messages.
Syntax
-msgstyle keyword
The options for keyword are: gcc
Uses the message style that the Gnu Compiler Collection tools use.
parseable
Uses context-free machine parseable message style.
std
Uses standard message style. This is the default.
6.7 -nofail
66
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 6 Command-Line Options for Diagnostic Messages
Continues processing after getting error messages in earlier files.
Syntax
-nofail
6.8 -progress
Shows progress and version information.
Syntax
-progress
6.9 -S
Disassembles all files and send output to a file. This command is global and casesensitive.
Syntax
-S
6.10 -stderr
Uses the standard error stream to report error and warning messages.
Syntax
-stderr
-nostderr
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
67
-verbose
Remarks
The
-stderr
option specifies to the compiler, and other tools that it invokes, that error and warning messages should be sent to the standard error stream.
The
-nostderr
option specifies that error and warning messages should be sent to the standard output stream.
6.11 -verbose
Instructs the compiler to provide extra, cumulative information in messages.
Syntax
-v[erbose]
Remarks
This option also gives progress and version information.
6.12 -version
Displays version, configuration, and build data.
Syntax
-v[ersion]
6.13 -timing
Shows the amount of time that the tool used to perform an action.
Syntax
68
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 6 Command-Line Options for Diagnostic Messages
-timing
6.14 -warnings
Specifies which warning messages the command-line tool issues. This command is global.
Syntax
-w[arnings] keyword [,...]
The options for keyword
are: off
Turns off all warning messages. Passed to all tools. Equivalent to
#pragma warning off on
Turns on most warning messages. Passed to all tools. Refer
for a list of warning messages turned on by the
-w[arnings] on
command.
Equivalent to
#pragma warning on most
Turns on most warnings.
all
Turns on almost all warnings and require prototypes.
full
Turns on all warning messages and require prototypes. This option is likely to generate spurious warnings.
-warnings full
NOTE
should be used before using any other options that affect warnings. For example, use
-warnings full -warnings noanyptrintconv
instead of
-warnings noanyptrintconv -warnings full.
[no]cmdline
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
69
-warnings
Passed to all tools.
[no]err[or] | [no]iserr[or]
Treats warnings as errors. Passed to all tools. Equivalent to
#pragma warning_errors
[no]pragmas | [no]illpragmas
Issues warning messages on invalid pragmas. Enabled when most
is used. Equivalent to
#pragma warn_illpragma
[no]empty[decl]
Issues warning messages on empty declarations. Enabled when most
is used. Equivalent to
#pragma warn_emptydecl
[no]possible | [no]unwanted
Issues warning messages on possible unwanted effects. Enabled when most
is used.
Equivalent to
#pragma warn_possunwanted
[no]unusedarg
Issues warning messages on unused arguments. Enabled when most
is used. Equivalent to
#pragma warn_unusedarg
[no]unusedvar
Issues warning messages on unused variables. Enabled when most
is used. Equivalent to
#pragma warn_unusedvar
[no]unused
Same as
70
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 6 Command-Line Options for Diagnostic Messages
-w [no]unusedarg,[no]unusedvar
Enabled when most
is used.
[no]extracomma | [no]comma
Issues warning messages on extra commas in enumerations. The compiler ignores terminating commas in enumerations when compiling source code that conforms to the
ISO/IEC 9899-1999 ("C99") standard. Enabled when most
is used. Equivalent to
#pragma warn_extracomma
[no]extended
Extended error checking. Enabled when most
is used. Equivalent to either:
#pragma extended_errorcheck
[no]hidevirtual | [no]hidden[virtual]
Issues warning messages on hidden virtual functions. Enabled when most
is used.
Equivalent to
#pragma warn_hidevirtual
[no]implicit[conv]
Issues warning messages on implicit arithmetic conversions. Enabled when all
is used.
Implies
-warn impl_float2int,impl_signedunsigned
[no]impl_int2float
Issues warning messages on implicit integral to floating conversions. Enabled when all
is used. Equivalent to
#pragma warn_impl_i2f_conv
[no]impl_float2int
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
71
-warnings
Issues warning messages on implicit floating to integral conversions. Enabled when all
is used. Equivalent to
#pragma warn_impl_f2i_conv
[no]impl_signedunsigned
Issues warning messages on implicit signed/unsigned conversions. Enabled when all
is used.
[no]notinlined
Issues warning messages for functions declared with the inline
qualifier that are not inlined. Enabled when full
is used. Equivalent to
#pragma warn_notinlined
[no]largeargs
Issues warning messages when passing large arguments to unprototyped functions.
Enabled when most
is used. Equivalent to
#pragma warn_largeargs
[no]structclass
Issues warning messages on inconsistent use of class
and struct
. Enabled when most
is used. Equivalent to
#pragma warn_structclass
[no]padding
Issue warning messages when padding is added between struct
members. Enabled when full
is used. Equivalent to
#pragma warn_padding
[no]notused
Issues warning messages when the result of non-void-returning functions are not used.
Enabled when full
is used. Equivalent to
72
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 6 Command-Line Options for Diagnostic Messages
#pragma warn_resultnotused
[no]missingreturn
Issues warning messages when a return without a value in non-void-returning function occurs. Enabled when most
is used. Equivalent to
#pragma warn_missingreturn
[no]unusedexpr
Issues warning messages when encountering the use of expressions as statements without side effects. Equivalent to
#pragma warn_no_side_effect
[no]ptrintconv
Issues warning messages when lossy conversions occur from pointers to integers.
Enabled when full
is used.
[no]anyptrintconv
Issues warning messages on any conversion of pointers to integers. Enabled when full
is used. Equivalent to
#pragma warn_ptr_int_conv
[no]undef[macro]
Issues warning messages on the use of undefined macros in
#if
and
#elif
conditionals.
Enabled when full
is used. Equivalent to
#pragma warn_undefmacro
[no]filecaps
Issues warning messages when
# include ""
directives use incorrect capitalization.
Enabled when most
is used. Equivalent to
#pragma warn_filenamecaps
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
73
-warnings
[no]sysfilecaps
Issue warning messages when
# include <>
statements use incorrect capitalization.
Enabled when most
is used. Equivalent to
#pragma warn_filenamecaps_system
[no]tokenpasting
Issue warning messages when token is not formed by the
##
preprocessor operator.
Enabled when most
is used. Equivalent to
#pragma warn_illtokenpasting
[no]relax_i2i_conv
Relax implicit arithmetic conversion warnings on certain implicit conversions. Equivalent to
#pragma relax_i2i_conv noSymRedef
Suppress symbol redefined warnings. Passed to linker (By default linker emits Symbol
Redefined warnings).
display | dump
Display list of active warnings.
Remarks
The table below lists the equivalent command option of the warning messages turned on by the
-w[arnings] on
command.
Table 6-1. Warnings turned on by the
-w[arnings] on
command on/most
all (includes most)
full (includes all and most)
Table continues on the next page...
74
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 6 Command-Line Options for Diagnostic Messages
Table 6-1. Warnings turned on by the
-w[arnings] on
command (continued) on/most
[no]hidevirtual | [no]hidden[virtual]
all (includes most) full (includes all and most)
6.15 -wraplines
Controls the word wrapping of messages.
Syntax
-wraplines
-nowraplines
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
75
-wraplines
76
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 7
Command-Line Options for Preprocessing
This chapter explains the following command-line options for preprocessing:
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
7.1 -convertpaths
Instructs the compiler to interpret
# include file paths specified for a foreign operating system. This command is global.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
77
-cwd
Syntax
-[no]convertpaths
Remarks
The CodeWarrior compiler can interpret file paths from several different operating systems. Each operating system uses unique characters as path separators. These separators include:
• Mac OS® - colon "
:
" (
:sys:stat.h
)
• UNIX - forward slash "
/
" ( sys/stat.h
)
• Windows® operating systems - backward slash "
\
" ( sys\stat.h
)
When convertpaths
is enabled, the compiler can correctly interpret and use paths like
<sys/ stat.h> or
<:sys:stat.h>
. However, when enabled, (
/
) and (
:
) separate directories and cannot be used in filenames.
NOTE
This is not a problem on Windows systems since these characters are already disallowed in file names. It is safe to leave this option on
.
When noconvertpaths
is enabled, the compiler can only interpret paths that use the
Windows form, like
<\sys\stat.h>
.
7.2 -cwd
Controls where a search begins for
# include
files.
Syntax
-cwd keyword
The options for keyword are: explicit
No implicit directory. Search
-I
or
-ir
paths.
include
78
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Begins searching in directory of referencing file.
Chapter 7 Command-Line Options for Preprocessing
proj
Begins searching in current working directory (default).
source
Begins searching in directory that contains the source file.
Remarks
The path represented by keyword is searched before searching access paths defined for the build target.
7.3 -D+
Same as the
-define
option.
Syntax
-D+
name
The parameters are: name
The symbol name to define. Symbol is set to 1.
7.4 -define
Defines a preprocessor symbol.
Syntax
-d[efine] name[=value]
The parameters are: name
The symbol name to define.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
79
-E
value
The value to assign to symbol name. If no value is specified, set symbol value equal to 1.
7.5 -E
Tells the command-line tool to preprocess source files.
Syntax
-E
Remarks
This option is global and case sensitive.
7.6 -EP
Tells the command-line tool to preprocess source files that are stripped of
#line directives.
Syntax
-EP
Remarks
This option is global and case sensitive.
7.7 -gccincludes
Controls the compilers use of GCC
#include
semantics.
Syntax
80
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 7 Command-Line Options for Preprocessing
-gccinc[ludes]
Remarks
Use
-gccincludes
to control the CodeWarrior compiler understanding of Gnu Compiler
Collection (GCC) semantics. When enabled, the semantics include:
• Adds
-I-
paths to the systems list if
-I-
is not already specified
• Search referencing file's directory first for
#include
files (same as
-cwd include
) The compiler and IDE only search access paths, and do not take the currently
#include
file into account.
This command is global.
7.8 -I-
Changes the build target's search order of access paths to start with the system paths list.
Syntax
-I-
-i-
Remarks
The compiler can search
#include
files in several different ways. Use
-I-
to set the search order as follows:
• For include statements of the form
#include "xyz"
, the compiler first searches user paths, then the system paths
• For include statements of the form
#include <xyz>
, the compiler searches only system paths
This command is global.
7.9 -I+
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
81
-include
Appends a non-recursive access path to the current
#include
list.
Syntax
-I+path
-i path
The parameters are: path
The non-recursive access path to append.
Remarks
This command is global and case-sensitive.
7.10 -include
Defines the name of the text file or precompiled header file to add to every source file processed.
Syntax
-include file file
Name of text file or precompiled header file to prefix to all source files.
Remarks
With the command line tool, you can add multiple prefix files all of which are included in a meta-prefix file.
7.11 -ir
Appends a recursive access path to the current
#include
list. This command is global.
Syntax
82
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 7 Command-Line Options for Preprocessing
-ir path path
The recursive access path to append.
7.12 -P
Preprocesses the source files without generating object code, and send output to file.
Syntax
-P
Remarks
This option is global and case-sensitive.
7.13 -precompile
Precompiles a header file from selected source files.
Syntax
-precompile file | dir | "" file
If specified, the precompiled header name.
dir
If specified, the directory to store the header file.
""
If
""
is specified, write header file to location specified in source code. If neither argument is specified, the header file name is derived from the source file name.
Remarks
The driver determines whether to precompile a file based on its extension. The option
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
83
-preprocess
-precompile
filesource is equivalent to
-c -o
filesource
7.14 -preprocess
Preprocesses the source files. This command is global.
Syntax
-preprocess
7.15 -ppopt
Specifies options affecting the preprocessed output.
Syntax
-ppopt keyword [,...]
The arguments for keyword are:
[no]break
Emits file and line breaks. This is the default.
[no]line
Controls whether #line directives are emitted or just comments. The default is line
.
[no]full[path]
Controls whether full paths are emitted or just the base filename. The default is fullpath
.
[no]pragma
Controls whether #pragma directives are kept or stripped. The default is pragma
.
[no]comment
Controls whether comments are kept or stripped.
84
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 7 Command-Line Options for Preprocessing
[no]space
Controls whether whitespace is kept or stripped. The default is space
.
Remarks
The default settings is break
.
7.16 -prefix
Adds contents of a text file or precompiled header as a prefix to all source files.
Syntax
-prefix file
7.17 -noprecompile
Do not precompile any source files based upon the filename extension.
Syntax
-noprecompile
7.18 -nosyspath
Performs a search of both the user and system paths, treating
#include
statements of the form
#include <xyz>
the same as the form
#include "xyz"
.
Syntax
-nosyspath
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
85
-stdinc
This command is global.
7.19 -stdinc
Uses standard system include paths as specified by the environment variable
%MWCIncludes
%
.
Syntax
-stdinc
-nostdinc
Remarks
Add this option after all system
-I
paths.
7.20 -U+
Same as the -undefine option.
Syntax
-U+
name
7.21 -undefine
Undefines the specified symbol name.
Syntax
-u[ndefine] name
-U+name
86
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
name
The symbol name to undefine.
Remarks
This option is case-sensitive.
Chapter 7 Command-Line Options for Preprocessing
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
87
-undefine
88
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 8
Command-Line Options for Library and Linking
This chapter explains the following command-line options for library and linking.
•
•
•
8.1 -keepobjects
Retains or deletes object files after invoking the linker.
Syntax
-keepobj[ects]
-nokeepobj[ects]
Remarks
Use
-keepobjects
to retain object files after invoking the linker. Use
-nokeepobjects
to delete object files after linking. This option is global.
NOTE
Object files are always kept when compiling.
8.2 -nolink
Compiles the source files, without linking.
Syntax
-nolink
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
89
-o
Remarks
This command is global.
8.3 -o
Specifies the output filename or directory for storing object files or text output during compilation, or the output file if calling the linker.
Syntax
-o file | dir file
The output file name.
dir
The directory to store object files or text output.
90
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 9
Command-Line Options for Object Code
This chapter explains the following command-line options for object code:
•
•
•
•
•
•
9.1 -c
Instructs the compiler to compile but not invoke the linker to link the object code.
Syntax
-c
Remarks
This option is global.
9.2 -codegen
Instructs the compiler to compile without generating object code.
Syntax
-codegen
-nocodegen
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
91
-enum
Remarks
This option is global.
9.3 -enum
Specifies the default size for enumeration types.
Syntax
-enum keyword
The arguments for keyword are: int
Uses int
size for enumerated types.
min
Uses minimum size for enumerated types. This is the default.
9.4 -min_enum_size
Specifies the size, in bytes, of enumerated types.
Syntax
-min_enum_size 1 | 2 | 4
Remarks
Specifying this option also invokes the
-enum min
option by default.
9.5 -ext
Specifies which file name extension to apply to object files.
Syntax
-ext extension
92
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 9 Command-Line Options for Object Code
extension
The extension to apply to object files. Use these rules to specify the extension:
• Limited to a maximum length of 14 characters
• Extensions specified without a leading period replace the source file's extension. For example, if extension is " o
" (without quotes), then source.cpp
becomes source.o
.
• Extensions specified with a leading period (
.
extension
) are appended to the object files name. For example, if extension
is "
.o
" (without quotes), then source.cpp
becomes source.cpp.o
.
Remarks
This command is global. The default setting is
.o
.
9.6 -strings
Controls how string literals are stored and used.
Remarks
-str[ings] keyword[, ...]
The keyword arguments are:
[no]pool
All string constants are stored as a single data object so your program needs one data section for all of them.
[no]reuse
All equivalent string constants are stored as a single data object so your program can reuse them. This is the default.
[no]readonly
Make all string constants read-only. This is the default.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
93
-strings
94
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 10
Command-Line Options for Optimization
This chapter explains the following command-line options for optimization:
•
•
•
•
•
10.1 -inline
Specifies inline options. Default settings are smart
, noauto
.
Syntax
-inline keyword
The options for keyword are: off | none
Turns off inlining.
on | smart
Turns on inlining for functions declared with the inline
qualifier. This is the default.
auto
Attempts to inline small functions even if they are declared with inline
.
noauto
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
95
-ipa
Does not auto-inline. This is the default auto-inline setting.
deferred
Refrains from inlining until a file has been translated. This allows inlining of functions in both directions.
level=n
Inlines functions up to n levels deep. Level 0 is the same as
-inline on
. For n
, enter 1 to 8 levels. This argument is case-sensitive.
all
Turns on aggressive inlining. This option is the same as
-inlineon
,
-inlineauto
. This does not turn on the aggressive_inline
feature.
10.2 -ipa
Controls Interprocedural Analysis (IPA) that lets the compiler generate better optimizations by evaluating all the functions and data objects in a file or program before generating code.
Syntax
-ipa file | function | off | program | program-final function | off
Per-function optimization. This is the default option if –ipa is specified.
file
Per file optimization.
program
Per-program optimization
Remarks
NOTE
-ipa
program is available only when using the command line compiler.
96
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 10 Command-Line Options for Optimization
Using IPA mode from command-line tools is more complicated. If you specify all source files on the command-line you can just use
-ipa program
:
Use the off
or function
arguments to turn interprocedural analysis off. This is the default setting.
Use the file
argument to apply interprocedural analysis at the file level. For example, if the name of the compiler is ccaiop
, the command: ccaiop -ipa file -c file1.c file2.c
generates object code and applies this optimization to file file1.c
and then file2.c
, but does not apply the optimization across both files. For each source file, this command generates a regular object code file (a file with a name that ends with ".o" or ".obj"), which is empty. It also generates an additional file ending with ".irobj". This additional object code file contains the object code to which the compiler has applied interprocedural analysis.
This example compiles the same source files again, applies file-level analysis, then links object code into an output file named myprog
: ccaiop -o myprog -ipa file -c file1.c file2.c
Use the program
argument to apply this optimization among all files being compiled and linked. For example: ccaiop -o myprog -ipa program file1.c file2.c
generates object code, applies this optimization among all resulting object code files to link it into an output file named myprog
.
To separate compiling and linking steps for program-level interprocedural analysis, use the program
and program-final
arguments. For example: ccaiop -ipa program –c file1.c ccaiop -ipa program -c file2.c compiles file1.c
and file2.c
into empty regular object files named file1.o
and file2.o
, respectively, and optimized object files named file1.irobj
and file2.irobj
.
To link object files, refer to the regular object files or the optimized object files. For example: ccaiop -o myprog -ipa program file1.o file2.o or, equivalently: ccaiop -o myprog -ipa program file1.irobj file2.irobj
To invoke the linker directly, use the program-final
argument. For example, these commands prepare object code for linking, then invoke the linker directly (named mwld
in this example): ccaiop -ipa program-final file1.irobj file2.irobj ldaiop -o myprog file1.o file2.o
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
97
-O
NOTE
It is possible to mix program and and other IPA modes using the command-line tool options. To mix modes in the plug-in compiler you will have to use "
#pragma ipa
".
10.3 -O
Sets optimization settings to
-opt level=2
.
Syntax
-O
Remarks
Provided for backwards compatibility.
10.4 -O+
Controls optimization settings.
Syntax
-O+keyword [,...]
The keyword arguments are:
0
Equivalent to
-opt off
.
1
Equivalent to
-opt level=1
.
2
Equivalent to
-opt level=2
, peephole
3
Equivalent to
-opt level=3
, peephole.
4
98
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
Chapter 10 Command-Line Options for Optimization
Equivalent to
-opt level=4
, peephole, schedule, autoinline, func_align 16.
p
Equivalent to
-opt speed
.
s
Equivalent to
-opt space
.
Remarks
Options can be combined into a single command. Command is case-sensitive.
10.5 -opt
Specifies code optimization options to apply to object code.
Remarks
-optkeyword [,...]
The keyword arguments are: off | none
Suppresses all optimizations. This is the default.
on
Same as -opt level=2, peephole
all | full
Same as -opt speed,level=4, schedule, intrinsics, noframe peephole, autoline, func_align
16
l[evel]=num
Sets a specific optimization level. The options for num are:
•
0
- Global register allocation only for temporary values. Equivalent to
#pragma optimization_level 0
.
•
1
- Adds dead code elimination, branch and arithmetic optimizations, expression simplification. Equivalent to
#pragma optimization_level 1
.
•
2
- Adds more optimizations including, but not limited to, common subexpression elimination, copy and expression propagation, stack frame compression, stack alignment, fast floating-point to integer conversions. Equivalent to:
#pragma optimization_level 2
.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
99
-opt
•
3
- Adds more optimizations including, but not limited to, dead store elimination, live range splitting, loop-invariant code motion, strength reduction, loop transformations, loop unrolling (with
-opt speed
only), loop vectorization, lifetime-based register allocation. Equivalent to optimization_level 3
.
• 4 - Like level 3, but with more comprehensive optimizations from levels 1 and 2.
Equivalent to
#pragma optimization_level 4
.
For num
options 0 through 4 inclusive, the default is 0.
[no]space
Optimizes object code for size. Equivalent to
#pragma optimize_for_size on
.
[no]speed
Optimizes object code for speed. Equivalent to
#pragma optimize_for_size off
.
Table 10-1 explains the supported optimizations available in the compiler.
Table 10-1. Optimization Routines
Optimization Routine
Global Register Allocation or Global Register Allocation Only for Temporary Values
Dead Code Elimination
Branch Optimizations
Arithmetic Operations
Expression Simplification
Common Subexpression Elimination
Copy Propagation or Copy and Expression Propagation
Peephole Optimization
Dead Store Elimination
Live Range Splitting
Loop-Invariant Code Motion
Strength Reduction
Loop Transformations
Loop Unrolling or Loop Unrolling (Opt for Speed Only)
Explanation
Stores working values of heavily used variables in registers instead of memory.
Removes statements never logically executed or referred to by other statements.
Merges and restructures portions of the intermediate code translation in order to reduce branch instructions.
Replaces intensive computational instructions with faster equivalent instructions that produce the same result.
Replaces complex arithmetic expressions with simplified equivalent expressions.
Replaces redundant expressions with a single expression.
Replaces multiple occurrences of one variable with a single occurrence.
Applies local optimization routines to small sections of code.
Removes assignments to a variable that goes unused before being reassigned again.
Reduces variable lifetimes to achieve optimal allocation.
Shorter variable lifetimes reduce register spilling.
Moves static computations outside of a loop
Inside loops, replaces multiplication instructions with addition instructions.
Reorganizes loop object code in order to reduce setup and completion-test overhead.
Duplicates code inside a loop in order to spread branch and completion-test overhead over more operations.
Table continues on the next page...
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
100 Freescale Semiconductor, Inc.
Vectorization
Instruction Scheduling
Repeated
Stack alignment
Stack frame compression
Chapter 10 Command-Line Options for Optimization
Table 10-1. Optimization Routines (continued)
Optimization Routine
Lifetime Based Register Allocation or Register Coloring
Explanation
For processors that support vector optimizations, translates computations with code-loop arrays into equivalent vector instructions.
In a particular routine, uses the same processor register to store different variables, as long as no statement uses those variables simultaneously.
Rearranges the instruction sequence to reduce conflicts among registers and processor resources.
Iterates the optimization routines listed between {* and *}.
Aligns the stack pointer. Required when load/store instruction has an alignment requirement.
Performs live/dead analysis on all non-static symbols
(compiler generated temporaries, automatics, input parameters, and outgoing parameters). Using the live/dead information, all the dead stores are eliminated.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
101
-opt
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
102 Freescale Semiconductor, Inc.
Chapter 11
Command-Line
This chapter describes how to use the command-line tools to generate, examine, and manage source code and object code.
•
•
Specifying Source File Locations
•
Licensing Command-Line Options
•
Diagnostic Command-Line Options
•
Library and Linking Command-Line Options
•
Code Generation Command-Line Options
•
Optimization Command-Line Options
11.1 Naming Conventions
Table 11-1 lists the names of the CodeWarrior command line tools.
asmaiop ccaiop ldaiop
This tool...
Table 11-1. AIOP command line tools does these tasks...
translates assembly language source code into object code translates C and C++ source code into object code links object code into a loadable image file
11.2 Specifying Source File Locations
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
103
Licensing Command-Line Options
The build tools use several environment variables at build time to search for include
files , libraries, and other source files. All of the variables mentioned here are lists which are separated by semicolons ("
;
") in Windows operating systems and colons ("
:
") in Linux operating systems.
Unless
-nodefaults
is passed to on the command line, the compiler searches for an environment variable called
MWCEABIPPCIncludes
or
MWCIncludes
(in that order). These variables contain a list of system access paths to be searched after the system access paths specified by the user. The assembler also does this, using the variables
MWAsmEABIPPCIncludes
or
MWAsmIncludes
.
Analogously, unless
-nodefaults
or
-disassemble
is given, the linker will search the environment for a list of system access paths and system library files to be added to the end of the search and link orders. The variable
MWEABIPPCLibraries
or
MWLibraries
contains a list of system library paths to search for files, libraries, and command files.
Associated with this list is the variable
MWEABIPPCLibraryFiles
or
MWLibraryFiles
which contains a list of libraries (or object files or command files) to add to the end of the link order. These files may be located in any of the cumulative access paths at runtime.
If you are only building for one target, it is okay to use
MWCIncludes
,
MWAsmIncludes
,
MWLibraries
, and
MWLibraryFiles
. The target-specific versions of the variables come in handy when targeting multiple targets, since the target-specific variables override the generic variables. Note that if the target-specific variable exists, the generic variable will not be used; the contents of the two variables will not be combined.
11.3 Licensing Command-Line Options
This chapter explains the following command-line options:
•
•
11.3.1 -fullLicenseSearch
Continues the search for a license file on the host computer.
Syntax
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
104 Freescale Semiconductor, Inc.
Chapter 11 Command-Line
-fullLicenseSearch
Remarks
A license file unlocks features and capabilities in CodeWarrior tools. This option extends the normal search for a valid license.dat
file.
Each time they are invoked, the command-line compiler, stand-alone assembler, and linker search on the host computer in this order until they find a valid license file in this order:
• the directory specified in a
-license
option
• the directory containing the command-line tool
• the current working directory
• the directory containing the CodeWarrior IDE
When this option is not used, the tool stops when it finds a valid license file. With this option, the tool searches all paths to read all valid licenses.
11.3.2 -license
Specifies a location on the host computer to search for a license file.
Syntax
-license
location where location is the path of a directory that contains a valid license file named license.dat
.
Remarks
A license file unlocks features and capabilities in CodeWarrior tools.
11.4 Diagnostic Command-Line Options
This chapter explains the following command-line options:
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
105
Diagnostic Command-Line Options
•
•
•
•
•
•
•
•
•
•
11.4.1 -g
Generates DWARF 1. x-conforming debugging information.
Syntax
-g[dwarf]
Remarks
This option is global. This option is equivalent to
-sym dwarf-1,full
11.4.2 -gdwarf-2
Generates DWARF-2. x-conforming debugging information.
Syntax
-gdwarf-2
Remarks
This option is global. This option is equivalent to
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
106 Freescale Semiconductor, Inc.
-sym dwarf-2,full
11.4.3 -fmt
Equivalent to the
-format
option.
Syntax
-fmt x | nox
11.4.4 -format
Specifies the style of mnemonics to show in disassemblies.
Syntax
-format x | nox
Remarks
To show extended mnemonics in a disassembly, use
-format x
This option is the default.
To show regular mnemonics in a disassembly, use
-format nox
This is a linker option.
Chapter 11 Command-Line
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
107
Diagnostic Command-Line Options
11.4.5 -listclosure
Controls the appearance of symbol closures in the linker's map file.
Syntax
-listclosure
-nolistclosure
Remarks
This option also generates a map file if the
-map
option has not already been specified.
This is a linker option.
11.4.6 -listdwarf
Controls the appearance of DWARF debugging information in the linker's map file.
Syntax
-listdwarf
-nolistdwarf
Remarks
This option also generates a map file if the
-map
option has not already been specified.
This is a linker option.
11.4.7 -map
Generates a text file that describes the contents of the linker's output file.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
108 Freescale Semiconductor, Inc.
Chapter 11 Command-Line
Syntax
-map [
filename]
Remarks
The default value for filename is the name of the linker's output file with a
.MAP
file name extension.
This is a linker option.
11.4.8 -mapunused
Controls the appearance of a list of unused symbols in the map file.
Syntax
-mapunused
-nomapunused
Remarks
This option also generates a map file if the
-map
option has not already been specified.
This is a linker option.
11.4.9 -sym
Specifies global debugging options.
Syntax
-sym
keyword[,...]
The choices for keyword are:
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
109
Library and Linking Command-Line Options
off
Do not generate debugging information. This option is the default.
on
Generate DWARF-1-conforming debugging information.
dwarf-1
Generate DWARF-1-conforming debugging information.
full[path]
Store absolute paths of source files instead of relative paths.
dwarf-2
Generate DWARF-2-conforming debugging information.
11.4.10 -unused
Equivalent to the
-mapunused
option.
Syntax
-unused
-nounused
11.5 Library and Linking Command-Line Options
This chapter explains the following command-line options:
•
•
•
•
•
•
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
110 Freescale Semiconductor, Inc.
Chapter 11 Command-Line
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
• -sortsrec
•
•
•
•
•
•
•
•
•
•
•
•
11.5.1 -codeaddr
Sets the runtime address of the executable code.
Syntax
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
111
Library and Linking Command-Line Options
-codeaddr
addr
Remarks
The addr value is an address, in decimal or hexadecimal format. Hexadecimal values must begin with
0x
. The default is
65536
.
The linker ignores this option if you invoke the linker with the
-lcf
option.
This is a linker option.
11.5.2 -dataaddr
Sets the loading address of the data.
Syntax
-dataaddr
addr
Remarks
The addr value is an address, in decimal or hexadecimal format. Hexadecimal values must begin with
0x
. The default is the address after the code and large constant sections.
The linker ignores this option if the linker is invoked with the
-lcf
option.
This is a linker option.
11.5.3 -genbinary
Controls the generation of a binary file.
Syntax
-genbinary
none
|
one
|
multiple
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
112 Freescale Semiconductor, Inc.
Chapter 11 Command-Line
Remarks
To generate no binary file even if s-record generation is on, use
-genbinary none
This option is the default.
To generate a single binary file with all the loadable code and data, even if s-record generation is off, use
-genbinary one
To generate separate binary files for each MEMORY directive, even if s-record generation is off, use
-genbinary multiple
This is a linker option.
11.5.4 -gap_fill
Fills the gap in s-record file.
Syntax
-gap_fill
Remarks
This option fill the gaps caused by alignment.
11.5.5 -heapaddr
Sets the runtime address of the heap.
Syntax
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
113
Library and Linking Command-Line Options
-heapaddr
addr
Remarks
The addr value is an address, in decimal or hexadecimal format. Hexadecimal values must begin with
0x
. The default is
stack_address - (
heap_size +
stack_size) where stack_address is the address of the stack, heap_size is the size of the heap, and
stack_size
is the size of the stack.
This is a linker option.
11.5.6 -heapsize
Sets the runtime size of the heap, in kilobytes.
Syntax
-heapsize
size
Remarks
The default value for size is
1024
.
This is a linker option.
11.5.7 -lcf
Uses the code and data addresses specified in a linker command file.
Syntax
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
114 Freescale Semiconductor, Inc.
Chapter 11 Command-Line
-lcf
filename
Remarks
The filename argument is the name of a linker command file. The file must have a
.lcf
file name extension. The linker ignores the
-codeaddr
,
-dataaddr
,
-sdataaddr
, and
-sdata2addr options if it uses the
-lcf
option.
This is a linker option.
11.5.8 -library
Generates a static library.
Syntax
-library
Remarks
This option is global. This is a linker option.
11.5.9 -linkmode
Controls the performance of the linker.
Syntax
-linkmode
keyword
The choices for keyword are: lessram
Use little memory but take more processing time.
normal
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
115
Library and Linking Command-Line Options
Use a medium amount of memory for medium processing time. This is the default.
moreram
Use lots of memory to improve processing time.
Remarks
This is a linker option.
11.5.10 -main
Specifies the main entry point for the executable image.
Syntax
-m[ain]
symbol
Remarks
The maximum length of symbol is 63 characters. The default is
__start
.
This is a linker option.
11.5.11 -model
Specifies the addressing mode that the linker uses when resolving references.
Syntax
-model
keyword
The choices for keyword are: absolute
Use absolute executable and data addressing. This choice is the default.
sda_pic_pid
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
116 Freescale Semiconductor, Inc.
Chapter 11 Command-Line
Use position-independent addressing executable code and data.
Remarks
This is a linker option.
11.5.12 -noentry
Specifies no entry point for the executable image.
Syntax
-noentry
Remarks
The linker uses the main entry point to determine which objects/functions to add to your application that are referenced from that entry point. In absence of an entry point, the application will be empty (completely deadstripped) resulting in an linker error.
There are several ways to pass other entry points to the linker for objects that are not referenced from the main entry point.
• use the linker command file directives
TERM
or
INIT
• use
__declspec(export)
• use the lcf directives
FORCEFILES
or
FORCEACTIVE
For example, if you have a simple reset vector function which simply calls your startup code (call the startup code __start and __reset for the reset vector function for this example), you could do the following :
• use
-m __start
on the command line
• use
ENTRY(__start)
in the Linker Command File
• use
INIT(__reset)
in the Linker Command File
• use
FORCEACTIVE(__reset)
in the Linker Command File
• use
__declspec(export) void __reset(void) {__start;}
in the source.
11.5.13 -nomain
Equivalent to
Syntax
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
117
Library and Linking Command-Line Options
-nomain
11.5.14 -opt_partial
Finishes a partial link operation.
Syntax
-opt_partial
Remarks
This option allows the use of a linker command file, creates tables for C++ static constructors, C++ static destructors, and C++ exceptions. This option also tells the linker to build an executable image even if some symbols cannot be resolved.
This is a linker option.
11.5.15 -partial
Does not report error messages for unresolved symbols.
Syntax
-partial
Remarks
This option tells the linker to build a reloadable object file even if some symbols cannot be resolved.
This is a linker option.
11.5.16 -r
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
118 Freescale Semiconductor, Inc.
Chapter 11 Command-Line
Equivalent to
-partial
.
Syntax
-r
Remarks
This option tells the linker to build a reloadable object file even if some symbols cannot be resolved.
This is a linker option.
11.5.17 -r1
Equivalent to
-opt_partial
.
Syntax
-r1
Remarks
This option allows the use of a linker command file, creates tables for C++ static constructors, C++ static destructors, and C++ exceptions. This option tells the linker to build a reloadable object file even if some symbols cannot be resolved.
This is a linker option.
11.5.18 -r2
Equivalent to
-resolved_partial
.
Syntax
-r2
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
119
Library and Linking Command-Line Options
This option first allows the use of a linker command file, creates tables for C++ static constructors, C++ static destructors, and C++ exceptions.
This is a linker option.
11.5.19 -relocate
Controls the use of relocations contained in DWARF 1. x sections when generating a disassembly of non-executable files.
Syntax
-relocate
-norelocate
Remarks
The
-relocate
option tells the linker to use data in the
.rela.text
and
.rela.debug
parts of the DWARF 1.x when generating disassemblies.
The
-relocate
option is the default when disassembling executable files with DWARF 1. x sections and any type of file with DWARF 2. x sections.
This is a linker option
11.5.20 -resolved_partial
Finishes a partial link operation and issues error messages for unresolved symbols.
Syntax
-resolved_partial
Remarks
This option first allows the use of a linker command file, creates tables for C++ static constructors, C++ static destructors, and C++ exceptions.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
120 Freescale Semiconductor, Inc.
Chapter 11 Command-Line
This is a linker option.
11.5.21 -sdataaddr
Sets the loading address of small data.
Syntax
-sdataaddr addr
Remarks
The addr value is an address, in decimal or hexadecimal format. Hexadecimal values must begin with
0x
. The default is the address after the large data section.
This option essentially defines the value of the
__SDA_BASE__
symbol.
__SDA_BASE__
is used in startup code to set the initial value of the small data area base pointer, which, by definition in the EABI is r13.
The linker ignores this option if the linker is invoked with the
-lcf
option.
__SDA_BASE__ will automatically be generated by the linker if
–lcf <file>
is used.
This is a linker option.
11.5.22 -sdata2addr
Sets the loading address of small constant data.
Syntax
-sdata2addr addr
Remarks
The addr value is an address, in decimal or hexadecimal format. Hexadecimal values must begin with
0x
. The default is the address after the small data section.
This option essentially defines the value of the
__SDA2_BASE__
symbol.
__SDA2_BASE__
is used in startup code to set the initial value of the small constant data area base pointer, which, by definition in the EABI is r2.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
121
.
Library and Linking Command-Line Options
The linker ignores this option if the linker is invoked with the
-lcf
option.
__SDA_BASE__ will automatically be generated by the linker if
–lcf <file>
is used. This is a linker option.
11.5.23 -sdatathreshold
Limits the size of the largest objects in the small data section.
Syntax
-sdata[threshold] size
Remarks
The size value specifies the maximum size, in bytes, of any object in the small data section (typically named ".sdata"). The linker places objects that are greater than this size in the data section (typically named ".data") instead.
You can force a variable into the small data section in your source code like this:
__declspec(section ".sdata") extern int bigobj[25];
The default value for size is
8
.
This is a linker option.
11.5.24 -sdata2threshold
Limits the size of the largest objects in the small constant data section.
Syntax
-sdata2[threshold] size
Remarks
The size value specifies the maximum size, in bytes, of any object in the small constant data section (typically named ".sdata2"). The linker places constant objects that are greater than this size in the constant data section (typically named ".rodata") instead.
You can force a variable into the small constant data section in your source code like this:
__declspec(section ".sdata2") extern int bigobj[] = { 0, 1, 2, 3, 4, 5, 6, 7, 8, 9 };
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
122 Freescale Semiconductor, Inc.
Chapter 11 Command-Line
The default for size is
8
. This is a linker option.
11.5.25 -show
Specifies the information to list in a disassembly.
Syntax
-show
keyword[,...]
The choices for keyword are: only | none
Shows no disassembly. Begin a list of choices with only
or none
to prevent default information from appearing in the disassembly.
all
Shows binary, executable code, detailed, data, extended, and exception information in the disassembly.
binary | nobinary
Shows or does not show address and op-code values.
code | nocode
Shows or does not show executable code sections.
text | notext
Equivalent to the code
and nocode
choices, respectively.
data | nodata
Shows or does not show data sections.
detail | nodetail
Shows or does not show extra information.
extended | noextended
Shows or does not show extended mnemonics.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
123
Library and Linking Command-Line Options
exceptions | noexceptions
Shows or does not show C++ exception tables. This option also shows data sections.
xtab[les] | noxtab[les]
Equivalent to the exceptions
and noexceptions
choices, respectively.
headers | noheaders
Shows or does not show object header information.
debug | nodebug
Shows or does not show debugging information.
dwarf | nodwarf
Equivalent to the debug
and nodebug
choices, respectively.
tables | notables
Shows or does not show character string and symbol tables.
source | nosource
Interleaves the code dissassembly with c or c++ source code.
Remarks
The default setting for this option is
-show binary,code,data,extended,headers,tables
This is a linker option.
11.5.26 -srec
Generates an S-record file.
Syntax
-srec [
file-name]
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
124 Freescale Semiconductor, Inc.
Chapter 11 Command-Line
The default value for file-name is the name of the linker's output file with a
.mot
file name extension.
This is a linker option.
11.5.27 -sreceol
Specifies the end-of-line style to use in an S-record file.
Syntax
-sreceol
keyword
The choices for keyword are: mac
Use Mac OS®-style end-of-line format.
dos
Use Microsoft® Windows®-style end-of-line format. This is the default choice.
unix
Use a UNIX-style end-of-line format.
Remarks
This option also generates an S-record file if the
-srec
option has not already been specified.
This is a linker option.
11.5.28 -sreclength
Specify the length of S-records.
Syntax
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
125
Library and Linking Command-Line Options
-sreclength
value
The choices for value are from 8 to 255. The default is 26.
Remarks
This option also generates an S-record file if the
-srec
option has not already been specified.
This is a linker option.
11.5.29 -stackaddr
Sets the runtime address of the stack.
Syntax
-stackaddr addr
Remarks
The addr value is an address, in decimal or hexadecimal format. Hexadecimal values must begin with
0x
. The default is
0x3dff0
.
This option essentially defines the value of the
_stack_addr
symbol. The
_stack_addr symbol is used at startup to define the initial value of the stack pointer register, r1.
This is a linker option.
11.5.30 -stacksize
Sets the runtime size of the stack, in kilobytes.
Syntax
-stacksize size
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
126 Freescale Semiconductor, Inc.
Chapter 11 Command-Line
The default value for size is
64
. This option essentially defines the value of the
_stack_size symbol. The
_stack_size
symbol is not currently used.
This is a linker option.
11.5.31 -strip_partial
Removes unreferenced objects on a partially linked image.
Syntax
-strip_partial
Remarks
Use this option with either the
-opt_partial
or
-resolved_partial
options.
This is a linker option.
11.5.32 -tune_relocations
Ensures that references made by the linker conform to the Embedded application binary interface (EABI) or position-independent Application binary interface (ABI).
Syntax
-tune_relocations
Remarks
Use this option only with the
-abi eabi
and
-abi sda_pic_pid
option to ensure that references in the executable image conform to these ABIs.
To conform to both of these ABIs, the linker will modify relocations that do not reach the desired executable code. The linker first converts near branch instructions to far branch instructions. Then it will convert absolute branches to PC-relative branches. For branches that cannot be converted to far or PC-relative addressing, the linker will generate branch islands.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
127
Library and Linking Command-Line Options
To conform to the SDA PIC/PID ABI, the linker will generate the appropriate style of addressing.
This option is global. This is a linker option.
11.5.33 -xtables
Equivalent to
-show exceptions
or
-show noexceptions
.
Syntax
-xtables on | off
Remarks
This is a linker option.
11.5.34 -stdlib
Uses standard system library access paths as specified by the environment variable
%MWEABIPPCLibraries%to add system libraries as specified by the environment variable
%MWEABIPPCLibraryFiles% at the end of link order
.
Syntax
-stdlib
-nostdlib
Remarks
-nostdlib
disables searching for paths or files in the environment variables.
This command is global. This is a linker option.
11.5.35 -L+
Adds a new library search path to the default settings list.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
128 Freescale Semiconductor, Inc.
Chapter 11 Command-Line
Syntax
-L+path
-l path
is identical to
–L+path
.
The parameters are: path
The search path to append.
Remarks
This command is global and case-sensitive.
11.5.36 -lr
Adds a recursive library search path to the default settings list.
Syntax
-lr path
The parameters are: path
The recursive library search path to append.
Remarks
This command is global. This is a linker option.
11.5.37 -l+
Adds a library by searching access path for a specified library filename.
Syntax
-l+file
The parameters are:
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
129
Code Generation Command-Line Options
file
Name of the library file to search.
Remarks
The linker searches access path for the specified lib<file>.<ext>
, where
<ext>
is
.a
or
.lib
.
If the file is not found then search for
<file>
. This command is case-sensitive.
11.6 Code Generation Command-Line Options
This chapter explains the following command-line options:
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
11.6.1 -abi
Chooses which ABI to conform to.
Syntax
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
130 Freescale Semiconductor, Inc.
Chapter 11 Command-Line
-abi keyword
The choices for keyword are: eabi
Use the Embedded ABI. This choice is the default.
sda_pic_pid
Use position-independent addressing executable code and data.
Remarks
This option is global.
11.6.2 -align
Specifies structure and array alignment.
Syntax
-align keyword[,...]
The choices for keyword are: power[pc]
Use conventional alignment. This choice is the default.
array[members]
Align members of arrays, too.
11.6.3 -big
Generates object code and links an executable image to use big-endian data formats.
Syntax
-big
Remarks
This is the default setting for the compiler and linker.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
131
Code Generation Command-Line Options
11.6.4 -common
Moves uninitialized data into a common section.
Syntax
-common on | off
Remarks
The default is off
.
11.6.5 -fp
Controls floating-point code generation.
Syntax
-fp keyword
The choices for keyword are: none | off
No floating point code generation.
soft[ware]
Use software libraries to perform floating-point operations. This is the default.
hard[ware]
Use the processor's built-in floating-point capabilities to perform floating-point operations. This is not applicable to AIOP as AIOP contains no hardware floating point support.
11.6.6 -func_align
Specifies alignment of functions in executable code.
Syntax
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
132 Freescale Semiconductor, Inc.
Chapter 11 Command-Line
-func_align 4 | 8 | 16 | 32 | 64 | 128
Remarks
The default alignment is
4
. However, at an optimization level
4
, the alignment changes to
16
. If you are using
-func_align 4
(or none
) and if you are compiling for VLE, then the linker will compress gaps between VLE functions:
• if those functions are not called by a Classic PPC function
• the function has an alignment greater than
4
.
11.6.7 -little
Generates object code and links an executable image to use little-endian data formats.
Syntax
-little
11.6.8 -pool
Controls the grouping of similar-sized data objects.
Syntax
-pool[data] on | off
Remarks
Use this option to reduce the size of executable object code in functions that refer to many object of the same size. These similar-sized objects do not need to be of the same type. The compiler only applies this option to a function if the function refers to at least 3 similar-sized objects. The objects must be global or static.
At the beginning of the function, the compiler generates instructions to load the address of the first similar-sized object. The compiler then uses this address to generate 1 instruction for each subsequent reference to other similar-sized objects instead of the usual 2 instructions for loading an object using absolute addressing.
This option is equivalent to the pool_data
pragma.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
133
Code Generation Command-Line Options
The default is on
.
11.6.9 -processor
Generates and links object code for a specific processor.
Syntax
-proc[essor] keyword
The choices for keyword are:
Zen generic
Remarks
The keyword parameter specifies the processor core. The default for keyword is generic
.
11.6.10 -ppc_asm_to_vle
Converts regular assembler mnemonics to equivalent VLE (Variable Length Encoded) assembler mnemonics in the inline assembler and standalone assembler.
Syntax
-ppc_asm_to_vle
Remarks
C/C++ source is input to compiler inline assembler whereas assembly source is input to standalone assembler.
While translating assembly statements in C or C++ or assembly source code, the compiler inline assembler or standalone assembler will replace each regular assembler mnemonic with its matching VLE instruction if one exists. The mnemonics for VLE
(Variable Length Encoding) instructions begin with se_
or e_
. The compiler's inline assembler recognizes these mnemonics when the compiler is configured to generate VLE
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
134 Freescale Semiconductor, Inc.
Chapter 11 Command-Line
object code. The same is with assembler. If an equivalent VLE instruction does not exist for an UISA instruction then it might generate multiple VLE instructions for that operation.
Assembler changes the section
.text
to
.text_vle
. For user defined sections, LCF needs modification.
Example
Consider test.s
.section .mytext
tmp: do_loop00:
B exit0
Add r0,r0,r1
.text
exit0:
blr
Assemble with option
-ppc_asm_to_vle
In LCF
.text_vle (VLECODE) : {
*(.text)
*(.text_vle)
(.mytext)
}
NOTE
Assembler does not emit a 16-bit VLE instruction for branch instructions, if the label definition does not exist at that point.
11.6.11 -rambuffer
Specifies a runtime address in which to store the executable image in RAM so that it may be transferred to flash memory.
Syntax
-rambuffer
addr
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
135
Code Generation Command-Line Options
This option specifies information for a legacy flashing tool. This tool required that the executable image must first be loaded to an area in RAM before being transferred to
ROM. Do not use this option if your flash memory tool does not follow this behavior.
The CodeWarrior tools ignore this option if it is not used with the
-romaddr
option.
11.6.12 -readonlystrings
Places string constants in a read-only section named .rodata.
Syntax
-readonlystrings
11.6.13 -relax_ieee
Controls the use of relaxed IEEE floating point operations.
Syntax
-relax_ieee
-norelax_ieee
Remarks
The default is
-relax_ieee
.
11.6.14 -romaddr
Generates a ROM image and specifies the image's starting address at runtime.
Syntax
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
136 Freescale Semiconductor, Inc.
Chapter 11 Command-Line
-romaddr
addr
11.6.15 -rostr
Equivalent to the
-readonlystrings
option.
Syntax
-rostr
11.6.16 -schedule
Controls the rearrangement of instructions to reduce the effects of instruction latency.
Syntax
-schedule on | off
Remarks
The default is off
.
11.6.17 -strict_ieee
Specifies the use of strict IEEE floating point operations.
Syntax
-strict_ieee
Remarks
This option is the same as the
-norelax_ieee
option.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
137
Code Generation Command-Line Options
11.6.18 -use_lmw_stmw
Controls the use of multiple load and store instructions for function prologues and epilogues.
Syntax
-use_lmw_stmw on | off
Remarks
The default is off.
11.6.19 -use_isel
Controls the use of isel
instructions.
Syntax
-use_isel on | off
Remarks
The default is off
.
11.6.20 -vle
Controls the use of the Variable Length Encoded (VLE) instruction set.
Syntax
-vle
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
138 Freescale Semiconductor, Inc.
Chapter 11 Command-Line
This option tells the compiler and linker to generate and lay out Variable Length Encoded
(VLE) instructions. VLE instructions give extra flexibility in instruction encoding and alignment, allowing the compiler and linker to greatly reduce the size of runtime object code with only a small penalty in execution performance.
This option also turns on the
-processor Zen
option.
11.7 Optimization Command-Line Options
This chapter explains the following command-line options:
•
•
•
•
•
11.7.1 -code_merging
Removes duplicated functions to reduce object code size.
Syntax
-code_merging keyword[,...]
The choices for keyword are: all
Use the all
argument to specify that the linker should remove all duplicate functions except one.
safe
Use the safe
argument to specify that only duplicate functions marked as weak should be reduced to one function.
aggressive
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
139
Optimization Command-Line Options
Use the aggressive
option to specify that the linker should ignore references to function addresses when considering which duplicate functions to remove.
off
Use the off
argument to disable code merging optimization.
Remarks
This linker optimization removes duplicate copies of functions with identical executable code.
The linker does not apply this optimization to functions that have been declared with the
__declspec(no_linker_opts)
directive.
By default the code merging optimization is off.
11.7.2 -far_near_addressing
Simplifies address computations to reduce object code size and improve performance.
Syntax
-far_near_addressing
-nofar_near_addressing
Remarks
This linker optimization simplifies address computations in object code. If an address value is within the range that can be stored in the immediate field of the load immediate instruction, the linker replaces the address's two-instruction computation with a single instruction. An address value that is outside this range still requires two instructions to compute.
The ranges of values that may be stored in the immediate field is
-0x7fff
to
0x8000
for the regular li
instruction and
-0x7ffff
to
0x80000
for e_li
, the VLE (Variable Length
Encoding) instruction.
The linker does not apply this optimization to functions that have been declared with the
__declspec(no_linker_opts)
directive.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
140 Freescale Semiconductor, Inc.
Chapter 11 Command-Line
11.7.3 -vle_bl_opt
Replaces branch instructions to reduce object code size.
Syntax
-ble_bl_opt
-noble_bl_opt
Remarks
This linker optimization replaces each 32-bit e_bl
instruction with a 16-bit se_bl instruction for a function call when the span of memory between the calling function and called function is sufficiently close.
This optimization requires that the target processor has the Variable Length Encoding
(VLE) extension.
The linker does not apply this optimization to functions that have been declared with the
__declspec(no_linker_opts)
directive.
11.7.4 -vle_enhance_merging
Removes duplicated functions that are called by functions that use VLE instructions to reduce object code size.
Syntax
-vle_enhance_merging
-novle_enhance_merging
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
141
AIOP Specific Command-Line Options
Remarks
When applying the code merging optimization (
-code_merging
), this linker optimization ensures that function calls that use VLE (Variable Length Encoding) instructions will still be able to reach a function that has been removed. This optimization replaces the 16-bit se_bl
instruction with a 32-bit e_bl
instruction.
When this option is not used, the linker does not merge functions that are called by functions that use VLE instructions.
This optimization requires that the target processor has the Variable Length Encoding
(VLE) extension. This optimization has no effect when the linker is not applying the code merging optimization.
The linker does not apply this optimization to functions that have been declared with the
__declspec(no_linker_opts)
directive.
11.7.5 -volatileasm
Controls whether or not inline assembly statements will be optimized.
Syntax
-volatileasm
-novolatileasm
11.8 AIOP Specific Command-Line Options
This chapter explains the following command-line options:
•
•
•
•
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
142 Freescale Semiconductor, Inc.
Chapter 11 Command-Line
•
•
11.8.1 -perfmarks
Enables performance marker functionality.
Syntax
-perfmarks
Remarks
This option allows the compiler to generate performance marker symbols used by the point-to-point profiler in CodeWarrior for Advanced Packet Processing Development
Suite. Performance markers are specially names symbols in an ELF which the point-topoint profiler uses to identify regions of code to profile. These performance markers are generated when the user specifies one of following three types of performance marker pragmas:
•
#pragma perfmark start <marker_name>
•
#pragma perfmark stop <marker_name>
•
#pragma perfmark abort <marker_name>
The
–perfmarks
option must be specified for any of the above three pragmas to have any effect. If
–perfmarks
is not specified, the pragma will be parsed, but no performance markers will be generated.
The
–perfmarks
option may also be enabled using the CodeWarrior for Advanced Packet
Processing Development Suite by selecting the relevent option on the Power PC
Compiler->Processor Settings panel.
11.8.2 -aiop
Enables AIOP specific code generation.
Syntax
-aiop
Remarks
This option sets the VLE option, which further sets the processor to Zen(e200).
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
143
AIOP Specific Command-Line Options
11.8.3 -gen_ld_std
Generates e_ldw/e_stdw e200z490 instructions for 64-bit data load store operations.
Syntax
-gen_ld_std
Remarks
This option sets the processor to Zen(e200).
11.8.4 -fcoverage
Generate GCOV coverage information files.
Syntax
-fcoverage
11.8.5 -[no]stackinfo
Generate stack usage information.
Syntax
-[no]stackinfo
Remarks
Default value is enabled. Stack usage information generation can be disabled using the
– nostackinfo
option.
11.8.6 -[no]cases_array
Enables the
directive.
Syntax
-[no]cases_array
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
144 Freescale Semiconductor, Inc.
Chapter 11 Command-Line
Default value is on (pragma is enabled by default). Pragma operation can be disabled using the
–nocases_array
option. See the
#pragma
information on this pragma.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
145
AIOP Specific Command-Line Options
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
146 Freescale Semiconductor, Inc.
Chapter 12
Assembler
This chapter describes the assembler:
•
•
•
•
12.1 Syntax
This section describes the syntax of assembly language statements. It consists of these topics:
•
•
•
•
•
•
•
12.1.1 Assembly Language Statements
The three types of assembly language statements are:
• Machine instructions
• Macro calls
• Assembler directives
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
147
Syntax
Instructions, directives, and macro names are case insensitive: the assembler considers
LWZ, Lwz, and lwz to be the same instruction
Remember these rules for assembly language statements:
1. The maximum length of a statement or an expanded macro is 512 characters.
2. A statement must reside on a single line. However, you can concatenate two or more lines by typing a backslash (\) character at the end of the line.
3. Each line of the source file can contain only one statement unless the assembler is running in GNU mode. (This mode allows multiple statements on one line, with semicolon separators.)
12.1.2 Statement Syntax
Statement Syntax shows the syntax of an assembly language statement. Table 12-1
describes the elements of this syntax.
Listing: Statement Syntax
statement ::= [
symbol ]
operation [
operand ] [ ,
operand ]... [
comment ]
operation ::=
machine_instruction |
assembler_directive |
macro_call
operand ::=
symbol |
constant |
expression |
register_name symbol machine_instruction assembler_directive macro_call constant expression
Element
Table 12-1. Syntax Elements
Description
A combination of characters that represents a value.
A machine instruction for your target processor.
A special instruction that tells the assembler how to process other assembly language statements. For example, certain assembler directives specify the beginning and end of a macro.
A statement that calls a previously defined macro.
A defined value, such as a string of characters or a numeric value.
A mathematical expression.
Table continues on the next page...
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
148 Freescale Semiconductor, Inc.
register_name comment
Chapter 12 Assembler
Table 12-1. Syntax Elements (continued)
Element Description
The name of a register; these names are processor-specific.
Text that the assembler ignores, useful for documenting your code.
12.1.3 Symbols
A symbol is a group of characters that represents a value, such as an address, numeric constant, string constant, or character constant. There is no length limit to symbols.
The syntax of a symbol is:
symbol ::= label | equate
In general, symbols have file-wide scope. This means:
1. You can access the symbol from anywhere in the file that includes the symbol definition.
2. You cannot access the symbol from another file.
However, it is possible for symbols to have a different scope, as described in the following sub-sections.
•
•
•
•
•
•
12.1.3.1 Labels
A label is a symbol that represents an address. A label's scope depends on whether the label is local or non-local.
The syntax of a label is:
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
149
Syntax
label ::= local_label [ : ] | non-local_label[ : ]
The default settings are that each label ends with a colon (:), a label can begin in any column. However, if you port existing code that does not follow this convention, you should clear the Labels must end with ':' checkbox of the Assembler settings. After you clear the checkbox, you may use labels that do not end with colons, but such labels must begin in column 1.
12.1.3.2 Non-Local Labels
A non-local label is a symbol that represents an address and has file-wide scope. The first character of a non-local label must be a:
• letter (a-z or A-Z),
• period (.),
• question mark (?), or an
• underscore (_).
Subsequent characters can be from the preceding list or a:
• numeral (0-9), or
• dollar sign ($).
12.1.3.3 Local Labels
A local label is a symbol that represents an address and has local scope: the range forward and backward within the file to the points where the assembler encounters nonlocal labels.
The first character of a local label must be an at-sign (@). The subsequent characters of a local label can be:
• letters (a-z or A-Z)
• numerals (0-9)
• underscores (_)
• question marks (?)
• dollar sign. ($)
• periods (.)
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
150 Freescale Semiconductor, Inc.
Chapter 12 Assembler
NOTE
You cannot export local labels; local labels do not appear in debugging tables.
Within an expanded macro, the scope of local labels works differently:
• The scope of local labels defined in macros does not extend outside the macro.
• A non-local label in an expanded macro does not end the scope of locals in the unexpanded source.
Local Label Scope in a Macro shows the scope of local labels in macros: the
@SKIP
label defined in the macro does not conflict with the
@SKIP
label defined in the main body of code.
Listing: Local Label Scope in a Macro
MAKEPOS .MACRO
cmpwi 0,r3,0
bge @SKIP
neg r3,r3
@SKIP: ; Scope of this label is within
; the macro
.ENDM
START:
lwz r3,COUNT
cmpw 0,r3, r4
beq @SKIP
MAKEPOS
@SKIP: ; Scope of this label is START to
; END excluding lines arising
; from macro expansion
addic r3,r3,1
END: blr
12.1.3.4 Relocatable Labels
The assembler assumes a flat 32-bit memory space. You can use the expressions of
to specify the relocation of a 32-bit label.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
151
Syntax
label
label@l
label@h
label@ha
label@sdax
label@got
NOTE
The assembler for your target processor may not allow all of these expressions.
Table 12-2. Relocatable Label Expressions
Expression Represents
The offset from the address of the label to the base of its section, relocated by the section base address. It also is the
PC-relative target of a branch or call. It is a 32-bit address.
The low 16-bits of the relocated address of the symbol.
The high 16-bits of the relocated address of the symbol. You can OR this with label@l to produce the full 32-bit relocated address.
The adjusted high 16-bits of the relocated address of the symbol. You can add this to label@l to produce the full 32-bit relocated address.
For labels in a small data section, the offset from the base of the small data section to the label. This syntax is not allowed for labels in other sections.
For processors with a global offset table, the offset from the base of the global offset table to the 32-bit entry for label.
12.1.3.5 Equates
An equate is a symbol that represents any value. To create an equate, use the
.equ
or
.set
directive.
The first character of an equate must be a:
• letter (a-z or A-Z),
• period (.),
• question mark (?), or
• underscore (_)
Subsequent characters can be from the preceding list or a:
• numeral (0-9) or
• dollar sign ($).
The assembler allows forward equates. This means that a reference to an equate can be in a file before the equate's definition. When an assembler encounters such a symbol whose value is not known, the assembler retains the expression and marks it as unresolved. After the assembler reads the entire file, it reevaluates any unresolved expressions. If
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
152 Freescale Semiconductor, Inc.
Chapter 12 Assembler
necessary, the assembler repeatedly reevaluates expressions until it resolves them all or cannot resolve them any further. If the assembler cannot resolve an expression, it issues an error message.
NOTE
The assembler must be able to resolve immediately any expression whose value affects the location counter. If the assembler can make a reasonable assumption about the location counter, it allows the expression.
The code of
shows a valid forward equate.
Listing: Valid Forward Equate
.data
.long alloc_size alloc_size .set rec_size + 4
; a valid forward equate on next line rec_size .set table_start-table_end
.text
;...
table_start:
; ...
table_end:
However, the code of Invalid Forward Equate is not valid. The assembler cannot
immediately resolve the expression in the
.space
directive, so the effect on the location counter is unknown.
Listing: Invalid Forward Equate
;invalid forward equate on next line rec_size .set table_start-table_end
.space rec_size
.text; ...
table_start:
; ...
table_end:
12.1.3.6 Case-Sensitive Identifiers
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
153
Syntax
The Case-sensitive identifiers checkbox of the Assembler settings panel lets you control case-sensitivity for symbols:
• Check the checkbox to make symbols case sensitive -
SYM1
, sym1
, and
Sym1
are three different symbols.
• Clear the checkbox to make symbols not case-sensitive -
SYM1
, sym1
, and
Sym1
are the same symbol. (This is the default setting.)
12.1.4 Constants
The assembler recognizes three kinds of constants:
•
•
•
12.1.4.1 Integer Constants
Table 12-3 lists the notations for integer constants. Use the preferred notation for new
code. The alternate notations are for porting existing code.
Hexadecimal
Decimal
Binary
Type
Table 12-3. Preferred Integer Constant Notation
Preferred Notation
$ followed by string of hexadecimal digits, such as $deadbeef.
String of decimal digits, such as
12345678.
% followed by a string of binary digits, such as %01010001.
Alternate Notation
0x followed by a string of hexadecimal digits, such as 0xdeadbeef.
0 followed by a string of hexadecimal digits, ending with h, such as
0deadbeefh.
String of decimal digits followed by d, such as 12345678d.
String of binary digits followed by b, such as 01010001b.
NOTE
The assembler uses 32-bit signed arithmetic to store and manipulate integer constants.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
154 Freescale Semiconductor, Inc.
Chapter 12 Assembler
12.1.4.2 Floating-Point Constants
You can specify floating-point constants in either hexadecimal or decimal format. The decimal format must contain a decimal point or an exponent. Examples are
1E-10
and
1.0
.
You can use floating-point constants only in data generation directives such as
.float
and
.double
, or in floating-point instructions. You cannot such constants in expressions.
12.1.4.3 Character Constants
Enclose a character constant in single quotes. However, if the character constant includes a single quote, use double quotes to enclose the character constant.
NOTE
A character constant cannot include both single and double quotes.
\"
\\
\r
\t
\b
\n
\nnn
The maximum width of a character constant is 4 characters, depending on the context.
Examples are
'A'
,
'ABC'
, and
'TEXT'
.
A character constant can contain any of the escape sequences that
lists.
Table 12-4. Character Constant Escape Sequences
Sequence Description
Backspace
Line feed (ASCII character 10)
Return (ASCII character 13)
Tab
Double quote
Backslash
Octal value of \nnn
During computation, the assembler zero-extends a character constant to 32 bits. You can use a character constant anywhere you can use an integer constant.
12.1.5 Expressions
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
155
Syntax
The assembler uses 32-bit signed arithmetic to evaluates expressions; it does not check for arithmetic overflow.
As different processors use different operators, the assembler uses an expression syntax similar to that of the C language. Expressions use C operators and follow C rules for parentheses and associativity.
NOTE
To refer to the program counter in an expression, use a period
(.), dollar sign ($), or asterisk (*).
Table 12-5 lists the expression operators that the assembler supports.
Table 12-5. Expression Operators
Binary
Unary
Alternate
Category
~
<>
|
%
-
+
<
@
||
==
!=
<=
>=
>=
|
&
^
<<
>>
/
%
||
&&
-
+
*
Operator Description
add subtract multiply divide modulo logical OR logical AND bitwise OR bitwise AND bitwise XOR shift left shift right (zeros are shifted into high order bits) equal to not equal to less than or equal to greater than or equal to greater than or equal to less than at unary plus unary minus unary bitwise complement not equal to modulo logical OR logical XOR
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
156 Freescale Semiconductor, Inc.
Chapter 12 Assembler
Operator precedence is:
1. unary
+ - ~
2.
* / %
3. binary
+ -
4.
<< >>
5.
< <= > >=
6.
== !=
7.
&
8.
^
9.
| @
10.
&&
11.
||
12.1.6 Comments
There are several ways to specify comments:
1. Use either type of C-style comment, which can start in any column:
// This is a comment.
or
/* This is a comment. */
2. Start the comment with an asterisk (*) in the first column of the line.
NOTE
The asterisk (*) must be the first character of the line for it to specify a comment. The asterisk has other meanings if it occurs elsewhere in a line.
3. Clear the Allow space in operand field checkbox of the Assembler settings panel.
Subsequently, if you type a space in an operand field, all the remaining text of the line is a comment.
4. Anything following a # character is considered to be a comment. For example, st r3,0(r4) # Store total
5. Anything following a
; character is considered to be a comment, except in GNU compatibility mode, where
; is a statement separator.
12.1.7 Data Alignment
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
157
Directives
The assembler's default alignment is on a natural boundary for the data size and for the target processor family. To turn off this default alignment, use the alignment
keyword argument with to the
. option
directive.
NOTE
The assembler does not align data automatically in the
.debug
section.
12.2 Directives
• Some directives may not be available for the assembler for your target processor.
• The default starting character for most directives is the period (.). However, if you clear the Directives begin with '.' checkbox of the Assembler settings panel, you can omit the period.
• You can use the C/C++ preprocessor format to specify several preprocessor directives .
Explanations are in these sections:
•
•
Conditional Preprocessor Directives
•
•
•
•
•
•
12.2.1 Macro Directives
These directives let you create macros:
•
•
•
•
For more information on macros, see
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
158 Freescale Semiconductor, Inc.
Chapter 12 Assembler
12.2.1.1 macro
Starts the definition of a macro.
label .macro [ parameter ] [ ,parameter ] ...
Parameters
label
Name you give the macro.
parameter
Optional parameter for the macro.
12.2.1.2 endm
Ends the definition of a macro.
.endm
12.2.1.3 mexit
Stops macro execution before it reaches the .endm directive. Program execution continues with the statement that follows the macro call.
.mexit
12.2.1.4 #define
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
159
Directives
Defines a C pre-processor macro with the specified parameters. Note that the C preprocessor is run on the assembler file before normal assembly. C pre-processor macros should not be confused with normal macros declared using the
MACRO
and
ENDM
directives.
#define
name [ (
parms) ]
assembly_statement [ ; ] [ \ ]
assembly_statement [ ; ] [ \ ]
assembly_statement
parms ::=
parameter [ ,
parameter ]...
Parameters
name
Name you give the macro.
parms
List of parameters, separated by commas.
assembly_statement
Any valid assembly statement.
Remarks
To extend an assembly_statement, type a backslash (\) and continue the statement on the next line. To specify multiple assembly statements in the macro, type a semicolon and backslash (;\), then type a new assembly statement on the next line. If the assembler is in
GNU mode, multiple statements can be on one line of code - separate them with semicolon characters (;).
NOTE
For more information, see
"Using the #define Directive" on page 181
.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
160 Freescale Semiconductor, Inc.
Chapter 12 Assembler
12.2.2 Conditional Preprocessor Directives
Conditional directives let you control whether compilation includes a block of code.
These directives let you make multiple builds that are slightly different.
You must use conditional directives together to form a complete block. Several conditional directives are variations of
.if
that make it easier to establish blocks that test strings for equality, test whether a symbol is defined, and so on.
NOTE
You can use the C/C++ preprocessor format to specify these conditional directives:
#if #ifdef #ifndef
#else #elif #endif
With two exceptions, these directives function identically whether their starting character is a pound sign (#) or a period.
The exceptions are:
1. You cannot use the pound sign format in a macro.
2. The period form of
#elif
is
.elseif
.
The conditional preprocessor directives are:
•
•
•
•
•
•
•
•
•
Compatibility Conditional Directives
12.2.2.1 if
Starts a conditional assembly block, making assembly conditional on the truth of a boolean expression.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
161
Directives
.if bool-expr
Parameter
bool-expr
Any boolean expression.
Remarks
If bool-expr
is true, the assembler processes the statements of the block. If bool-expr
is false, the assembler skips the statements of the block.
Each
.if
directive must have a matching
.endif
directive.
12.2.2.2 ifdef
Starts a conditional assembly block, making assembly conditional on the definition of a symbol.
#ifdef symbol
Parameter
symbol
Any valid symbol.
Remarks
If previous code includes a definition for symbol
, the assembler processes the statements of the block. If symbol
is not defined, the assembler skips the statements of the block.
Each
.ifdef
directive must have a matching
.endif
directive.
12.2.2.3 ifndef
Starts a conditional assembly block, making assembly conditional on a symbol not being defined.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
162 Freescale Semiconductor, Inc.
Chapter 12 Assembler
.ifndef symbol
Parameter
symbol
Any valid symbol.
Remarks
If previous code does not include a definition for symbol
, the assembler processes the statements of the block. If there is a definition for symbol
, the assembler skips the statements of the block.
Each
.ifndef
directive must have a matching
.endif
directive.
12.2.2.4 ifc
Starts a conditional assembly block, making assembly conditional on the equality of two strings.
.ifc string1, string2
Parameters
string1
Any valid string.
string2
Any valid string.
Remarks
If string1 and string2 are equal, the assembler processes the statements of the block. (The equality comparison is case-sensitive.) If the strings are not equal, the assembler skips the statements of the block.
Each
.ifc
directive must have a matching
.endif
directive.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
163
Directives
12.2.2.5 ifnc
Starts a conditional assembly block, making assembly conditional on the inequality of two strings.
.ifnc string1, string2
Parameters
string1
Any valid string.
string2
Any valid string.
Remarks
If string1 and string2 are not equal, the assembler processes the statements of the block.
(The inequality comparison is case-sensitive.) If the strings are equal, the assembler skips the statements of the block.
Each
.ifnc
directive must have a matching
.endif
directive.
12.2.2.6 endif
Ends a conditional assembly block. A matching
.endif
directive is mandatory for each type of
.if
directive.
.endif
12.2.2.7 elseif
Starts an alternative conditional assembly block, making assembly conditional on the truth of a boolean expression.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
164 Freescale Semiconductor, Inc.
Chapter 12 Assembler
.elseif bool-expr
Parameter
bool-expr
Any boolean expression.
Remarks
If bool-expr
is true, the assembler processes the statements of the block. If bool-expr
is false, the assembler skips the statements of the block.
You can use this directive to create a logical, multilevel if-then-else statement, according to this syntax:
.if bool-expr statement-group
[ .elseif bool-expr statement-group ]...
[ .else statement-group ]
.endif
(In this syntax, statement-group
is any group of assembly-language statements.)
The
.elseif
directive can be part of more complicated logical structures, such as:
.if bool-expr-1
statement-group-1
.elseif bool-expr-2
statement-group-2
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
165
Directives
.elseif bool-expr-3
statement-group-3
.elseif bool-expr-4
statement-group-4
.else
statement-group-5
.endif
• If this structure's bool-expr-1
is true, the assembler executes the statement-group-1 assembly-language statements, then goes to the
.endif
directive.
• If bool-expr-1
is false, the assembler skips statement-group-1
, executing the first
.elseif
directive. If bool-expr-2
is true, the assembler executes statement-group-2
, then goes to the
.endif
directive.
• If bool-expr-2
also is false, the assembler skips statement-group-2
, executing the second
.elseif
directive.
• The assembler continues evaluating the boolean expressions of succeeding
.elseif
directives until it comes to a boolean expression that is true.
• If none of the boolean expressions are true, the assembler processes statement-group-5
, because this structure includes an
.else
directive. (If none of the boolean values were true and there were no
.else
directive, the assembler would not process any of the statement groups.)
12.2.2.8 else
Starts an alternative conditional assembly block.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
166 Freescale Semiconductor, Inc.
Chapter 12 Assembler
.else
Remarks
This directive is optional. The assembler processes the statements of the alternative conditional assembly block only if the expressions for an
.if
directive and any associated
.elseif
directives are false.
12.2.3 Compatibility Conditional Directives
For compatibility with other assemblers, the assembler supports these additional conditional directives:
•
•
•
•
•
•
.ifge if greater than or equal
12.2.3.1 .ifeq if equal
Starts a conditional assembly block, making assembly conditional on a string value being equal to zero.
.ifeq string
Parameter
string
Any valid string.
Remarks
If the string
value equals 0, the assembler processes the statements of the block. If the string
value does not equal 0, the assembler skips the statements of the block.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
167
Directives
12.2.3.2 .ifne if not equal
Starts a conditional assembly block, making assembly conditional on a string value not being equal to zero.
.ifne string
Parameter
string
Any valid string.
Remarks
If the string value is not equal to 0, the assembler processes the statements of the block. If the string
value does equal 0, the assembler skips the statements of the block.
12.2.3.3 .iflt if less than
Starts a conditional assembly block, making assembly conditional on a string value being less than zero.
.iflt string
Parameter
string
Any valid string.
Remarks
If the string
value is less than 0, the assembler processes the statements of the block. If the string
value equals or exceeds 0, the assembler skips the statements of the block.
12.2.3.4 .ifle if less than or equal
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
168 Freescale Semiconductor, Inc.
Chapter 12 Assembler
Starts a conditional assembly block, making assembly conditional on a string value being less than or equal to zero.
.ifle string
Parameter
string
Any valid string.
Remarks
If the string value is less than or equal to 0, the assembler processes the statements of the block. If the string
value is greater than 0, the assembler skips the statements of the block.
12.2.3.5 .ifgt if greater than
Starts a conditional assembly block, making assembly conditional on a string value being greater than zero.
.ifgt string
Parameter
string
Any valid string.
Remarks
If the string
value is greater than 0, the assembler processes the statements of the block. If the string value is less than or equal to 0, the assembler skips the statements of the block.
12.2.3.6 .ifge if greater than or equal
Starts a conditional assembly block, making assembly conditional on a the string value being greater than or equal to zero.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
169
Directives
.ifge string
Parameter
string
Any valid string.
Remarks
If the string
value is greater than or equal to 0, the assembler processes the statements of the block. If the string
value is less than 0, the assembler skips the statements of the block.
12.2.4 Section Control Directives
These directives identify the different sections of an assembly file:
•
•
•
•
•
•
•
•
•
•
•
•
12.2.4.1 text
Specifies an executable code section; must be in front of the actual code in a file.
.text
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
170 Freescale Semiconductor, Inc.
Chapter 12 Assembler
12.2.4.2 data
Specifies an initialized read-write data section.
.data
12.2.4.3 rodata
Specifies an initialized read-only data section.
.rodata
12.2.4.4 bss
Specifies an uninitialized read-write data section.
.bss
12.2.4.5 sdata
Specifies a small data section as initialized and read-write.
.sdata
12.2.4.6 sdata2
Specifies a small data section as initialized and read-only.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
171
Directives
.sdata2
12.2.4.7 sbss
Specifies a small data section as uninitialized and read-write.
.sbss
12.2.4.8 text_vle
Specifies a Variable length encoded section as read/execute.
.text_vle
12.2.4.9 debug
Specifies a debug section.
.debug
Remarks
If you enable the debugger, the assembler automatically generates some debug information for your project. However, you can use special directives in the debug section that provide the debugger with more detailed information. For more information on the debug directives, see
12.2.4.10 previous
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
172 Freescale Semiconductor, Inc.
Chapter 12 Assembler
Reverts to the previous section; toggles between the current section and the previous section.
.previous
12.2.4.11 offset
Starts a record definition, which extends to the start of the next section.
.offset [expression]
Parameter
expression
Optional initial location-counter value.
Remarks
Table 12-6 lists the only directives your can use inside a record.
Table 12-6. Directives within a Record
.align
.ascii
.asciz
.byte
.double
.equ
.float
.long
.org
.set
.short
.space
.textequ
Data declaration directives such as
.byte
and
.short
update the location counter, but do not allocate any storage.
Example
Figure 12-1 shows a sample record definition.
Figure 12-1. Record Definition with Offset Directive
.offset
top: .short 0 left: .short 0 bottom: .short 0 right: .short 0
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
173
Directives
rectSize .equ *
12.2.4.12 section
Defines a section of an ELF (Executable and Linkable Format) object file.
.section name [,alignment [,type [,flags]]]
Parameters
name
Name of the section.
alignment
Alignment boundary.
type
Numeric value for the ELF section type, per Table 12-7 . The default
type
value is
1
:
(SHT_PROGBITS).
flags
0
1
5
6
2
3
4
Numeric value for the ELF section flags, per
. The default flags
value is
0x00000002, 0x00000001
: (SHF_ALLOC+SHF_WRITE).
7
Type
Table 12-7. ELF Section Header Types (SHT)
NULL
PROGBITS
SYMTAB
STRTAB
RELA
HASH
DYNAMIC
NOTE
Name Meaning
Section header is inactive.
Section contains information that the program defines.
Section contains a symbol table.
Section contains a string table.
Section contains relocation entries with explicit addends.
Section contains a symbol hash table.
Section contains information used for dynamic linking.
Section contains information that marks the file, often for compatibility purposes between programs.
Table continues on the next page...
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
174 Freescale Semiconductor, Inc.
8
9
10
11
0x00000001
0x00000002
0x00000004
0xF0000000
Type
Chapter 12 Assembler
Table 12-7. ELF Section Header Types (SHT) (continued)
NOBITS
REL
SHLIB
DYNSYM
Name Meaning
Section occupies no space in the object file.
Section contains relocation entries without explicit addends.
Section has unspecified semantics, so does not conform to the Application
Binary Interface (ABI) standard.
Section contains a minimal set of symbols for dynamic linking.
Flag
Table 12-8. ELF Section Header Flags (SHF)
WRITE
ALLOC
EXECINSTR
MASKPROC
Name Meaning
Section contains data that is writable during execution.
Section occupies memory during execution.
Section contains executable machine instructions.
Bits this mask specifies are reserved for processor-specific purposes.
Remark
Use this directive to create arbitrary relocatable sections, including sections to be loaded at an absolute address.
Possible syntax forms
The section directive accepts a number of different syntax forms, partly for convenience and partly for compatibility with other assemblers. A section declaration requires four pieces of information: a section name, alignment, ELF section type (for example,
SHT_PROGBITS) and ELF section flags (for example, SHF_ALLOC
+SHF_EXECINSTR).
The possible syntax forms are as follows:
• Specify built-in section name.
.section text
This example specifies a built-in section name text
. Equivalently, .text is also a valid syntax form.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
175
Directives
Table 12-9 provides a list of all the possible values, together with their ELF types
and ELF Section Header Flags.
.text
.data
.rodata
.bss
.sdata
.sdata0
.sdata2
.sbss
.sbss0
.sbss2
.debug
.text_vle
Name
Table 12-9. Built-in Section names with their ELF Types and
Flags
.PPC.EMB.sdata0
.PPC.EMB.sbss0
ELF Type
SHT_PROGBITS
SHT_PROGBITS
SHT_PROGBITS
SHT_NOBITS
SHT_PROGBITS
SHT_PROGBITS
SHT_PROGBITS
SHT_NOBITS
SHT_NOBITS
SHT_PROGBITS
SHT_PROGBITS
SHT_PROGBITS
SHT_PROGBITS
SHT_PROGBITS
ELF Flag
SHF_ALLOC
+SHF_EXECINSTR
SHF_ALLOC+SHF_WRITE
SHF_ALLOC
SHF_ALLOC+SHF_WRITE
SHF_ALLOC+SHF_WRITE
SHF_ALLOC+SHF_WRITE
SHF_ALLOC
SHF_ALLOC+SHF_WRITE
SHF_ALLOC+SHF_WRITE
SHF_ALLOC
0
SHF_ALLOC
+SHF_EXECINSTR+
SHF_PE_EXECINSTR
SHF_ALLOC+SHF_WRITE
SHF_ALLOC+SHF_WRITE
In general, .text is for instructions, .data for initialised data, .rodata for read-only data
(constants) and .bss for uninitialised data. The additional forms like .sdata are for small data areas. The built-in section names are architecture-specific, and are intended to make access to data more efficient. The alignment used for these sections is architecture-specific and is usually 4.
• Specify your own section name but get the attributes of the built-in section type.
.section mySection,text
This example is equivalent to writing .text except that the section will be called mySection.
• Specify the alignment, and optionally the ELF type and flags.
.section name [,alignment [,type [,flags]]]
In the syntax above, if the alignment is not specified it defaults to 16. If the type or flags are not specified, the defaults are as follows:
• If the name parameter is a built-in section name, the type and the flags are taken
as specified in the Table 12-9 . For example, in the syntax form.section text,8 the
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
176 Freescale Semiconductor, Inc.
Chapter 12 Assembler
type is SHT_PROGBITS and the flags value is SHF_ALLOC
+SHF_EXECINSTR.
• In all other cases, the default type is SHT_PROGBITS and the default flags value is SHF_ALLOC+SHF_WRITE, corresponding to a writeable data section.
• Specify the type and flags parameters in pre-defined characters, optionally in double quotes.
.section mySection,4,"rx" or .section mySection,4,rx
The values are additive. For example, rx is equivalent to SHF_ALLOC
+SHF_WRITE+SHF_EXECINSTR
NOTE
If the syntax doesn't specify a type it defaults to
SHT_PROGBITS
provides a list of all the possible characters and their corresponding ELF
Type and ELF Flags.
r w x d m b c
Table 12-10. Characters and their corresponding ELF Type and ELF
Flags
Character
0
0
0
ELF Type
SHT_NOBITS
SHT_PROGBITS
SHT_PROGBITS
SHT_PROGBITS
ELF Flag
SHF_ALLOC+SHF_WRITE
SHF_ALLOC
+SHF_EXECINSTR
SHF_ALLOC+SHF_WRITE
SHF_ALLOC+SHF_WRITE+
SHF_EXECINSTR
SHF_ALLOC
SHF_ALLOC+SHF_WRITE
SHF_ALLOC
+SHF_EXECINSTR
12.2.5 Scope Control Directives
These directives let you import and export labels:
•
•
•
For more information on labels, see "Labels".
.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
177
Directives
NOTE
You cannot import or export equates or local labels.
12.2.5.1 global
Tells the assembler to export the specified labels, that is, make them available to other files.
.global label [ ,label ]...
Parameter
label
Any valid label.
12.2.5.2 extern
Tells the assembler to import the specified labels, that is, find the definitions in another file.
.extern label [ ,label ]...
Parameter
label
Any valid label.
12.2.5.3 public
Declares specified labels to be public.
.public label [ ,label ]...
Parameter
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
178 Freescale Semiconductor, Inc.
Chapter 12 Assembler
label
Any valid label.
Remarks
If the labels already are defined in the same file, the assembler exports them (makes them available to other files). If the labels are not already defined, the assembler imports them
(finds their definitions in another file).
12.2.6 Symbol Definition Directives
These directives let you create equates:
•
•
•
•
12.2.6.1 set
Defines an equate, assigning an initial value. You can change this value at a later time.
equate .set expression
Parameters
equate
Name of the equate.
expression
Temporary initial value for the equate.
12.2.6.2 equal sign (=)
Defines an equate, assigning an initial value. You can change this value at a later time.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
179
Directives
equate = expression
Parameters
equate
Name of the equate.
expression
Temporary initial value for the equate.
Remarks
This directive is equivalent to
.set
. It is available only for compatibility with assemblers provided by other companies.
12.2.6.3 equ
Defines an equate, assigning a permanent value. You cannot change this value at a later time.
equate .equ expression
Parameters
equate
Name of the equate.
expression
Permanent value for the equate.
12.2.6.4 textequ
Defines a text equate, assigning a string value.
equate .textequ "string"
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
180 Freescale Semiconductor, Inc.
Chapter 12 Assembler
Parameters
equate
Name of the equate.
string
String value for the equate, in double quotes.
Remarks
This directive helps port existing code. You can use it to give new names to machine instructions, directives, and operands.
Upon finding a text equate, the assembler replaces it with the string value before performing any other processing on that source line.
Examples
dc.b .textequ ".byte"
endc .textequ ".endif"
12.2.7 Data Declaration Directives
These directive types initialize data:
•
•
•
12.2.7.1 Integer Directives
These directives let you initialize blocks of integer data:
•
•
•
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
181
Directives
•
•
•
12.2.7.1.1 byte
Declares an initialized block of bytes.
[ label ] .byte expression [ ,expression ]...
Parameters
label
Name of the block of bytes.
expression
Value for one byte of the block; must fit into one byte.
12.2.7.1.2 fill
Declares a block of bytes, initializing each byte to zero.
[ label ] .fill expression
Parameters
label
Name of the block of bytes.
expression
Number of bytes in the block.
12.2.7.1.3 incbin
Tells the assembler to include the binary content of the specified file.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
182 Freescale Semiconductor, Inc.
Chapter 12 Assembler
. incbin filename[,start[,length]]
Parameters
filename
Name of a binary file.
start
Offset from start of file at which to begin including data.
length
Number of bytes of data to include.
Remarks
This directive is useful when you wish to include arbitrary data directly in the program being assembled, e.g.
logoPicture: .incbin "CompanyLogo.jpg"
12.2.7.1.4 long
Declares an initialized block of 32-bit short integers.
[ label ] .long expression [ ,expression ]...
Parameters
label
Name of the block of integers.
expression
Value for 32 bits of the block; must fit into 32 bits.
12.2.7.1.5 short
Declares an initialized block of 16-bit short integers.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
183
Directives
[ label ] .short expression [ ,expression ]...
Parameters
label
Name of the block of integers.
expression
Value for 16 bits of the block; must fit into 16 bits.
12.2.7.1.6 space
Declares a block of bytes, initializing each byte to zero.
[ label ] .space expression
Parameters
label
Name of the block of bytes.
expression
Number of bytes in the block.
12.2.7.2 String Directives
These directives initialize blocks of character data:
•
•
A string can contain any of the escape sequences
\b
\n
Sequence
Table 12-11. Escape sequences
Description
Backspace
Line feed (ASCII character 10)
Table continues on the next page...
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
184 Freescale Semiconductor, Inc.
\r
\t
\"
\\
\nnn
Table 12-11. Escape sequences (continued)
Sequence Description
Return (ASCII character 13)
Tab
Double quote
Backslash
Octal value of \nnn
Chapter 12 Assembler
12.2.7.2.1 ascii
Declares a block of storage for a string; the assembler allocates a byte for each character.
[ label ] .ascii "string"
Parameters
label
Name of the storage block.
string
String value to be stored, in double quotes.
12.2.7.2.2 asciz
Declares a zero-terminated block of storage for a string.
[ label ] .asciz "string"
Parameters
label
Name of the storage block.
string
String value to be stored, in double quotes.
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
185
Directives
The assembler allocates a byte for each string
character. The assembler then allocates an extra byte at the end, initializing this extra byte to zero.
12.2.7.3 Floating-Point Directives
These directives initialize blocks of floating-point data:
•
•
12.2.7.3.1 float
Declares an initialized block of 32-bit, floating-point numbers; the assembler allocates 32 bits for each value.
[ label ] .float value [ ,value ]...
Parameters
label
Name of the storage block.
value
Floating-point value; must fit into 32 bits.
12.2.7.3.2 double
Declares an initialized block of 64-bit, floating-point numbers; the assembler allocates 64 bits for each value.
[ label ] .double value [ ,value ]...
Parameters
label
Name of the storage block.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
186 Freescale Semiconductor, Inc.
Chapter 12 Assembler
value
Floating-point value; must fit into 64 bits.
12.2.8 Assembler Control Directives
These directives let you control code assembly:
•
•
•
•
•
•
•
12.2.8.1 align
Aligns the location counter to the next multiple of the specified power of 2.
.align expression
Parameter
expression
Power of 2, such as 2, 4, 8, 16, or 32.
12.2.8.2 endian
Specifies byte ordering for the target processor; valid only for processors that permit change of endianness.
.endian big | little
Parameters
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
187
Directives
big
Big-endian specifier.
little
Little-endian specifier.
12.2.8.3 error
Prints the specified error message to the IDE Errors and Warnings window.
.error "error"
Parameter
error
Error message, in double quotes.
12.2.8.4 include
Tells the assembler to take input from the specified file.
.include filename
Parameter
filename
Name of an input file.
Remarks
When the assembler reaches the end of the specified file, it takes input from the assembly statement line that follows the
.include
directive. The specified file can itself contain an
.include
directive that specifies yet another input file.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
188 Freescale Semiconductor, Inc.
Chapter 12 Assembler
12.2.8.5 pragma
Tells the assembler to use a particular pragma setting as it assembles code.
.pragma pragma-type setting
Parameters
pragma-type
Type of pragma.
setting
Setting value.
12.2.8.6 org
Changes the location-counter value, relative to the base of the current section.
.org expression
Parameter
expression
New value for the location counter; must be greater than the current location-counter value.
Remarks
Addresses of subsequent assembly statements begin at the new expression value for the location counter, but this value is relative to the base of the current section.
Example
In
Listing 12-1 on page 189 , the label
Alpha
reflects the value of
.text + 0x1000
. If the linker places the
.text
section at
0x10000000
, the runtime
Alpha
value is
0x10001000
.
Listing 12-1. Address-Change Example
.text
.org 0x1000
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
189
Directives
Alpha:
...
blr
NOTE
You must use the linker to place code at an absolute address.
12.2.8.7 option
Sets an assembler control option as
describes.
.option keyword setting
Parameters
keyword
Control option.
setting
Setting value appropriate for the option: OFF, ON, RESET, or a particular number value.
RESET returns the option to its previous setting.
alignment off | on | reset case off | on | reset colon off | on | reset period off | on | reset space off | on | reset
Keyword
Table 12-12. Option Keywords
Description
Controls data alignment on a natural boundary.
Specifies case sensitivity for identifiers.
Specifies whether labels must end with a colon (:). The OFF setting means that you can omit the ending colon from label names that start in the first column.
Controls period usage for directives. The ON setting means that each directive must start with a period.
Controls spaces in operand fields. The OFF setting means that a space in an operand field starts a comment.
12.2.9 Debugging Directives
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
190 Freescale Semiconductor, Inc.
Chapter 12 Assembler
When you enable the debugger, the assembler automatically generates some debug information for your project. However, you can use these directives in the debug section to provide additional information to the debugger:
•
•
•
•
•
NOTE
These debugging directives are valid only in the
.debug
and
.text
sections of an assembly file. Additionally, you must enable debugging for the file that contains the debugging directives; you use the Project window to enable this debugging.
12.2.9.1 file
Specifies the source-code file; enables correlation of generated assembly code and source code.
.file "filename"
Parameter
filename
Name of source-code file, in double quotes.
Remarks
Must precede other debugging directives in the assembly file. If you write your own
DWARF code, you must use the
.function
and
.line
directives as well as this
.file
directive.
Example
shows how to use the
.file
directive for your own DWARF code.
Listing 12-2. DWARF Code Example
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
191
Directives
.file "MyFile.c"
.text
.globl _MyFunction
.function "MyFunction",_MyFunction, _MyFunctionEnd -_MyFunction
_MyFunction:
.line 1
lwz r3, 0(r3)
.line 2
blr
_MyFunctionEnd:
12.2.9.2 function
Tells the assembler to generate debugging data for the specified subroutine.
.function "func", label, length
Parameters
func
Subroutine name, in double quotes.
label
Starting label of the subroutine.
length
Number of bytes in the subroutine.
12.2.9.3 line
Specifies the absolute line number (of the current source file) for which the assembler generates subsequent code or data.
.line number
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
192 Freescale Semiconductor, Inc.
Chapter 12 Assembler
Parameter
number
Line number of the file; the file's first line is number 1.
12.2.9.4 size
Specifies a length for a symbol.
.size symbol, expression
Parameters
symbol
Symbol name.
expression
Number of bytes.
12.2.9.5 type
Specifies the type of a symbol.
.type symbol, @function | @object
Parameters
symbol
Symbol name.
@function
Function type specifier.
@object
Variable specifier.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
193
Macros
12.3 Macros
This chapter explains how to define and use macros. You can use the same macro language regardless of your target processor.
This chapter includes these topics:
•
•
12.3.1 Defining Macros
This section explains how to define macros.
•
•
•
Creating Unique Labels and Equates
•
Referring to the Number of Arguments
12.3.1.1 Macro Definition Syntax
A macro definition is one or more assembly statements that define:
• the name of a macro
• the format of the macro call
• the assembly statements of the macro
To define a macro, use either the
,macro
or the
#define
directive.
NOTE
If you use a local label in a macro, the scope of the label is limited to the expansion of the macro. (Local labels begin with the @ character.)
12.3.1.1.1 Using the .macro Directive
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
194 Freescale Semiconductor, Inc.
Chapter 12 Assembler
The
.macro
directive is part of the first line of a macro definition. Every macro definition ends with the
.endm
directive . Listing 12-3 on page 195
shows the full syntax, and
12-13 explains the syntax elements.
Listing 12-3. Macro Definition Syntax: .macro Directive
name: .macro [
parameter ] [ ,
parameter ] ...
macro_body
.endm
name parameter macro_body
Table 12-13. Syntax Elements: .macro Directive
Element Description
Label that invokes the macro.
Operand the assembler passes to the macro for us in the macro body.
One or more assembly language statements. Invoking the macro tell the assembler to substitutes these statements.
The body of a simple macro consists of just one or two statements for the assembler to execute. Then, in response to the
.endm
directive, the assembler resumes program execution at the statement immediately after the macro call.
But not all macros are so simple. For example, a macro can contain a conditional assembly block, The conditional test could lead to the
.mexit
directive stopping execution early, before it reaches the
.endm
directive.
is the definition of macro addto
, which includes an
.mexit
directive.
shows the assembly-language code that calls the addto
macro.
Listing 12-6 on page 196 shows the expanded
addto
macro calls.
Listing 12-4. Conditional Macro Definition
//define a macro addto: .macro dest,val
.if val==0
nop
.elseif val >= -32768 && val <= 32767
addi dest,dest,val // use compact instruction
.else
addi dest,dest,val@l // use 32-bit add
addis dest,dest,val@ha
.endif
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
195
Macros
// end macro definition
.endm
Listing 12-5. Assembly Code that Calls addto Macro
// specify an executable code section
.text
li r3,0
// call the addto macro addto r3,0 addto r3,1 addto r3,2 addto r3,0x12345678
Listing 12-6. Expanded addto Macro Calls
li r3,0 nop addi r3,r3,1 addi r3,r3,2 addi r3,r3,0x12345678@l addis r3,r3,0x12345678@ha
12.3.1.2 Using Macro Arguments
You can refer to parameters directly by name. Listing 12-7 on page 196
shows the setup macro, which moves an integer into a register and branches to the label
_final_setup
.
shows a way to invoke the setup
macro., and Listing 12-9 on page 197
shows how the assembler expands the setup
macro.
Listing 12-7. Setup Macro Definition
setup: .macro name
li r3,name
bl _final_setup
.endm
Listing 12-8. Calling Setup Macro
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
196 Freescale Semiconductor, Inc.
Chapter 12 Assembler
VECT: .equ 0
setup VECT
Listing 12-9. Expanding Setup Macro
li r3,VECT
bl _final_setup
If you refer to named macro parameters in the macro body, you can precede or follow the macro parameter with
&&
. This lets you embed the parameter in a string. For example,
Listing 12-10 on page 197 shows the
smallnum
macro, which creates a small float by appending the string
E-20
to the macro argument. Listing 12-11 on page 197
shows a way to invoke the smallnum
macro, and
shows how the assembler expands the smallnum
macro.
Listing 12-10. Smallnum Macro Definition
smallnum: .macro mantissa
.float mantissa&&E-20
endm
Listing 12-11. Invoking Smallnum Macro
smallnum 10
Listing 12-12. Expanding Smallnum Macro
.float 10E-20
12.3.1.3 Creating Unique Labels and Equates
Use the backslash and at characters (\@) to have the assembler generate unique labels and equates within a macro. Each time you invoke the macro, the assembler generates a unique symbol of the form
??
nnnn
, such as
??0001
or
??0002
.
In your code, you refer to such unique labels and equates just as you do for regular labels and equates. But each time you invoke the macro, the assembler replaces the \@ sequence with a unique numeric string and increments the string value.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
197
Macros
Listing 12-13 on page 198 shows a macro that uses unique labels and equates. Listing
12-14 on page 198 shows two calls to the
putstr
macro.
shows the expanded code after the two calls.
Listing 12-13. Unique Label Macro Definition
putstr: .macro string
lis r3,(str\@)@h
oris r3,r3,(str\@)@l
bl put_string
b skip\@ str\@: .asciz string
.align 4 skip\@:
.endm
Listing 12-14. Invoking putstr Macro
putstr 'SuperSoft Version 1.3' putstr 'Initializing...'
Listing 12-15. Expanding putstr Calls
lis r3,(str??0000)@h
oris r3,r3,(str??0000)@l
bl put_string
b skip??0000
str??0000: .asciz 'SuperSoft Version
.align 4 skip??0000:
lis r3,(str??0001)@h
oris r3,r3,(str??0001)@l
bl put_string
b skip??0001
str??0001: .asciz 'Initializing...'
.align 4 skip??0001:
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
198 Freescale Semiconductor, Inc.
Chapter 12 Assembler
12.3.1.4 Referring to the Number of Arguments
To refer to the number of non-null arguments passed to a macro, use the special symbol narg
. You can use this symbol during macro expansion.
12.3.2 Invoking Macros
To invoke a macro, use its name in your assembler listing, separating parameters with commas. To pass a parameter that includes a comma, enclose the parameter in angle brackets.
For example, Listing 12-16 on page 199 shows macro
pattern
, which repeats a pattern of
bytes passed to it the number of times specified in the macro call. Listing 12-17 on page
199 shows a statement that calls
pattern
, passing a parameter that includes a comma.
Listing 12-16. Pattern Macro Definition
pattern: .macro times,bytes
.rept times
.byte bytes
.endr
.endm
Listing 12-17. Macro Argument with Commas
.data
halfgrey: pattern 4,<0xAA,0x55>
Listing 12-18. Alternate Byte-Pattern Method
halfgrey: .byte 0xAA,0x55,0xAA,0x55,0xAA,0x55,0xAA,0x55
12.3.2.1 Using the #define Directive
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
199
GNU Compatibility
Another way to define a macro is to use the #define directive. This will define a preprocessor style macro using a syntax that will be familiar to C programmers. Note that C pre-processor macros are complementary to the assembler's main
MACRO...ENDM
macro language.
Listing 12-19 on page 200 shows the full syntax, and explains the syntax elements.
Listing 12-19. Macro Definition Syntax: #define Directive
#define
name [ (
parms) ]
assembly_statement [ ; ] [ \ ]
assembly_statement [ ; ] [ \ ]
assembly_statement
parms ::=
parameter [ ,
parameter ]...
NOTE
If you specify parameters for a macro, you must enclose them in parentheses.
name parameter assembly_statement
Table 12-14. Syntax Elements: #define Directive
Element Description
Label that invokes the macro.
Operand the assembler passes to the macro.
An assembly language statement. To extend the statement beyond the length of one physical line, type a backslash (\) at the end of a line, then continue the statement on the next line.
To specify multiple statements on the same line, separate then with semicolon and backslash characters (;\).
12.4 GNU Compatibility
The CodeWarrior Assembler supports several GNU-format assembly language extensions.
•
•
•
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
200 Freescale Semiconductor, Inc.
Chapter 12 Assembler
12.4.1 GNU Compatible Syntax option
Only in cases where GNU's assembler format conflicts with that of the CodeWarrior assembler does the GNU Compatible Syntax option have any effect. Specifically:
• Defining Equates
Whether defined using .equ or .set, all equates can be re-defined.
• Ignored directives
The .type directive ignored.
• Undefined Symbols
Undefined symbols are automatically treated as imported
• Arithmetic Operators
< and > mean left-shift and right-shift instead of less than and greater than.
! means bitwise-or-not instead of logical not.
• Precedence Rules
Precedence rules for operators are changed to be compatible with GNU rather than with C.
• Local Labels
Local labels with multi-number characters are supported (example: "1000:"). There is no limit on the number of digits in the label name. Multiple instances of the label are allowed. When referenced, you get the nearest one - forwards or backwards depending on whether you append 'f' or 'b' to the number.
• Numeric Constants
Numeric constants beginning with 0 are treated as octal.
• Semicolon Use
Semicolons can be used as a statement separator.
• Unbalanced Quotes
A single unbalanced quote can be used for character constants. For example: .byte 'a
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
201
GNU Compatibility
12.4.2 Supported Extensions
Some GNU extensions are always available, regardless whether you enable GNU
compatible syntax . Specifically:
• Lines beginning with # * or ; are always treated as comment, even if the comment symbol for that assembler is someting different.
• Escape characters in strings extended to include
\xNN
for hex digits and
\NNN
for octal.
• Binary constants may begin with
0b
.
• Supports the GNU macro language, with macros defined by:
.macro name,arg1[=default1],arg2...s1
...
.endm
Arguments may have default values as shown, and when called may be specified by value or position. See the GNU documentation for details.
• New or enhanced directives (see GNU documentation for details)
Table 12-15. Supported GNU Assembler
Directives
Directive
.abort
.align N,[pad]
Description
End assembly
Align
.app-file name
.balign[wl] N,[pad]
.comm name,length
.def
.desc
.dim
.eject
.endr
.endef
.fill N,[size],[val]
.hword val..
.ident
.ifnotdef name
.include name
Source name
Align
Common data
Debugging
Debugging
Debugging
Eject page
End repeat
Debugging
Repeat data
Half-word
Tags
Conditional
Include file
Comment
Supported
Now accepts optional padding byte
Synonym for .file
Align to N (with optional padding value)
Reserve space in BSS for global symbol
Accepted but ignored
Accepted but ignored
Accepted but ignored
Accepted but ignored
See .irp, .irpc
Accepted but ignored
Emit N copies of width 'size', value 'val'
Synonym for .short
Accepted but ignored
Synonym for .ifndef
Now accepts single, double or no quotes
Table continues on the next page...
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
202 Freescale Semiconductor, Inc.
Directive
.int val..
.irp name,values
.irpc name,chars
.lcomm name,length
.lflags
.ln lineno
.list
.local name
.macro name, args..
.nolist
.org pos,fill
.p2align[wl] N[,pad]
.psize
.rept N
.sbttl
.scl
.size name,N
.skip N[,pad]
.space N[,pad]
.stabd
.stabn
.stabs
.str "string"
.string "string"
.tag
.title
.type
.val
.word
Chapter 12 Assembler
Table 12-15. Supported GNU Assembler Directives
(continued)
Word
Repeat
Description
Repeat
Local common
Ignored
Line number
Listing on
Local macro var
Macros
Listing off
Origin
Align
Page size
Repeat
Subtitle
Debugging
Set size
Space
Space
Debugging
Debugging
Debugging
Constant string
Constant string
Debugging
Title
Debugging
Debugging
Word
Comment
Synonm for .long
Repeat up to .endr substituting values for name
Repeat up to .endr substituting chars for name
Reserve length bytes in bss
Accepted but ignored
Synonym for .line
Switch on listing
Declare name as local to macro
Supports Gnu syntax, default values, etc
Disable listing
Now allows fill value ot be specified
Align to 2**N, using pad value
'pad'
Accepted but ignored
Repeat block up to .endr N times
Accepted but ignored
Accepted but ignored
Set size of name to N
Skip N bytes, pad with 'pad'
Skip N bytes, pad with 'pad'
Accepted but ignored
Accepted but ignored
Accepted but ignored
Synonym for .asciz
Synonym for .asciz
Accepted but ignored
Accepted but ignored
Ignored in Gnu mode
Accepted but ignored
Synonym for .long
12.4.3 Unsupported Extensions
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
203
GNU Compatibility
Among the GNU extensions that the CodeWarrior Assembler does not support are:
• Sub-sections (such as "
.text 2
"). The sub-section number will be ignored.
As a workaround, you can create your own sections with the
.section <name>
directive.
You may have an arbitrary number of text subsections with the names
.text1
,
.text2
, etc.
• Assignment to location counter (such as "
. = .+4
")
As a workaround, you can advance the location counter with
.space <expr>
• Empty expressions defaulting to 0. Example: ". byte ,
" equivalent to "
.byte 0,0
")
There is no workaround for this. You must always supply the arguments.
•
.linkonce
directive
The linker automatically detects logically-identical sections, and uses the following factors to determine whether to keep only one or both in the final image:
• the binding of the symbols associated with each section
• the location of these two sections. For example, are the sections in the same
•
.octa
overlay or overlay group? Is one in main, and the other in an overlay group?
We do not support 16-byte numbers directly. As a workaround, you may use consecutive
.long
directives to build a large number in memory.
•
.quad
We do not support eight-byte numbers directly. As a workaround, you may use consecutive
.long
directives to build a large number in memory.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
204 Freescale Semiconductor, Inc.
Chapter 13
Linker for AIOP
The compiler organizes its object code into sections that the linker arranges when it creates its output file.
To generate an output file, the linker reads from input ELF (Executable and Linkable
Format) files generated by compiler and other tools. The linker also reads a linker command file to determine how to build its output file. The linker then writes to its output file, an ELF file. This output file is the executable image, ready to load and run on the target platform.
This chapter explains the sections in the object code of and how to arrange them in the linker's output file:
•
Specifying Link Order in the IDE
•
•
Defining the Target's Memory Map
•
Defining Sections in the Output File
•
Associating Input Sections With Output Sections
•
•
Specifying Memory Area Locations and Sizes
•
•
•
•
Commands, Directives, and Keywords
13.1 Specifying Link Order in the IDE
To specify link order, use the Link Order page of the CodeWarrior IDE's Project window.
(For certain targets, the name of this page is Segments .)
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
205
Dead-Stripping
Regardless of the order that the Link Order page specifies, the The linker always processes source code files before it processes relocatable (
.o
) files or archive (
.a
) files.
This policy means that the linker prefers using a symbol definition from a source file rather than a library file definition for the same symbol.
There is an exception, however: if the source file defines a weak symbol, the linker uses a global-symbol definition from a library. Use
#pragma overload
to create weak symbols.
Well-constructed projects usually do not have strong link-order dependencies.
The linker ignores executable files of the project. You may find it convenient to keep the executable files in the project folder so that you can disassemble it. If a build is successful, a check mark disappears in the touch column on the left side of the project window. The check mark indicates that the new file in the project is out of date. If a build is unsuccessful, the IDE will not be able to find the executable file and it stops the build with an appropriate message.
13.2 Dead-Stripping
Normally, the CodeWarrior linker ignores object code that is not referred to by other object code. If the linker detects that an object is not referred to by the rest of the program being linked, the linker will not place that object in its output file. In other words, the linker "dead-strips" objects that are not used.
Dead-stripping ensures the smallest possible output file. Also, dead-stripping relieves you from having to manually exclude unused source code from the compiler and unused object code from the linker.
There are some objects, however, that need to be in the linker's output file even if these objects are not explicitly referred to by other parts of your program. For example, an executable image might contain an interrupt table that the target platform needs, but this interrupt table is not referred to by the rest of the image.
Use the
FORCEACTIVE
directive in a linker command file to specify to the linker which objects must not be dead-stripped.
shows an example from a linker command file that tells the linker not to dead-strip an object named
InterruptVectorTable
.
Listing 13-1. FORCEACTIVE example
FORCEACTIVE { InterruptVectorTable }
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
206 Freescale Semiconductor, Inc.
Chapter 13 Linker for AIOP
Use FORCEFILES directive to prevent deadstripping entire files.
206 shows an example from a linker command file that prevents the linker dead-stripping
entire files.
Listing 13-2. FORCEFILES example
FORCEFILES { segfault.o }
13.3 Defining the Target's Memory Map
Use the linker command file's
MEMORY
directive to delineate areas in the target platform's memory map and associate a name for each of these areas. Names defined in a
MEMORY directive may be used later in the linker command file to specify where object code
should be stored. Listing 13-3 on page 207
shows an example.
Listing 13-3. MEMORY directive example
MEMORY
{
ISR_table : org = 0x00000000, len = 0x400
data : org = 0x00000400, len = 0x10000
flash: org = 0x10000000, len = 0x10000
text : org = 0x80000000
}
This example defines 4 memory areas named
ISR_table
, data, flash
, and text
. The org argument specifies the beginning byte address of a memory area. The len
argument is optional, It specifies how many bytes of data or executable code the linker may store in an area. The linker issues a warning message if an attempt to store object code in an area exceeds its length.
13.4 Defining Sections in the Output File
Use the linker command file's
SECTIONS
directive to
• define sections in the linker's output file
• to specify in which memory area on the target platform a section in the output file should be loaded at runtime
Use
GROUP
directives in a
SECTIONS
directive to organize objects.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
207
Associating Input Sections With Output Sections
The linker will only create a section in the output file if the section is not empty, even if the section is defined in a
SECTIONS
or
GROUP
directive.
shows an example.
Listing 13-4. SECTIONS and GROUP example
SECTIONS
{
GROUP :
{
.text : {}
.rodata : {}
} > text
GROUP
{
.sdata : {}
.sbss : {}
} > data
GROUP
{
.sdata2 : {}
.sbss2 : {}
} > data
}
This example defines the
.text
and
.rodata
sections in the output file and specifies that they should be loaded in the memory area named text
on the target platform at runtime.
The example then defines sections named
.sdata
and
.sbss
. These sections will be loaded in the memory named data
. The last
GROUP
directive in the example defines sections named
.sdata2
, and
.sbss2
. These sections will also be loaded in the memory area named data
, after the sections
.sdata
and
.sbss
.
13.5 Associating Input Sections With Output Sections
Normally the linker stores sections from input object code in the sections of the linker's output file that have the same name. The linker command file's
SECTIONS
and
GROUP directives allow you to specify other ways to associate input object code with sections in
linker output. Listing 13-5 on page 208 shows an example.
Listing 13-5. Associating object code with sections in linker output
SECTIONS
{
GROUP :
{
.myText : { main.o (.text) }
.text : ( *(.text) }
} > text
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
208 Freescale Semiconductor, Inc.
Chapter 13 Linker for AIOP
}
This example defines a section in the output file named
.myText
. This section will contain the objects that are in the
.text
section in the object code taken from the input file named main.o
. The example also defines a section in the output file named
.text
. This section will contain all objects in the
.text
sections of all input files containing object code. Both these sections in the output file,
.myText
and
.text
, will be loaded in the memory area named text
on the target platform.
The
SECTIONS
and
GROUP
directives also allow you to filter what kinds of object code from
input files will be stored in a section in the output file. Table 13-1 shows the kinds of data
that may be filtered.
TEXT
CODE
DATA
BSS
CONST
MIXED
VLECODE
Table 13-1. Filter types for object code in input files
This filter allows input objects that have these permissions
readable, executable readable, executable readable, writable readable, writable readable readable, writable, executable readable, executable
and contain this kind of object code
initialized initialized initialized uninitialized initialized initialized initialized
shows an example.
Listing 13-6. Filtering objects from input files
SECTIONS
{
.text (TEXT) : { } > text
.bss (BSS) : { } > data
}
This example defines a section in the output file named
.text
. The linker will only store objects from input object code that are readable, executable, and initialized. This example also defines a section in the output file named
.bss
. This section will only contain objects from the linker's input files that are readable, writable, and uninitialized.
13.6 Controlling Alignment
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
209
Specifying Memory Area Locations and Sizes
Use the
ALIGN
argument in a
SECTIONS
or
GROUP
directive to specify an alignment relative to the start of the physical address.
shows an example.
Listing 13-7. Example of the ALIGN directive
SECTIONS
{
GROUP:
{
.init ALIGN(0x1000) : {}
.text ALIGN(0x1000) : {}
} > text
}
This example defines two sections named
.init
and
.text
. At runtime, each section will be loaded at the next available address that is evenly divisible by
0x1000
in the memory area named text
on the target platform.
13.7 Specifying Memory Area Locations and Sizes
Normally, the linker stores sections in the output file in sequential order. Each object from the linker's output is stored after the last object in the output file. Use the
BIND
,
ADDR
, and
SIZEOF
keywords in
SECTIONS
and
GROUP
directives to precisely specify where sections in the output file will be loaded.
shows an example.
Listing 13-8. BIND, ADDR, and SIZEOF example
SECTIONS
{
.text BIND(0x00010000) : ()
.rodata : {}
.data BIND(ADDR(.rodata + SIZEOF(.rodata)) ALIGN(0x010) : {}
}
This example defines a section in the output file named
.text
. This section will be loaded at address
0x00010000
on the target platform at runtime. The next section,
.rodata
, will be loaded at the address immediately proceeding the last byte in the
.text
section. The last section,
.data
, will be loaded at the address that is the sum of the beginning of the
.rodata
section's address and the size of the
.rodata
section. This last section will be aligned at the next address that is evenly divisible by
0x10
.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
210 Freescale Semiconductor, Inc.
Chapter 13 Linker for AIOP
The dot keyword ("
.
"), is a convenient way to set the linker's place in the current output section.
shows an example.
Listing 13-9. Skipping areas of memory
SECTIONS
{
GROUP :
{
.ISR_Table : {}
. = 0x2000
} > flash
GROUP :
{
.paramsection : {}
} > flash
}
This example defines two sections. The first section,
.ISRTable
, will be loaded at beginning of the memory area named flash
on the target platform at runtime. The second section,
.paramsection
, will be loaded at the address that is
0x2000
bytes past the beginning of the memory area named flash
.
13.8 Creating Memory Gaps
You can create gaps in memory by performing alignment calculations such as
. = (. + 0x20) & ~0x20;
This kind of calculation can occur between output_specs
, between input_specs,
or even in address_modifiers
. A "." refers to the current address. You may assign the
.
to a specific unallocated address or just do alignment as the example shows. The gap is filled with zeroes, in the case of an alignment (but not with
ALIGN()
).
You can specify an alternate fill pattern with
= <short_value>
, as in
.text : { . = (. + 0x20) & ~0x20; *(.text) } = 0xAB > text short_value
is 2 bytes long. Note that the fill pattern comes before the memory_spec
. You can add a fill to a
GROUP
or to an individual output_spec
section. Fills cannot be added between
.bss
type sections. All calculations must end in a "
;
".
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
211
Linker Command File Syntax
13.9 Creating Symbols
You can create symbols that you can use in your program by assigning a symbol to some value in your linker command file.
.text : { _red_start = .; *(.text) _red_end = .;} > text
In the example above, the linker generates the symbols
_red_start
and
_red_end as 32 bit values that you can access in your source files.
_red_start
is the address of the first byte of the
.text
section and
_red_end
is the byte that follows the last byte of the
.text section.
You can use any of the pseudo functions in the address_modifiers
in a calculation.
The CodeWarrior linker automatically generates symbols for the start address, the end address, and the start address for the section if it is to be burned into ROM. For a section
.red
, we create
_f_red
,
_e_red, and
_f_red_rom
. In all cases, any
"."
in the name is replaced with a "
_
". Addresses begin with an "
_f
", addresses after the last byte in section begin with an "
_e
", and ROM addresses end in a "
_rom
". See the header file __ ppc_eabi_linker.h
for further details.
All user defined sections follow the preceding pattern. However, you can override one or more of the symbols that the linker generates by defining the symbol in the linker command file.
NOTE
BSS sections do not have a ROM symbol.
13.10 Linker Command File Syntax
Linker command file syntax is a notation and implies what an LCF file includes.
13-10 on page 212 shows the syntax for linker command files.
Listing 13-10. Linker Command File Syntax
linker-command-file =
command* memory? command* sections? command*
This syntax implies that an LCF file can contain:
Zero or more command directives followed by Zero or at most one memory directive followed by Zero or more command directives followed by Zero or at most one sections directive followed by Zero or more command directives.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
212 Freescale Semiconductor, Inc.
*
?
Chapter 13 Linker for AIOP
The table below lists the notations used in the linker command file syntax.
Table 13-2. Linker Command File Notations
Notation Description
Implies zero or any number of directives
Implies zero or at most one directive
Listing 13-11 on page 213 shows the syntax of all valid linker command file keywords,
directives, and commands.
Listing 13-11. Linker Command File Syntax (Commands, Directives, and Keywords)
command =
exclude-files |
force-active |
force-files |
include-dwarf |
keep |
ref-include |
shorten-names-for-tornado-101 |
cats-bss-mod |
cats-header-mod |
data-type-converts |
entry |
init |
term |
external-symbol |
internal-symbol |
memory-gaps exclude-files =
"EXCLUDEFILES" "{" file-name+ "}" force-active =
"FORCEACTIVE" "{" symbol+ "}" letter =
'a'|'b'|'c'|'d'|'e'|'f'|'g'|'h'|'i'|'j'|'k'|'l'|'m'|
'n'|'o'|'p'|'q'|'r'|'s'|'t'|'u'|'v'|'w'|'x'|'y'|'z'|
'A'|'B'|'C'|'D'|'E'|'F'|'G'|'H'|'I'|'J'|'K'|'L'|'M'|
'N'|'O'|'P'|'Q'|'R'|'S'|'T'|'U'|'V'|'W'|'X'|'Y'|'Z' file-name =
(letter |"_") (letter |digit |"_")*
(".")?(letter |digit |"_")* section-name =
(letter |"_") (letter |digit |"_")* object-file =
(letter |"_") (letter |digit |"_")* (".") ("o"|"O") archive-file =
(letter |"_") (letter | digit |"_")* (".") ("a"|"A") include-dwarf =
"INCLUDEDWARF" "{" file-name "}" keep =
"KEEP""(" *( section-name )")" ref-include =
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
213
Linker Command File Syntax
"REF_INCLUDE" "{" section-name+ "}" shorten-names-for-tornado-101=
"SHORTEN_NAMES_FOR_TOR_101" cats-bss-mod =
"CATS_BSS_MOD" cats-header-mod =
"CATS_HEADER_MOD" data-type-converts =
"DATA_TYPE_CONVERTS" entry =
"ENTRY" "(" symbol ")" init =
"INIT" "(" symbol ")" term =
"TERM" "(" symbol ")" external-symbol =
"EXTERNAL_SYMBOL" "{" symbol ["," symbol] "}" internal-symbol =
"INTERNAL_SYMBOL" "{" symbol ["," symbol] "}" group=
"GROUP" address-modifiers ":"
"{" (section-spec )* "}" ["=" fill-shortnumber ]
[ "> " mem-area-symbol ] hexadigit =
'0'|'1'|'2'|'3'|'4'|'5'|'6'|'7'|'8'|'9'|
'A'|'B'|'C'|'D'|'E'|'a'|'b'|'c'|'d'|'e' digit =
'0'|'1'|'2'|'3'|'4'|'5'|'6'|'7'|'8'|'9' hexadecimal-number =
"0"("x"|"X")hexadigit+ decimal-number =
digit+ number =
hexadecimal-number | decimal-number binary-op =
'+'|'-'|'*'|'/'| '%' '==' | '!=' | '>'| '>='| '<'| '<='|
'&&' | '||' '>>' | '<<' | '&' | '|' | '' | '' unary-op =
'++' | '--' | '~' | '!' postfix-unary-op =
'++' | '--' symbol-declaration =
( symbol "=" address-spec ) |
( "PROVIDE" "(" identifier "=" address=spec ")" ) |
( "EXPORTSYMTAB") |
( "EXPORTSTRTAB") |
( "BYTE" "(" address-spec ")" |
( "SHORT" "(" address-spec ")" |
( "LONG" "(" address-spec ")" symbol =
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
214 Freescale Semiconductor, Inc.
Chapter 13 Linker for AIOP
(letter |"_") (letter|"_"|digit )* operand =
number |
("ADDR" "(" output-section-spec | address-expr ")" ) |
("ROMADDR" "(" output-section-spec | address-expr ")" |
("SIZEOF" "(" output-section-spec | address-expr ")" ) |
("SIZEOF_ROM" "(" output-section-spec | address-expr ")" ) address-spec =
number |
"." |
operand |
(address-spec binary-op operand ) |
(unary-op address-spec ) |
(address-spec postfix-unary-op ) memory-spec =
memory-area-symbol ":" "origin" |
"org" |
"o" "=" number "," "length"|
"len" |
"l" "=" number
[">" file-name] memory-gaps =
"." "=" address-spec memory =
"MEMORY" "{" memory-spec + "}" sections =
"SECTIONS" "{"
(section-spec | memory-gaps | symbol-declaration | group )*
"}" section-spec =
output-section-name ":"
["(" input-type ")"]
[address-modifiers ] "{"
[( input-section-spec )*] "}"
[= fill-shortnumber] [(">"|">>") memory-area-symbol ] output-section-name =
section-name input-type =
[ "TEXT" | "DATA" | "BSS" | "CONST" | "ZTEXT" | "ZCODE"|
"VLECODE" ] address-modifiers =
["BIND" "(" address-spec ")" ]
["ALIGN" "(" address-spec ")" ]
["NEXT" "(" address-spec ")"]
["LOAD" "(" address-spec ")"]
["INTERNAL_LOAD" "(" address-spec ")"] input-section-spec =
( file-name |
file-name "(" section-name ")" |
"*(" section-name ")" |
symbol-declaration |
data-write )+ data-write =
("LONG" | "SHORT" | "BYTE" ) "(" number ")" fill-shortnumber =
number
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
215
Commands, Directives, and Keywords
13.10.1 Linker Command File Operator Precedence and Support
||
=
|
&
&&
!
()
~
* / %
+
<< >>
-
==
!=
< <=
The table below lists the LCF operators supported and their order of precedence (highest to lowest). Their associativity indicates in what order operators of equal precedence in an expression are applied.
> >=
Table 13-3. Linker Command File Operator Precedence and Support
Operator Description
Parenthesis
Logical negation
Bitwise complement
Multiplication/Division/Modulus
Addition
Bitwise shift left, Bitwise shift right
Is equal
Subtraction
Is not equal to
Relational less than/Less than or equal to
Relational greater than/Greater than or equal to
Bitwise and
Bitwise inclusive or
Logical and
Logical or
Assignment left-to-right left-to-right left-to-right left-to-right left-to-right left-to-right left-to-right left-to-right left-to-right left-to-right left-to-right left-to-right left-to-right left-to-right left-to-right right-to-left
Associativity
13.11 Commands, Directives, and Keywords
The rest of this chapter consists of explanations of all valid LCF functions, keywords, directives, and commands, in alphabetic order.
•
•
•
•
•
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
216 Freescale Semiconductor, Inc.
Chapter 13 Linker for AIOP
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
13.11.1 . (location counter)
Denotes the current output location.
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
217
Commands, Directives, and Keywords
The period always refers to a location in a sections segment, so is valid only in a sectionssection definition. Within such a definition, '
.
' may appear anywhere a symbol is valid.
Assigning a new, greater value to '
.
' causes the location counter to advance. But it is not possible to decrease the location-counter value, so it is not possible to assign a new, lesser value to '
.
' You can use this effect to create empty space in an output section, as the
Listing 13-12 on page 218 example does.
Example
The code of
moves the location counter to a position 0x10000 bytes past the symbol
__start
.
Listing 13-12. Moving the Location Counter
..data :
{
*.(data)
*.(bss)
*.(COMMON)
__start = .;
. = __start + 0x10000;
__end = .;
} > DATA
13.11.2 ADDR
Returns the address of the named section or memory segment.
ADDR (sectionName | segmentName)
Parameters
sectionName
Identifier for a file section.
segmentName
Identifier for a memory segment
Example
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
218 Freescale Semiconductor, Inc.
Chapter 13 Linker for AIOP
The code of
uses the
ADDR
function to assign the address of
ROOT
to the symbol
__rootbasecode
.
Listing 13-13. ADDR() Function
MEMORY{
ROOT : origin = 0x80000400, length = 0
}
SECTIONS{
.code :
{
__rootbasecode = ADDR(ROOT);
*.(text);
} > ROOT
}
13.11.3 ALIGN
Returns the location-counter value, aligned on a specified boundary.
ALIGN(alignValue)
Parameter
alignValue
Alignment-boundary specifier; must be a power of two.
Remarks
The
ALIGN
function does not update the location counter; it only performs arithmetic.
Updating the location counter requires an assignment such as:
. = ALIGN(0x10); #update location counter to 16-byte alignment
13.11.4 ALIGNALL
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
219
Commands, Directives, and Keywords
Forces minimum alignment for all objects in the current segment to the specified value.
ALIGNALL(alignValue);
Parameter
alignValue
Alignment-value specifier; must be a power of two.
Remarks
ALIGNALL
is the command version of the
ALIGN
function. It updates the location counter as each object is written to the output.
Example
Listing 13-14 on page 220 is an example use for
ALIGNALL()
command.
Listing 13-14. ALIGNALL Example
.code :
{
ALIGNALL(16); // Align code on 16-byte boundary
* (.init)
* (.text)
ALIGNALL(64); //align data on 64-byte boundary
* (.rodata)
} > .text
13.11.5 BIND
Specifies a section modifier for setting an address.
BIND(address)
Parameter
address
A memory address, such as 0x80000400.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
220 Freescale Semiconductor, Inc.
Chapter 13 Linker for AIOP
13.11.6 BIN_FILE_TYPE
Controls the generation of a binary file. By default, the linker does not generate any binary file, even if the s-record generation is enabled.
BIN_FILE_TYPE (NO_BIN_FILE | ONE_BIN_FILE
|MULTIPLE_BIN_FILES)
Parameters
NO_BIN_FILE
No binary file will be generated even if s-record generation is on.
ONE_BIN_FILE
Generates a single binary file with all the loadable code and data, even if s-record generation is off.
MULTIPLE_BIN_FILES
A separate binary file is generated for each MEMORY directive. When selected, the hexidecimal address of the location, the file should be loaded is inserted between the filename and extension.
Example
Listing 13-15 on page 221 is an example of a binary file generated using the
MULTIPLE_BIN_FILES parameter for a non-ROM Image build.
Listing 13-15. Binary File for a non-ROM Image Build Example
Listing 13-16 on page 221 is an example of a binary file generated using the
MULTIPLE_BIN_FILES parameter for a ROM Image build.
Listing 13-16. Binary File for a ROM Image Build Example
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
221
Commands, Directives, and Keywords
The code of
uses the MULTIPLE_BIN_FILES parameter to change the name of a binary file in the MEMORY directive.
Listing 13-17. Change Binary File Name in the MEMORY Directive
MEMORY {
code : org = 0x00002000 > my_binary_file.xyz
Any MEMORY directive without a custom name will be given a name like
Test.ffe00000.bin, where the binary file reflects the ROM address of the segement.
The code of
uses the MULTIPLE_BIN_FILES parameter to merge some binary files together, assuming that there are no segments between them.
Listing 13-18. Merge Binary Files Together Example
MEMORY {
code : org = 0x00002000 > my_binary_file.xyz
special : org = 0x00004000 > my_binary_file.xyz
13.11.7 BYTE
Inserts a byte of data at the current address of a section.
BYTE (
expression
) ;
Parameter
expression
Any expression that returns a value
0x00
to
0xFF
.
13.11.8 EXCEPTION
Creates the exception table index in the output file.
EXCEPTION
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
222 Freescale Semiconductor, Inc.
Chapter 13 Linker for AIOP
Only C++ code requires exception tables. To create an exception table, add the
EXCEPTION command, with symbols
__exception_table_start__
and
__exception_table_end__
, to the end of your code section segment, just as
Listing 13-19 on page 223 shows. (At runtime, the
system knows the values of the two symbols.)
Example
Listing 13-19 on page 223 shows the code for creating an exception table.
Listing 13-19. Creating an Exception Table
__exception_table_start__ = .;
EXCEPTION
__exception_table_end__ = .;
13.11.9 EXCLUDEFILES
Ignores object code in files.
Syntax
EXCLUDEFILES {
file-name }
Remarks
This directive is for partial link projects only. It makes your partial link file smaller.
EXCLUDEFILES
can be used independently of
INCLUDEDWARF
,
EXCLUDEFILES
can take any number of executable files.
In this example
EXCLUDEFILES { kernel.elf } kernel.elf
is added to your project but the linker does not add any section from kernel.elf
to your project. However, it does delete any weak symbol from your partial link that also exists in kernel.elf
. Weak symbols can come from templates or out-of-line inline functions.
13.11.10 EXPORTSTRTAB
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
223
Commands, Directives, and Keywords
Creates a string table from the names of exported symbols.
EXPORTSTRTAB
Remarks
Table 13-4 shows the structure of the export string table. As with an ELF string table, the
system zero-terminates the library and symbol names.
0x00 library name
symbol1 name symbol2 name
Table 13-4. Export String Table Structure
1 byte varies varies varies
Example
Listing 13-20 on page 224 shows the code for creating an export string table.
Listing 13-20. Creating an Export String Table
.expstr:
{
EXPORTSTRTAB
} > EXPSTR
13.11.11 EXPORTSYMTAB
Creates a jump table of the exported symbols.
EXPORTSYMTAB
Remarks
Table 13-5 shows the structure of the export symbol table. The start of the export symbol
table must be aligned on at least a four-byte boundary.
Table 13-5. Export Symbol Table Structure
Size (in bytes) of export table
Index to library name in export symbol table
Index to symbol1 name in export symbol table
Address of symbol1
4 bytes
4 bytes
4 bytes
4 bytes
Table continues on the next page...
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
224 Freescale Semiconductor, Inc.
Chapter 13 Linker for AIOP
Table 13-5. Export Symbol Table Structure (continued)
A5 value for symbol1
Index to symbol2 name in export symbol table
Address of symbolf2
A5 value for symbol2
4 bytes
4 bytes
4 bytes
4 bytes
Example
Listing 13-21 on page 225 shows the code for creating an export symbol table.
Listing 13-21. Creating an Export Symbol Table
.expsym:
{
EXPORTSYMTAB
} > EXPSYM
13.11.12 EXTERNAL_SYMBOL
Specifies objects that may be referred to by modules outside of an object code file.
Syntax
EXTERNAL_SYMBOL {
symbol [,
symbol ]* }
Remarks
The symbols must be the link time symbol names. In the case of C++ object code, these names must be the mangled.
13.11.13 FORCEACTIVE
Specifies objects that must not be deadstripped.
Syntax
FORCEACTIVE {
symbol [,
symbol ]* }
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
225
Commands, Directives, and Keywords
13.11.14 GROUP
Organizes objects listed in a linker command file.
Syntax
GROUP
address-modifiers : {
section-spec
[,
section-spec ]* }
[ >
memory-area-name ]
Remarks
The
GROUP
directive lets you organize the linker command file.
Listing 13-22 on page 226 shows that each group starts at a specified address. If no
address_modifiers
are present, it would start following the previous section or group.
Although you normally do not have an address_modifier for an output_spec
within a group; all sections in a group follow contiguously unless there is an address_modifier for that output_spec
.
The optional memory-area-name clause specifies a name defined with the
MEMORY directive. Using this clause specifies the memory space in the target platform where the group's object code should be placed.
Listing 13-22. Example of linker GROUP directive
SECTIONS {
GROUP BIND(0x00010000) : {
.text : {}
.rodata : {*(.rodata) *(extab) *(extabindex)}
}
GROUP BIND(0x2000) : {
.data : {}
.bss : {}
.sdata BIND(0x3500) : {}
.sbss : {}
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
226 Freescale Semiconductor, Inc.
Chapter 13 Linker for AIOP
.sdata2 : {}
.sbss2 : {}
}
GROUP BIND(0xffff8000) : {
.PPC.EMB.sdata0 : {}
.PPC.EMB.sbss0 : {}
}
}
13.11.15 IMPORTSTRTAB
Creates a string table from the names of imported symbols.
IMPORTSTRTAB
Remarks
Table 13-6 shows the structure of the import string table. As with an ELF string table, the
system zero-terminates the library and symbol names.
0x00 library name
symbol1 name symbol2 name
Table 13-6. Import String Table Structure
1 byte varies varies varies
Example
Listing 13-23 on page 227 shows the code for creating an import string table.
Listing 13-23. Creating an Import String Table
.impstr:
{
IMPORTSTRTAB
} > IMPSTR
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
227
Commands, Directives, and Keywords
13.11.16 IMPORTSYMTAB
Creates a jump table of the imported symbols.
IMPORTSYMTAB
Remarks
Table 13-7 shows the structure of the import symbol table. The start of the import symbol
table must be aligned on at least a four-byte boundary.
Table 13-7. Import Symbol Table Structure
Size (in bytes) of import table
Index to library1 name in import string table
Number of entries in library1
Index to symbol1 name in import string table
Address of symbol1 vector in export string table
Index to symbol2 name in import string table
Address of symbol2 vector in export string table
Index to library2 name in import string table
Number of entries in library2
4 bytes
4 bytes
4 bytes
4 bytes
4 bytes
4 bytes
4 bytes
4 bytes
4 bytes
Example
Listing 13-24 on page 228 shows the code for creating an import symbol table.
Listing 13-24. Creating an Import Symbol Table
.expsym:
{
IMPORTSYMTAB
} > EXPSYM
13.11.17 INCLUDE
Include a specified binary file in the output file.
INCLUDE filename
Parameter
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
228 Freescale Semiconductor, Inc.
Chapter 13 Linker for AIOP
filename
Name of a binary file in the project. The File Mappings target settings panel must specify resource file for all files that have the same extension as this file.
13.11.18 INCLUDEDWARF
Allows source-level kernel debugging.
Syntax
INCLUDEDWARF { file-name }
Remarks
In this example
INCLUDEDDWARF { kernel.elf } the linker inserts the
.debug
and
.line
sections of kernel.elf
to your application. These sections let you debug source level code in the kernel while debugging your application.
You are limited to one executable file when using this directive. If you need to process more than one executable, add this directive to another file.
13.11.19 INTERNAL_LOAD
Loads one or several segments at an address not specified at link time.
Syntax
INTERNAL_LOAD(address)
Parameter
address
A memory address, such as 0x80000400.
Remarks
Use INTERNAL_LOAD directive to specify an internal ROM addr_mode.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
229
Commands, Directives, and Keywords
13.11.20 INTERNAL_SYMBOL
Symbols created in a linker command file are considered external unless they are redefined using INTERNAL_SYMBOL
Syntax
INTERNAL_SYMBOL {
symbol [,
symbol ]* }
13.11.21 KEEP
Forces the linker to not dead strip the unused symbols in the specified section.
Syntax
KEEP(*(sectionType))
Parameter
sectionType
Identifier for any user-defined or predefined section.
Example
Listing 13-25 on page 230 shows the sample usage.
Listing 13-25. KEEP Directive Usage
GROUP : {
.text (TEXT) : {}
.mycode (TEXT) : {KEEP(*(.mycode))}
...}
13.11.22 LOAD
Loads one or several segments at a specific address.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
230 Freescale Semiconductor, Inc.
Chapter 13 Linker for AIOP
Syntax
LOAD(address)
Parameter address
A memory address, such as 0x80000400.
Remarks
• Use LOAD directive to specify an external ROM addr_mode.
• The LOAD directive can be used on all the sections going into ROM.
• The directive is ignored in links if Generate ROM image is not checked in the linker preference panel.
• Inconsistent initialized data sections copied from ROM to RAM may need a reference to a ROM address. For example, to store the .data and .sdata sections in
ROM such that they are immediately after the .text section, try the following:
.text LOAD(ADDR(.text)) : {} > rom
.data LOAD(ROMADDR(.text) + SIZEOF(.text)): {} > code
.sdata LOAD(ROMADDR(.data) + SIZEOF(.data)): {} > code
13.11.23 LONG
Inserts a word of data at the current address of a section.
LONG (
expression );
Parameter
expression
Any expression that returns a value
0x00000000
to
0xFFFFFFFF
.
13.11.24 MEMORY
Starts the LCF memory segment, which defines segments of target memory.
MEMORY { memory_spec[, memory_spec] }
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
231
Commands, Directives, and Keywords
Parameters
memory_spec segmentName: origin = address, length = length [> fileName] segmentName
Name for a new segment of target memory. Consists of alphanumeric characters; can include the underscore character.
accessFlags
ELF-access permission flags -
R
= read,
W
= write, or
X
= execute.
address
A memory address should be a numeric value, such as 0x80000400.
length
Size of the new memory segment: a value greater than zero. Optionally, the value zero for autolength , in which the linker allocates space for all the data and code of the segment. (Autolength cannot increase the amount of target memory, so the feature can lead to overflow.) fileName
Optional, binary-file destination. The linker writes the segment to this binary file on disk, instead of to an ELF program header. The linker puts this binary file in the same folder as the ELF output file. This option has two variants:
•
> fileName: writes the segment to a new binary file.
•
>> fileName: appends the segment to an existing binary file.
Remarks
The LCF contains only one
MEMORY
directive, but this directive can define as many memory segments as you wish.
For each memory segment, the
ORIGIN
keyword introduces the starting address, and the
LENGTH
keyword introduces the length value.
There is no overflow checking for the autolength feature. To prevent overflow,the segment start address should be specified correctly.
Example
Listing 13-26 on page 232 is an example use of the
MEMORY
directive.
Listing 13-26. MEMORY Directive Example
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
232 Freescale Semiconductor, Inc.
Chapter 13 Linker for AIOP
MEMORY {
TEXT (RX) : ORIGIN = 0x00003000, LENGTH = 0
DATA (RW) : ORIGIN = 0x5000, LENGTH = 0
}
13.11.25 MAX_BIN_GAP
Controls the maximum gap size value between two segments.
MAX_BIN_GAP(nnnnn)
Parameters
nnnnn
Size of the maximum gap allowed between segments.
Remarks
The directive can be placed in the LCF anyway except within the MEMORY and
SECTIONS directives.
This directive can only be used if you are generating a single binary file.
13.11.26 NEXT
Specifies an expression for setting an address.
NEXT(address)
Parameter
address
A memory address, such as 0x80000400.
13.11.27 NO_INIT_BSS
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
233
Commands, Directives, and Keywords
Prevents initializing a BSS section to zero to avoid uninitialized variable to be iniatialized to zero by the startup.
Syntax
NO_INIT_BSS
Remarks
If the variable is initialized to 0 while using the directive
NO_INIT_BSS
then the variable will be placed in
BSS
section and will not be initialized to zero ( Listing 13-27 on page 234 ).
Example
Following are few examples of using the NO_INIT_BSS directive.
Listing 13-27. NO_INIT_BSS Directive Usage - Example 1
LCF:
.......
GROUP : {
.__uninitialized_intc_handlertable ALIGN(0x10) : {}
.data : {}
.user_def_Init (DATA):{}
.sdata : {}
.sbss : {}
.sdata2 : {}
.sbss2 : {}
.bss : {}
.user_def_Uninit (BSS) NO_INIT_BSS :{}
} > internal_ram
.......
Test case:
#pragma section RW ".user_def_Init" ".user_def_Uninit"
// first user defined section treated as initialized section and second
// user defined section treated as uninitialized section
__declspec (section ".user_def_Init") int myvar_a;
// myvar_a goes to user_def_Uninit section as it is not initialized
__declspec (section ".user_def_Init") int myvar_b=3;
// myvar_b goes to user_def_Init section as it is initialized
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
234 Freescale Semiconductor, Inc.
Chapter 13 Linker for AIOP
__declspec (section ".user_def_Init") int myvar_c=0;
// myvar_c goes to user_def_Init section as it is initialized to zero
// But var myvar_c is not initialized to zero as it comes under
NO_INIT_BSS int main(void) {
volatile int i = 0;
/* Loop forever */
for (;;) {
i++;
myvar_a ++;
myvar_b ++;
}
}
Listing 13-28. NO_INIT_BSS Directive Usage - Example 2
LCF:
.......
GROUP : {
.__uninitialized_intc_handlertable ALIGN(0x10) : {}
.data : {}
.user_def_Init (DATA):{}
.sdata : {}
.sbss : {}
.sdata2 : {}
.sbss2 : {}
.bss : {}
.user_def_Uninit (BSS) NO_INIT_BSS :{}
} > internal_ram
.......
Test case:
#pragma push //Save the original pragma context
#pragma section data_type sdata_type ".user_def_Init"
".user_def_Uninit" unsigned int test_var; //Here variable test_var will place in the section user_def_Uninit and NOT initializes to zero.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
235
Commands, Directives, and Keywords
#pragma pop unsigned int test_var2 = 0; //Here variable test_var2 will place in
.bss/.sbss section and initializes to zero.
Listing 13-29. NO_INIT_BSS Directive Usage - Example 3
LCF:
.......
GROUP : {
.__uninitialized_intc_handlertable ALIGN(0x10) : {}
.data : {}
.user_def_Init (DATA):{}
.sdata : {}
.sbss : {}
.sdata2 : {}
.sbss2 : {}
.bss : {}
.user_def_Uninit (BSS) NO_INIT_BSS :{}
} > internal_ram
.......
Test case:
#pragma push //Save the original pragma context
#pragma section data_type sdata_type ".user_def_Init"
".user_def_Uninit" unsigned int test_var; // Variable test_var will place in the section user_def_Uninit
#pragma explicit_zero_data on unsigned int test_var2 = 0; // Variable test_var2 will place in the section user_def_Init as the pragma // explicit_zero_data is enabled
#pragma explicit_zero_data reset
(OR #pragma explicit_zero_data off) unsigned int test_var3 = 5; //Variable test_var3 will place in the section user_def_Init unsigned int test_var4 = 0; //Variable test_var3 will place in the section user_def_Uninit
#pragma pop
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
236 Freescale Semiconductor, Inc.
Chapter 13 Linker for AIOP
13.11.28 NO_TRAILING_BSS_IN_BIN_FILES
Removes uninitiallized data contained in a binary file.
Syntax
NO_TRAILING_BSS_IN_BIN_FILES
Remarks
This directive can only be used if the last section or sections of a binary file contains uninitialized data.
Example
Listing 13-30 on page 237 is an example use of the
NO_TRAILING_BSS_IN_BIN_FILES directive.
Listing 13-30. NO_TRAILING_BSS_IN_BIN_FILES Directive Example
SECTIONS
{
GROUP {
.text:{}
}
>code
}
NO_TRAILING_BSS_IN_BIN_FILES
13.11.29 OBJECT
Sections-segment keyword that specifies a function. Multiple
OBJECT
keywords control the order of functions in the output file.
OBJECT (function, sourcefile.c)
Parameters
function
Name of a function.
sourcefile.c
Name of the C file that contains the function.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
237
Commands, Directives, and Keywords
Remarks
If an
OBJECT
keyword tells the linker to write an object to the output file, the linker does not write the same object again, in response to either the
GROUP
keyword or the '*' wildcard character.
13.11.30 REF_INCLUDE
Starts an optional LCF closure segment that specifies sections the linker should not deadstrip, if program code references the files that contain these sections.
REF_INCLUDE{ sectionType[, sectionType] }
Parameter
sectionType
Identifier for any user-defined or predefined section.
Remarks
Useful if you want to include version information from your source file components.
13.11.31 REGISTER
Use the REGISTER directive to assign one of the AIOP's non-volatile registers to a userdefined small data section.
REGISTER(nn [ , limit])
Parameter
nn
Specifies one of the predefined small data base registers, a non-volatile AIOP register, or any of the following values:
• 0, 2, 13
These registers are for the predefined small data sections:
0 - .PPC.EMB.sdata0/.PPC.EMB.sbss0
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
238 Freescale Semiconductor, Inc.
Chapter 13 Linker for AIOP
2 - .sdata2/sbss2
13 - .sdata/sbss
You do not have to define these sections using REGISTER because they are predefined.
• 14 - 31
Match any value in this range with the register reserved by your global register variable declaration.
• -1
This "register" value instructs the linker to treat relocations that refer to objects in your small data section as non-small data area relocations. These objects are converted to near absolute relocations, which means that the objects referenced must reside within the first
32 KB of memory. If they do not, the linker emits a "relocation out of range" error. To fix this problem, rewrite your code such that the offending objects use large data relocations.
limit
Specifies the maximum size of the small data section to which register nn is bound. This value is the size of the initialized and uninitialized sections of the small data section combined. If limit is not specified, 0x00008000 is used.
NOTE
Each small data section you create makes one less register available to the compiler; it is possible to starve the compiler of registers. As a result, create only the number of small data sections you need.
13.11.32 ROMADDR
Equivalent to ADDR. Returns ROM address.
ROMADDR (sectionName | segmentName)
Parameters
sectionName
Identifier for a file section.
segmentName
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
239
Commands, Directives, and Keywords
Identifier for a memory segment
Example
The code of
uses the
ROMADDR
function to assign the address of
ROM
to the symbol
__rootbasecode
.
Listing 13-31. ROMADDR() Function
MEMORY{
ROM : origin = 0x80000400, length = 0
}
SECTIONS{
.code :
{
__rootbasecode = ROMADDR(ROM);
*.(text);
} > ROM
}
13.11.33 SECTIONS
Starts the LCF sections segment, which defines the contents of target-memory sections.
Also defines global symbols to be used in the output file.
SECTIONS { section_spec[, section_spec] }
Parameters
section_spec sectionName :
[LOAD (loadAddress)]
{contents} > segmentName sectionName
Name for the output section. Must start with a period.
LOAD
Loads one or several segments at a specific address.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
240 Freescale Semiconductor, Inc.
Chapter 13 Linker for AIOP
contents
Statements that assign a value to a symbol or specify section placement, including input sections.
segmentName
Predefined memory-segment destination for the contents of the section. The two variants are:
•
> segmentName:
puts section contents at the beginning of memory segment segmentName
.
•
>> segmentName:
appends section contents to the end of memory segment segmentName
.
Example
Listing 13-32 on page 241 is an example sections-segment definition.
Listing 13-32. SECTIONS Directive Example
SECTIONS {
.text : {
_textSegmentStart = .;
alpha.c (.text)
. = ALIGN (0x10);
beta.c (.text)
_textSegmentEnd = .;
}
.data : { *(.data) }
.bss : { *(.bss)
*(COMMON)
}
}
13.11.34 SHORT
Inserts a halfword of data at the current address of a section.
SHORT
(
expression );
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
241
Commands, Directives, and Keywords
Parameter
expression
Any expression that returns a value
0x0000
to
0xFFFF
13.11.35 SIZEOF
Returns the size (in bytes) of the specified segment or section.
SIZEOF(segmentName | sectionName)
Parameters
segmentName
Name of a segment; must start with a period.
sectionName
Name of a section; must start with a period.
13.11.36 SIZEOF_ROM
Returns the size (in bytes) that a segment occupies in ROM.
SIZEOF_ROM (segmentName)
Parameter
segmentName
Name of a ROM segment; must start with a period.
Remarks
Always returns the value 0 until the ROM is built. Accordingly, you should use
SIZEOF_ROM only within an expression inside a
BYTE
,
SHORT
, or
LONG
function.
Furthermore, you need
SIZEOF_ROM
only if you use the
COMPRESS
option on the memory segment. Without compression, there is no difference between the return values of
SIZEOF_ROM
and
SIZEOF
.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
242 Freescale Semiconductor, Inc.
Chapter 13 Linker for AIOP
13.11.37 WRITES0COMMENT
Inserts an S0 comment record into an S-record file.
WRITES0COMMENT "comment"
Parameter
comment
Comment text: a string of alphanumerical characters
0-9
,
A-Z
, and a-z
, plus space, underscore, and dash characters. Double quotes must enclose the comment string. (If you omit the closing double-quote character, the linker tries to put the entire LCF into the
S0 comment.)
Remarks
This command, valid only in an LCF sections segment, creates an S0 record of the form:
S0aa0000bbbbbbbbbbbbbbbbdd
• aa
- hexadecimal number of bytes that follow
• bb
- ASCII equivalent of comment
• dd
- the checksum
This command does not null-terminate the ASCII string.
Within a comment string, do not use these character sequences, which are reserved for
LCF comments:
# /* */ //
Example
This example shows that multi-line
S0
comments are valid:
WRITES0COMMENT "Line 1 comment
Line 2 comment"
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
243
Commands, Directives, and Keywords
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
244 Freescale Semiconductor, Inc.
Chapter 14
Linker
This chapter describes how to use the features in the CodeWarrior linker that are specific to AIOP software development.
•
•
•
•
CodeWarrior Linker Command File (LCF)
•
•
•
Relocating Code and Data in Internal RAM
•
Relocating Code and Data in External MRAM
•
•
14.1 Predefined Sections
Table 14-1 describes the sections that the compiler creates.
NOTE
The Compiler-defined section names are case sensitive. For example, using .binary instead of .BINARY will not give expected results.
.bss
.BINARY
.ctors
Name
Table 14-1. Compiler-defined sections
Description
uninitialized global data
Binary files.
C++ constructors and Altivec vector constructors
Table continues on the next page...
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
245
Predefined Sections
.dtors
.data
extab extabindex
.init
.init_vle
.PPC.EMB.sdata0
.PPC.EMB.sbss0
.rodata
.sdata
.sdata2
.sbss
.sbss2
.text
.text_vle
.tdata
.tbss
Table 14-1. Compiler-defined sections (continued)
Name Description
C++ destructors initialized global data
C++ exception tables
C++ exception tables initialization executable code from the runtime library
Initialization executable code for VLE compilers
Initialized data with addressing relative to address 0
Uninitialized data with addressing relative to address 0 literal values and initialization values in the application's source code initialized small global data initialized global small data defined with the const keyword uninitialized global small data uninitialized global constant small data defined with the const keyword application code application code for VLE compilers
AIOP task variable data section
AIOP task variable uninitialized data section
14.1.1 Linking Binary Files
You can link external binary files/data (tables, Bitmap graphics, sound records) into the project image. The following sections explain how to link binary files using CodeWarrior
IDE and Command line:
•
•
14.1.1.1 Using CodeWarrior IDE
To link a binary file using CodeWarrior IDE, perform the following steps:
1. Launch CodeWarrior and open the desired project to add the binary file.
2. Add a binary file ( bin_data.bin
) to project.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
246 Freescale Semiconductor, Inc.
Chapter 14 Linker
3. Select Properties > C/C++ Build > Settings > Tool Settings > AIOP Linker >
General and add the binary file to the Other Flags text box.
4. Update linker command file (.lcf) and place .BINARY section into memory. Listing
shows a sample linker command file with .BINARY section.
Listing 14-1. Linker Command File with .BINARY section
MEMORY
{
resetvector: org = 0x00000000, len = 0x00000008
init: org = 0x00000020, len = 0x00000FE0
exception_handlers: org = 0x00001000, len = 0x00001000
internal_flash: org = 0x00002000, len = 0x001FD000
my_binary_data: org = 0x001FE000, len = 0x00001000
...
}
SECTIONS
{
.__bam_bootarea LOAD (0x00000000): {} > resetvector
...
.binary1_area:
{
binary1Start = .;
bin_data1.bin
binary1End = .;
} > my_binary_data
}
.binary2_area:
{
binary2Start = .;
bin_data2.bin
binary2End = .;
} > my_binary_data
}
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
247
Additional Small Data Sections
}
14.1.1.2 Using Command-Line
To link a binary file using Command line, perform the following steps:
1. Linker recognizes
.bin
extension as a binary data input file. If binary file has another extension it may not be recognized correctly by the command line linker.
2. Update linker command file (
.lcf
) and place
.BINARY
section into memory.
Command File with .BINARY section
shows a sample linker command file with
.BINARY
section.
3. Add a binary file (
.bin
) as an input file for linker (ldaiop.exe) ldaiop main.o msl.lib bin_data.bin -o myapp.elf -lcf commandfile.lcf
14.2 Additional Small Data Sections
The AIOP EABI specification mandates that compliant build tools predefine three small data sections. The Properties > C/C++ Build > Settings > Tool Settings > Linker >
Input panel lets you specify the address at which the CodeWarrior linker puts two of these sections (if the default locations are unsatisfactory).
CodeWarrior Development Studio, lets you create small data sections in addition to those mandated by the AIOP EABI specification. The CodeWarrior tools let you specify that the contents of a given user-defined section will be accessed by the small data base register selected from the available non-volatile registers. To do this, you use a combination of source code statements and linker command file directives.
To create one additional small data area, follow these steps:
1. Open the CodeWarrior project in which you want to create an additional small data section.
2. Select the build target in which you want to create an additional small data section.
3. Select Project > Properties.
The IDE displays the Properties for <Project> window.
4. In the left pane of the Properties for <Project>, select Compiler > Preprocessor.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
248 Freescale Semiconductor, Inc.
Chapter 14 Linker
5. Click Browse to open the prefix file whose name appears in the Prefix File text box in an editor window.
6. Add the statements that define a small data section to the top of the prefix file: a. Add a statement that creates a global register variable.
For example, to create a global register variable for register 14, add this statement to the prefix file:
// _dummy does not have to be defined extern int _dummy asm("r14"); b. Create a user-defined section using the section pragma; include the clause data_mode = sda_rel
so the section can use small data area addressing.
For example:
// you do not have to use the names in this example
// .red is the initialized part of the section
// .blue is the uninitialized part of the section#pragma section RW ".red" ".blue" data_mode = sda_rel
NOTE
If you want your small data area to be the default section for all small data, use the following form of the section pragma instead of the one above:
#pragma section sdata_type ".red" "blue" data_mode = sda_rel
7. Save the prefix file and close the editor window.
8. In each header or source file that declares or defines a global variable that you want to put in a small data section, put the storage-class modifier
__declspec(section
"initialized_small_sect_nm")
in front of the definition or declaration.
For example, the statement:
__declspec(section ".red") int x = 5; instructs the compiler to put the global variable x into the small data section named .red
CAUTION
The section name specified in the __declspec(section
<section_name>) statement must be the name of an initialized data section. It is an error to use the uninitialized data section name.
NOTE
The semantics of __declspec(section ".sdata") int x; is to use the section pair .sdata and .sbss to store x. The location where x is stored is determined by whether or not x is explicitly initialized.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
249
Additional Small Data Sections
NOTE
If you want your small data section to be the default section for all small data, use #pragma section sdata_type ".foo"
".bar" data_mode = sda_relUse __declspec(section
".foo")only when the object is greater than the size threshold for small data.
9. On the Tool Settings tab, select Linker > Input .
10. Click Browse to specify a Link Command File (.lcf).
11. On the Tool Settings tab, select CPU.
12. From the Code Model listbox, select Absolute Addressing.
13. From the ABI listbox, select EABI .
14. Click OK.
The IDE saves your settings and closes theProperties for <Project> window.
15. Modify the project's linker command file such that it instructs the linker to use the global register declared above as the base register for your new small data section.
To do this, follow these steps: a. In the linker command file, add two REGISTER directives, one for the initialized part of the small data section and one for uninitialized part.
For example, to make register 14 the base register, add statements like these:
.red REGISTER(14) : {} > ram
.blue REGISTER(14) : {} > ram b. Add the linker command file to each build target in which you want to use the new small data section.
16. Open the CodeWarrior project for the runtime library used by your project.
17. In the build target listbox of the runtime library project window, select the build target of the runtime library that your main project uses.
18. Open this build target's prefix file in a CodeWarrior editor window.
19. Add the same statements to this prefix file that you added to the prefix file of the main project.
20. Save the prefix file and close the editor window.
21. Open __start.c in a CodeWarrior editor window.
22. Find the string __init_registers(void) and add statements that initialize the small data section base register you are using near the end of this function (immediately above the terminating blr instruction).
For example, to initialize register 14, add these statements:
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
250 Freescale Semiconductor, Inc.
Chapter 14 Linker
lis r14, _SDA14_BASE_@ha addi r14, r14, _SDA14_BASE_@l
23. Save __start.c and close the editor window.
24. Open __ppc_eabi_linker.h in a CodeWarrior editor window.
25. Find the string _SDA_BASE_[] in this file and add this statement after the block of statements that follow this string:
// SDAnn_BASE is defined by the linker if
// the REGISTER(nn) directive appears in the .lcf file
__declspec(section ".init") extern char _SDA14_BASE_[];
26. Save __ppc_eabi_linker.h and close the editor window.
27. Return to your main project.
28. Select Project > Build Project.
The IDE builds your project.
You can now use the new small data section in this project.
NOTE
You can create more small data segments by following the procedure above. Remember, however, that for each small data section created, the compiler loses one non-volatile register to use for other purposes.
14.3 Linker Map File
A linker map file is a text file containing information about a program's global symbols, source file and source line numbers. The linker names the map file with the base name of the program and the extension .map. The linker map consists of the following sections:
•
•
•
•
14.3.1 Closure
The linker lists all the required objects under the closure section with the following details:
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
251
Linker Map File
• Level of closure: Object B is in Object A's closure if and only if, the level of B is higher than the level of A and one of the following conditions is true:
Condition 1: There is no object in between B and A.
Condition 2: There are objects between B and A, and the level of A is lower than the levels of all the objects between B and A.
• Object name: specifies the name of an object.
• Object characteristics: specifies the characteristics of an object. They can be one of the following:
• function, local | global | weak
• section, local | global | weak
• object, local | global | weak
• notype, local | global | weak
• Object locations: specifies an object location.
The listing below shows a sample closure section.
Listing 14-2. Sample closure section
1] reset (func,global) found in reset.o
2] __reset (func,global) found in 8568mds_init.o
3] __start (func,global) found in Runtime.AIOPEABI.E2.UC.a __start.o
4] __init_registers (func,weak) found in Runtime.AIOPEABI.E2.UC.a __start.o
5] _stack_addr found as linker generated symbol
5] _SDA2_BASE_ found as linker generated symbol
5] _SDA_BASE_ found as linker generated symbol
4] __init_hardware (func,global) found in __AIOP_eabi_init.o
5] usr_init (func,global) found in 8568mds_init.o
6] gInterruptVectorTable (notype,global) found in AIOP_exception.o
7] gInterruptVectorTableEnd (notype,global) found in AIOP_exception.o
7] .intvec (section,local) found in AIOP_exception.o
8] InterruptHandler (func,global) found in interrupt.o
9] @21 (object,local) found in interrupt.o
9] printf (func,global) found in MSL_C.AIOPEABI.bare.E2.UC.a printf.o
9] __msl_count_trailing_zero64 (func,weak) found in MSL_C.AIOPEABI.bare.E.a math_double.o
9] >>> UNREFERENCED DUPLICATE __msl_count_trailing_zero64
9] >>> (func,weak) found in MSL_C.AIOPEABI.bare.E.a math_float.o
9] >>> (func,weak) found in MSL_C.AIOPEABI.bare.E.a math_longdouble.o
9] >>> (func,weak) found in MSL_C.AIOPEABI.bare.E.a math_AIOP.o
In the sample above:
•
__reset
is in the closure of reset because:
•
__reset
is of level 2, reset is of level 1 and
• there is no object in between
__reset
and reset
•
_SDA_BASE_
is in the closure of
__init_registers
because:
•
_SDA_BASE_
is of level 5,
__init_registers
is of level 4; and
• the objects between
__init_registers
and
_SDA_BASE_
are all of level 5 and are higher than the level of
__init_registers
•
InterruptHandler
is in the closure of
__init_hardware
because:
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
252 Freescale Semiconductor, Inc.
Chapter 14 Linker
•
InterruptHandler
is of level 8,
__init_hardware
is of level 4; and
• the objects between
__init_hardware
and
InterruptHandler
are of level 5, 6, 7 respectively and are all higher than the level of
__init_hardware
•
__init_hardware
is NOT in the closure of
_init_registers
because they both are of level
4
• gInterruptVectorTableEnd
is NOT in the closure of
__init_registers
because:
• the objects between gInterruptVectorTableEnd
and
__init_registers
are not all of a higher level than
__init_registers
•
__init_hardware
is of the same level as
__init_registers
.
Weak symbols are allowed by the ABI and are global. They can have the same name as another symbol. The line before the UNREFERENCED DUPLICATE lists the first weak symbol found by the linker, that appears in the executable.
The line after the UNREFERENCED DUPLICATE lists other versions of a same object found by the linker. Linker will not copy the duplicate objects to the executable.
14.3.2 Section Layout
The linker lists information of all the objects within a section in a section layout. Listing
shows a sample .text section layout.
Listing 14-3. Sample .text section layout
.text section layout
Starting Virtual File
address Size address offset
---------------------------------
00000084 000030 fffc1964 00001ce4 1 .text
00000084 00000c fffc1964 00001ce4 4 __init_user __AIOP_eabi_init.o
00000090 000020 fffc1970 00001cf0 4 exit __AIOP_eabi_init.o
000000b0 000004 fffc1990 00001d10 4 _ExitProcess __AIOP_eabi_init.o
...
UNUSED 000030 ........ ........ __copy Runtime.AIOPEABI.E2.UC.a CPlusLibAIOP.o
UNUSED 000084 ........ ........ __init_arr Runtime.AIOPEABI.E2.UC.a CPlusLibAIOP.o
...
The first line of a section layout specifies the name of a section. Starting from the 5th line
(the line after the dotted line separator), objects within section are listed with the following information:
•
Starting address
: specifies the starting address of an object. The object is listed as
UNUSED if it is dead-stripped.
•
Size
: specifies the size of each object in a section.
•
Virtual address
: specifies the virtual address of the object.
•
File offset
: specifies the offset of an object in the section.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
253
Linker Map File
•
Alignment
: specifies the alignment of an objects. For legacy reasons, the alignment of all section symbols is 1. In reality, a section symbol is the highest alignment of all symbols in its section which in the above listing is 4.
•
Object name
: specifies the name of an object. The names are the C or mangled C++, depending on the language. The name of an object is similar to the one in a disassembled file.
•
Object location
: specifies the location an object. This is usually a name of the object file (.o), when no other column exists. In presence of an other column, the library file information is listed here.
In the sample above, note that the 5th line has a section name and that its Starting address, Virtual address and File offset values are similar to
__init_user
values. The 5th line is the section symbol for the objects that follow it. Its Size column is the total size of the section in the executable file (after dead stripping) and its alignment (column 5) is 1.
Although
Listing 14-3 on page 253 is from a normal .o file, if this project had an input
file which was a partially linked file, then you would see a section symbol between each merged section from each input file used to create the partially linked file. For example, if plf.o is a partially linked file, and it is composed of a.o and b.o and each of those files had a .text section, then plf.o would have one .text section but two .text section symbols.
The .text content from a.o would follow the first section symbol and the content from b.o
would follow the second section symbol.
14.3.3 Memory Map
You can verify segment allocation from the Memory map section in a linker map file.
shows a sample Memory map section.
Listing 14-4. Verifying segment allocation in a .MAP file
•
Starting address
: specifies the starting address for each section. Constant and executable object code are allocated in ROM space and data object code is allocated in RAM space.
•
Size
: specifies the size of each sections.
• File offset: specifies the offset of a section in the file.
•
ROM Address
: specifies the address of the section in the ROM image. For executable code and constants sections,
Starting address
is equal to
ROM Address
. For data
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
254 Freescale Semiconductor, Inc.
Chapter 14 Linker
sections,
ROM Address
specifies the address in ROM where the initialization values are stored.
•
RAM Buffer Address
: specifies the address in RAM that is to be used as a buffer for the flash image programmer. It is important to note that the RAM buffer is not used when the RAM address equals to the ROM address.
•
S-Record Line
: specifies the line number of a section in the S-Record file in decimal format.
•
Bin File Offset
: specifies the offset of a section in the binary file.
•
Bin File Name
: specifies the binary file name of the section. The file name also reflects the ROM address of the section.
14.3.4 Linker Generated Symbols
You can find a complete list of the linker generated symbols and user-defined symbols in either the C include file
__ppc_eabi_linker.h
or the assembly include file
__ppc_eabi_linker.i
. The CodeWarrior linker automatically generates symbols for the start address, the end address (the first byte after the last byte of the section), and the start address for the section if it will be burned into ROM. With a few exceptions, all
CodeWarrior linker-generated symbols are immediate 32 bit values.
255 shows a sample list of linker-generated symbols.
Listing 14-5. Sample list of linker-generated symbols
_f_init 000034d8
_f_init_rom 000034d8
_e_init 000035b0
_f_init_vle 000035b0
_f_init_vle_rom 000035b0
_e_init_vle 00003864
_f_text 00003864
_f_text_rom 00003864
_e_text 00003864
_f_text_vle 00003870
_f_text_vle_rom 00003870
_e_text_vle 00003ad4
If addresses are declared in your source file as unsigned char _f_text[];
you can treat
_f_text
just as a C variable even though it is a 32-bit immediate value.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
255
Linker Map File
unsigned int textsize = _e_text - _f_text;
If you do need linker symbols that are not addresses, you can access them from C.
unsigned int size = (unsigned int)&_text_size;
The linker generates four symbols:
•
__ctors
- an array of static constructors
•
__dtors
- an array of destructors
•
__rom_copy_info
- an array of a structure that contains all of the necessary information about all initialized sections to copy them from ROM to RAM
•
__bss_init_info
- a similar array that contains all of the information necessary to initialize all of the bss-type sections. Please see
__init_data
in
__start.c
.
These four symbols are actually not 32-bit immediate values but are variables with storage. You access them just as C variables. The startup code now automatically handles initializing all bss type sections and moves all necessary sections from ROM to RAM, even for user defined sections.
14.3.5 Stack Usage Estimation
In the AIOP subsystem, stack space is a precious resource. The total available workspace allocation is divided among the total number of tasks the system is programmed to handle--the more tasks that are specified, the less available workspace memory for each task. The workspace memory serves as the stack area for AIOP. Furthermore, the workspace memory also serves as storage for the hardware context, presentation area, and for task local variables.
Stack space, of course, is used by each function in a call tree for resources and objects used by that function. As the depth of the call tree increases, the more stack space is used.
Stack space is affected by many factors, but generally, the linkage area (LR register and back link pointer), stack based local variables, and saved non-volatile registers usually accounts for most of the stack space used.
For a complete view of stack usage, a runtime trace of the program is required. However, a reasonable estimate of stack usage can be made statically. Issues such as recursion, variable length arrays (which use stack space based on a variable at program execution time), interrupts and exceptions, and calls through function pointers may all contribute to stack usage, but are difficult to measure statically.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
256 Freescale Semiconductor, Inc.
Chapter 14 Linker
To aid the programmer in tracking stack usage in AIOP, the compiler, linker, and linker control file cooperate to give an estimate of stack usage. This information is reported in the linker map file. With the support of compiler based pragmas, the build tools try to construct a picture of stack usage that is as accurate as possible at link time.
The listing below shows a sample output of the stack usage reported within the map file.
Listing 14-6. Stack usage estimation - Output
ESTIMATED STACK USAGE root-function-name: <stack_used> <cumulative_stack_usage> <stack_remaining>
child1: <stack_used> <cumulative_stack_usage> <stack_remaining>
child11: <stack_used> <cumulative_stack_usage> <stack_remaining>
child12: <stack_used>* <cumulative_stack_usage> <stack_remaining>
child2: <stack_size> <cumulative_stack_usage> <stack_remaining>
root-function-name2: <stack_used> <cumulative_stack_usage> <stack_remaining> etc. …
The report lists:
• each root function specified (or all entry point functions, if the root-function is “all” on the command line)
• function name followed by the size of the stack used in bytes (
<stack_used>
) by the function
• the accumulated usage of stack by the function and its children
(
<cumulative_stack_usage>
)
• the available stack remaining (
<stack_remaining>
).
Beneath each function and indented are the children of the function (all the function calls made by the parent function), and beneath those children and indented are its children and so on.
The
<cumulative_stack_usage>
value is determined by adding the largest
<stack_used>
value of the children to the currently used stack value. For example, if the parent function has several children at the same level (siblings), the largest
<stack_used>
value of all the siblings would be added to the
<stack_used>
value of the current function. If a parent calls the same child function multiple times, it is only reported once. For leaf functions
(functions with no children), the
<stack_used>
value and
<cumulative_stack_usage>
value are the same value.
The
<stack_remaining>
value represents the number of bytes remaining in the available stack allocation for the worst case call tree. This value is equivalent to the number of bytes used so far at this point in the call tree plus the current
<cumulative_stack_usage>
. This value is subtracted from the initial stack size. For example:
Initial stack size: 1024
F1: <2> <10> <1014>***
F2: <2> <2> <1020>
F3: <4> <8> <1014>***
F4: <4> <4> <1014>***
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
257
Linker Map File
F1’s
<stack_remaining>
is 1014 = 1024 – <10> (F1’s cumulative_stack_usage).
F2’s
<stack_remaining>
is 1020 = 1024 - <2> (F2’s cumulative_stack_usage ) - <2> (F1’s stack_used).
F3’s
<stack_remaining>
is 1014 = 1024 - <8> (F3’s cumulative_stack_usage) - <2> (F1’s stack_used).
F4’s
<stack_remaining>
is 1014 = 1024 - <4> (F4’s cumulative_stack_usage) - <4> (F3’s stack_used) - <2> (F1’s stack_used).
The worst case paths for stack usage are indicated by the
“***”
annotation and the
<stack_remaining>
value represents the worst case stack remaining value. All
<stack_remaining>
values along the worst case paths will be the same value. A negative
<stack_remaining>
value indicates that stack use has exceeded the stack allocation. There can be multiple worst case paths for stack usage. In the example above, the worst case stack usage path is F1->F3->F4.
In some cases, the compiler and linker are unable to determine the stack usage information for a child function (perhaps because the reference is to an assembler defined function). In this case, the stack usage estimator reports the referenced function usage as zero for
<stack_used>
and zero for
<cumulative_stack_usage>
values. The child function for which the stack usage information is missing is prepended with “[NA]”.
[NA]child_function: <0> <0> <stack_remaining>
Additionally, if the reference is via an indirect function call (function pointer call) and the callee cannot be determined, the unknown function is annotated with “*”. This will help the user identify unknown indirect calls which may be mitigated via the
#pragma fn_ptr_candidates()
.
(*[NA]unknown): <0> <0> <stack_remaining>
Indirect function calls identified using the
#pragma fn_ptr_candidates
are identified with the following annotations:
(*indirect_child_name): <stack_used> <cumulative_stack_usage>
<stack_remaining>
14.3.5.1 Enabling stack usage estimation
Stack usage estimation can be enabled using the following options:
•
•
•
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
258 Freescale Semiconductor, Inc.
Chapter 14 Linker
14.3.5.1.1 Using compiler options
By default, the compiler (ccaiop) collects call tree and stack usage information in a special section of the ELF object file. The following compiler command-line option controls the collection of stack information in the compiler:
•
14.3.5.1.1.1 -[no]stackinfo
Controls the collection of stack information in the compiler. Stack information collection is enabled by default in the compiler.
–nostackinfo disables collection of stack information in the compiler.
Syntax
-[no]stackinfo
Remarks
The option is supported on the command-line only.
14.3.5.1.2 Using linker options
The following linker command-line options are required to enable stack estimation feature:
•
•
•
•
•
14.3.5.1.2.1 -estimate_stack_usage
Performs the stack usage estimate analysis.
Syntax
-estimate_stack_usage all | 'root[;root2;...]' | @<path_to_file>
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
259
Linker Map File
Parameters
all
If the root_function is all
then the tool will iterate through all the entry point functions
(identified by
__declspec(entry_point)
) and perform the analysis for each entry point.
'root1;root2;root3;'
If the root functions are provided as input on the command-line seperated by a delimiter
; then the tool will iterate through each of the functions specified in the list.
NOTE
No spaces are allowed in between the root function names. The maximum size of the root function list parameter is 256 characters including both the function names and delimiters.
Some command line shells interpret
;
as a command seperator.
It is suggested to enclose the root function name list within single quotes, for example
‘root1;root2;root3’
, to prevent interpretation of
;
by the shell.
@<path_to_file>
If a list of functions is to be input from a file, prepend the parameter with
@
. The tool will iterate through each of the function in the list contained in the specified file.
NOTE
The functions names should be separated by a delimiter ';'.No
spaces are allowed in between root function names. The maximum size of the root function name list is 256 characters.
Only one root function name list is allowed in the file.
Comments are allowed in the file and are prepended with an '#' symbol. Multiple comment lines are also allowed.
Example
# Comment line
# Root list specification follows: root1;root2;root3
14.3.5.1.2.2 -map
Generates a linker map file with the stack usage information.
Syntax
-map
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
260 Freescale Semiconductor, Inc.
Chapter 14 Linker
Remarks
This is a linker option.
14.3.5.1.2.3 -recursion_depth
Controls level of recursion reporting in the map file report. The option can only be activated if
-estimate_stack_usage
is specified.
Syntax
-recursion_depth
Remarks
This is a linker option. The default
-recursion_depth
is 1 if not specified.
14.3.5.1.2.4 -root_sort
Sorts root function reports and outputs the results of the Stack Estimator in descending sorted order according to the cumulative stack usage of the root function.
Syntax
-root_sort
Remarks
This is a linker option.
14.3.5.1.2.5 -ws_size
Allows the user to specify the workspace size of each task.
Syntax
-ws_size <size>
Parameters
<size>
Workspace size in bytes. If not specified, the default is 2048 (2Kb).
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
261
Linker Map File
In cases where a single program is executing and has access to the entire workspace,
ws_size
can be set to the whole workspace size.
14.3.5.1.3 Using CodeWarrior
To enable stack usage estimation using the CodeWarrior IDE, perform the following steps:
1. Launch CodeWarrior and open the desired project.
2. Select Properties > C/C++ Build > Settings > Tool Settings > Linker > Output
Figure 14-1. Enabling stack usage estimation using CodeWarrior
3. Select Generate Link Map, if not already selected.
4. Select List Estimated Stack Usage.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
262 Freescale Semiconductor, Inc.
Chapter 14 Linker
5. Specify the Root Function. This is the function at which the analysis is performed.
Normally this would be main()
, but it could be any function. If the root_function is all then the tool will iterate through all the entry point functions (identified by
__declspec(entry_point)
) and perform the analysis for each entry point.
6. Specify the Workspace Size in bytes. If not specified, the default is 2048 (2 KB).
14.3.5.2 AIOP Workspace Layout
The workspace memory for a task is shared by the Hardware context area, the
Presentation area, and the Task local variables. The remaining memory in the task workspace area is available for the stack.
Figure 14-2. Task workspace SRAM - Detailed logical view
Since the stack usage estimation feature reports a value indicating the stack remaining for each function, there must be a method of determining the initial stack space available to a root function. The Hardware Context and Task Local Storage values are static and are determined by the linker. However, the Presentation Area can be different for each entry point and is determined by the semantics of the entry point function (determined by the application programmer). Obviously, we must know the value of the Presentation
Area in order to determine the remaining memory in the workspace that will serve as stack memory. By default, Presentation Area is assumed to be 0 bytes.
In order to calculate the initial workspace memory available as stack space, linker uses the following formula: initial stack size = (workspace_size – (_stack_end + _entry point presentation size)) where:
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
263
Linker Map File
• workspace_size
is the value defined by the linker option,
–ws_size
. The default is 2048 bytes.
•
_stack_end
is a user linker control file defined symbol pointing to the end of the stack allocation for the application.
_stack_end
is an address relative to the beginning of the workspace area. The stack grows towards lower addresses. Therefore, the value of
_stack_end
is always between
0
and the workspace size.
• entry point presentation size
is determined by the
#pragma presentation_size(<size>)
.
NOTE
It is the user’s responsibility to define
_stack_end
in the linker control file and it must be defined at the first address following the task local storage area. The Presentation Area size may vary with the entry point, but the default value of all entry points is 0 bytes.
For example, given the LCF settings below:
MEMORY {
…
stack: org = 0x400, len = 0x400
…
}
_stack_end = ADDR(stack);
If the workspace size is defined to be
2048
and presentation size is
0
, then the initial stack size is
2048 – 0x400 = 1024
bytes.
14.3.5.3 Compiler Pragma Support
There are several compiler pragmas that may assist in making the stack estimation feature more accurate. The supported pragmas are listed below.
NOTE
All the pragmas are optional and not necessary for the feature to function properly.
•
• #pragma stackinfo_ignore
•
•
14.3.5.3.1 #pragma stack_used
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
264 Freescale Semiconductor, Inc.
Chapter 14 Linker
Overrides the stack size determined by the compiler. The compiler will calculate the stack usage of a function during compilation. This pragma allows the user to set their own value for stack usage. The scope of the pragma is the next function definition.
Syntax
#pragma stack_used(<size>)
<function definition>
Example
#pragma stack_used(16) void func_a(void) {
}
The stack size for func_a
is recorded as 16 bytes, regardless of the whatwhat the actual stack frame size is.
14.3.5.3.2 #pragma presentation_size
Specifies the size of presentation area in bytes. The presentation size is used in calculating the available stack space for the scoped functions. The scope of this pragma is the file unless overridden by another
#pragma presentation_size
. The pragma has no effect on non-entry point functions (function declaration with
__declspec(entry_point)
). The net effect of increasing the presentation area is to decrease the workspace memory available for stack usage.
Syntax
#pragma presentation_size(<size>)
Example
#pragma presentation_size(192)
__declspec(entry_point) funcA() { … } static int funcB() { … }
__declspec(entry_point) funcC() { … }
Presentation size is 192 until the end of the file or until another
#pragma presentation_size is encountered. By default, the presentation size is 0 for all entry point functions.
14.3.5.3.3 #pragma fn_ptr_candidates
Calls through function pointers are not directly detectable by the compiler. The fn_ptr_candidates
pragma lists all the possible functions a caller may call. Some, all, or none of the candidates in the pragma may be actually called, but all candidates are placed in the caller’s call tree database.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
265
Deadstripping
The scope of the pragma is the next function defined and the pragma is reset after the function definition. There is no limit on the number of candidates in the pragma’s callee list.
Syntax
#pragma fn_ptr_candidates(<func>,<func2>,…<funcN>)
Example
#pragma fn_ptr_candidates(baz,bop,boo) typedef void (*FN)(void); void caller(FN f, FN g) {
f();
g();
}
In the above example, the list of callees for the function caller
is baz
, bop
, and boo
, which may be invoked via the f
and g
function pointers.. The user may provide this information using the fn_ptr_candidates
pragma. Otherwise, there will be no recorded callees for caller
.
Example output (boo is not defined in any file and thus reports as [NA]):
…
caller: <32> <160> <65220>
(*baz): <128> <128> <65220>
(*bop): <128> <128> <65220>
(*[NA]boo): <0> <0> <65348>
14.3.5.4 Limitations
Following is a list of limitations and known issues of stack usage estimation:
• Recursion, variable length arrays, interrupts and exceptions, and calls through function pointers are not handled by the stack estimator (unless the pragma fn_ptr_candidates
is used).
• Stack usage information is not recorded for assembler files.
14.4 Deadstripping
If the Pool Data checkbox is checked in the CodeWarrior IDE's panel, the pooled data is not stripped. However, all small data and code is still subject to deadstripping.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
266 Freescale Semiconductor, Inc.
Chapter 14 Linker
14.5 CodeWarrior Linker Command File (LCF)
Linker command files are an alternative way of specifying segment addresses.
Only one linker command file is supported per target in a project. The linker command filename must end in the
.lcf
extension.
The LCF along with other compiler directives, places pieces of code and data into ROM and RAM. You can do this by creating specific sections in the LCF and then matching them to the source code using pragma directives.
LCF consists of three kinds of segments, which must be in this order:
• A memory segment, which begins with the
MEMORY{} directive,
• An optional closure segment, which begins with the
FORCE_ACTIVE{}, KEEP_SECTION{}
, or
REF_INCLUDE{} directives, and
• A sections segment, which begins with the
SECTIONS{}
directive.
14.6 Creating an LCF from Scratch
Consider creating a LCF for a sample
BOOKE ROM
project.
First add the memory area for vectors (interrupts), exception tables, code, data, heap and stack.
Listing 14-7. Adding memory area
MEMORY
{
resetvector: org = 0x00000000, len = 0x00000008
exception_handlers_p0: org = 0x00001000, len = 0x00001000
internal_flash: org = 0x00003000, len = 0x001FD000
internal_ram: org = 0x40000000, len = 0x0007C000
heap : org = 0x4007C000, len = 0x00002000 /* Heap start location */
stack : org = 0x4007E000, len = 0x00002000 /* Start location for Stack
*/
}
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
267
Creating an LCF from Scratch
Place the sections to the above memory areas in LCF in the SECTIONS { } block.
.__bam_bootarea LOAD (0x00000000): {} > resetvector
The code to handle exceptions are grouped and placed in memory area exception_handlers_p0.
Listing 14-8. Grouping exceptions code
GROUP : {
.ivor_branch_table_p0 LOAD (0x00001000) : {}
.intc_hw_branch_table_p0 LOAD (0x00001800): {}
.__exception_handlers_p0 LOAD (0x00001100) : {}
} > exception_handlers_p0
The hardware initialization routines, application code, constants, code for constructors/ destructors, and C++ exception tables are grouped together and placed in Flash.
Listing 14-9. Grouping initialization routines
GROUP : {
.intc_sw_isr_vector_table_p0 ALIGN (2048) : {}
.init : {}
.text : {}
.rodata (CONST) : {
*(.rdata)
*(.rodata)
}
.ctors : {}
.dtors : {}
extab : {}
extabindex : {}
} > internal_flash
The uninitialized and initialized data are placed in RAM.
Listing 14-10. Placing data in RAM
GROUP : {
.__uninitialized_intc_handlertable ALIGN(2048) : {}
.data : {}
.sdata : {}
.sbss : {}
.sdata2 : {}
.sbss2 : {}
.bss : {}
} > internal_ram
The sections are allocated to segments in the order given in
SECTIONS/GROUP
block of lcf file.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
268 Freescale Semiconductor, Inc.
Chapter 14 Linker
For internal_flash
segment, following is the order of section allocation: .init, .text, .rodata, .ctors, .dtors, extab and extabindex.
Variables are added in LCF and these can be used in application as well as internally in linker tool for computation.
Listing 14-11. Adding variables in LCF
_stack_addr = ADDR(stack)+SIZEOF(stack);
_stack_end = ADDR(stack);
_heap_addr = ADDR(heap);
_heap_end = ADDR(heap)+SIZEOF(heap);
EXCEPTION_HANDLERS = ADDR(exception_handlers_p0);
L2SRAM_LOCATION = 0x40000000;
Let us take a simple example to see how the allocation of variables to the respective sections take place.
Listing 14-12. C Source file
#include "MPC5675K.h" int sdata_i = 10; int sbss_i; const char sdata2_array[] = "Hello";
__declspec(section ".rodata") const char rodata_array[40]="CodeWarior";
__declspec(section ".data") long bss_i;
__declspec(section ".data") long data_i = 10; int main(void) {
return sdata_i + sbss_i + sdata2_array[3] + data_i + bss_i + rodata_array[5];
}
NOTE
Above is a hypothetical example built to provide clarity on variables and their allocation to sections. __declspec is used to forcefully place the variables into sections.
The table below lists the details of the allocation of objects to the sections.
sdata_i sbss_i sdata2_array
Variable
Table 14-2. Allocating Objects
Section
.sdata
.sbss
.sdata2
Table continues on the next page...
0x400004d8
0x400004e8
0x40000500
Address
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
269
Relocating Code in ROM
rodata_array bss_i data_i
Variable
Table 14-2. Allocating Objects (continued)
Section
.rodata
.bss
.data
0x00003938
0x40000508
0x400004d0
Address
14.7 Relocating Code in ROM
To place data and code in a specific memory location there are two general steps that must be performed.
• Use pragma compiler directives to tell the compiler which part of the code is going to be relocated.
• Tell the linker where the code will be placed within the memory map using LCF definitions.
14.7.1 Relocating Function in ROM
To put code in a specific memory section it is needed first to create the section using the section pragma directive. In the following listing a new section called .
romsymbols
is created.
All the content in this section is going to be referenced in the LCF with the name
.romsymbols
. After defining a new section you can place code in this section by using the
__declspec()
directive.
In the following listing,
__declspec() directive is used to tell the compiler that function funcInROM() is going to be placed in section romsymbols
.
Create a stationary project for any target and add the following code to your main.c
file before the main() function and have a call to this function.
Listing 14-13. Code to add in the main.c
#pragma section RX ".romsymbols" data_mode=far_abs
__declspec(section ".romsymbols") void funcInROM(int flag); //Function Prototype
void funcInROM(int flag){ if (flag > 0)
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
270 Freescale Semiconductor, Inc.
Chapter 14 Linker
{ flag ++;
}
}
14.7.2 Placing Code in ROM
You have just edited a source file to tell the compiler which code will be relocated. Next, the LCF needs to be edited to tell the linker the memory addresses where these sections are going to be allocated.
First you need to define a new Microcontroller memory segment where new sections will be allocated.
You can have just one memory segment for all the new sections or one segment for each section.
14.7.2.1 Create New ROM Segment
Below you can find the memory segment of a LCF. Notice that the segment internal_flash has been edited and its length has been reduced by
0x10000
from its original size. This memory space is taken to create the new segment. In the following listing the new segment is called myrom
, it will be located next to segment internal_flash
and its length is going to be
0x10000
. You can calculate the address where segment code ends by adding its length plus the origin address.
Edit your LCF as shown in the following listing. Ensure you edit ROM target lcf.
Listing 14-14. Memory Segment of LCF
MEMORY
{
resetvector: org = 0x00000000, len = 0x00000008
init: org = 0x00000010, len = 0x0000FFF0
exception_handlers_p0: org = 0x00010000, len = 0x00010000
internal_flash: org = 0x00030000, len = 0x001C0000
myrom: org = 0x00220000, len = 0x00010000
internal_ram: org = 0x40000000, len = 0x0007C000
heap : org = 0x4007C000, len = 0x00002000 /* z7_0 Heap start location */
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
271
Relocating Code and Data in Internal RAM
stack : org = 0x4007E000, len = 0x00002000 /* z7_0 Start location for
Stack */
}
14.7.2.2 Create New ROM Section
The next step is to add the content of the new section into the Microcontroller memory segment you have reserved. This is done in the sections segment of the LCF.
The code below creates a new section called
.rom_symbols
, then the label
__ROM_SYMBOLS points to the address where the section begins. Then
*(.romsymbols)
instruction is used to tell the linker that all the code referenced with this word is going to be placed in section
.rom_symbols
.
Finally you close the section telling the linker that this content is going to be located in segment myrom.
Edit your LCF as shown below.
Listing 14-15. Code to add your LCF.
.rom_symbols :
{
__ROM_SYMBOLS = . ; #start address of the new symbol area
. = ALIGN (0x4);
*(.romsymbols) #actual data matching pragma directives.
. = ALIGN (0x4);
} > myrom
Please note that in the MAP file you can notice newly created ROM section.
14.8 Relocating Code and Data in Internal RAM
Since it is not possible to write a variable in ROM, data must be relocated in RAM. Code can be also relocated in RAM. Another reason to relocate code in RAM is that it is twice as fast as in Flash.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
272 Freescale Semiconductor, Inc.
Chapter 14 Linker
To relocate code and data in internal RAM, create a new section using section pragma directive and __declspec directives as shown in the listing below.
Listing 14-16. Using pragma Directives to Define a Section
#pragma section ".myCodeInRAM" data_mode=far_abs
__declspec(section ".myCodeInRAM") struct { unsigned char data0; unsigned char data1; unsigned char data2; unsigned char data3; unsigned char data4; unsigned char data5; unsigned char data6; unsigned char data7;
} CTMData = { 0x82, 0x65, 0x77, 0x32, 0x84, 0x69, 0x83, 0x84 };
__declspec(section ".myCodeInRAM") void funcInROM(int flag);
void funcInROM(int flag){ if (flag > 0)
}
}
{ flag++;
14.8.1 Placing Code and Data in RAM
Placing code and data into RAM is more complicated. As the content in RAM cannot be saved when turning power off, you first need to save the code and data in flash and then make a copy to RAM in runtime.
Following are the steps to relocate code and data in a new RAM segment.
1.
2.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
273
Relocating Code and Data in Internal RAM
14.8.1.1 Create New RAM Segment
As it was made for the new ROM segment, a piece of the user ram memory segment is taken to create a new memory segment called myram.
Edit your LCF as shown in Listing 5.
Listing 14-17. Memory Segment of LCF
MEMORY
{
exception_handlers_p0: org = 0x40000000, len = 0x00001000
pseudo_rom: org = 0x40001000, len = 0x00006800
init: org = 0x40007800, len = 0x00000800
internal_ram: org = 0x40008000, len = 0x00070000
myram: org = 0x40078000, len = 0x00004000
heap : org = 0x4007c000, len = 0x00002000 /* Heap start location
*/
stack : org = 0x4007e000, len = 0x00002000 /* Start location for
Stack */
}
14.8.1.2 Create New RAM Section
The memory segment specifies the intended location in RAM. The code below shows a new section called
.my_ram
which is going to be linked in segment
.myram
but is going to be resident in the Flash memory address calculated by label
___CodeStart
. This label is intended to find the first address available in flash.
In the listing section
.app_text
the linker places in the segment code all the code and then the read only data. After this it sets a label called
__ROM_AT
. Section .data is allocated in the address pointed by this label.
Add the following code to LCF. You can put this code just after Placing data in RAM .
The uninitialized and initialized data are placed in RAM.
Listing 14-18. Add this Code to LCF after Listing D.
___CodeStart = ___RAM_end;
.my_ram :
{
. = ALIGN (0x4);
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
274 Freescale Semiconductor, Inc.
Chapter 14 Linker
___myRAMStart = .;
*(.myCodeInRAM)
___myRAMEnd = .;
. = ALIGN (0x4);
} > myram
14.9 Relocating Code and Data in External MRAM
Many times the internal RAM in the Microcontroller you are using is not enough for the application. For this reason it is needed to use external memories as part of the solution.
The process to relocate code and data in external memories is exactly the same as you did for internal RAM. The only difference is that the external device needs to be communicated by an interface controller.
14.10 Unique LCF Examples
This topic describes the following LCF examples.
•
Configuring Linker File to Several ROM Blocks
•
Place the Library File in the LCF
•
Place Symbols in Specific Memory Location
•
14.10.1 Configuring Linker File to Several ROM Blocks
The following listing is an example to configure linker file to several ROM blocks.
Listing 14-19. ROM IMAGE address = 0x3000
MEMORY{
internal_flash: org = 0x00003000, len = 0x0010000
MyFlash: org = 0x00041000, len = 0x00000008
//org should match the LOAD address
}
SECTIONS{
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
275
Unique LCF Examples
.text {} > internal_flash
.my_flash ALIGN(0x08) LOAD(0x00041000) : {} > MyFlash
}
14.10.2 Place the Library File in the LCF
The following listing is an example to place the library file in the LCF.
Listing 14-20. Placing library file in LCF
GROUP : {
.libcode (VLECODE) LOAD (0x00004000) : {
Runtime.PPCEABI.VS.UC.a (.text)
}
.libconst:
{
Runtime.PPCEABI.VS.UC.a (.rodata)
}
} > lib_flash
NOTE
For small data sections, ctors, dtors section it's not allowed to have different output section name.
14.10.3 Place Symbols in Specific Memory Location
For placing the symbols in specific memory location, user has to define the memory region (say Memory_to_store ) in the lcf file and also define a new section
(say .user_defined_section) then use the same section in the source file to place the symbol.
Listing 14-21. Example for initialized variable
In the source file:
#pragma section <section_qualifier(R,RW)> ".user_defined_section"
__declspec(section ".user_defined_section") int temp = 5;
In the LCF file:
GROUP : {
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
276 Freescale Semiconductor, Inc.
Chapter 14 Linker
.user_defined_section :{}
} > Memory_to_store // Memory_to_store is the memory area where user want to store
Listing 14-22. Example for uninitialized variable
In the source file:
#pragma section ".user_defined_section"".data"
__declspec(section ".user_defined_section") /* We cannot have an uninitialized section name in The //uninitialized section must be paired with initialized section. */
__declspec(section ".user_defined_section") int temp;
In the LCF file:
GROUP : {
.user_defined_section :{}
} > Memory_to_store
14.10.4 How to Relocate Code in RAM
The following listings provide sample code examples to relocate the code in RAM.
Listing 14-23. Sample code in the source file
#pragma section ".myCodeInRAM" code_mode=far_abs
__declspec(section ".myCodeInRAM") int _add(int a , int b); int main(void) {
volatile int i = 0;
volatile int total = 0;
/* Loop forever */
for (;;) {
total = _add(i , i);
i++;
}
}
__declspec(section ".myCodeInRAM") int _add(int a , int b)
{
return a + b;
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
277
Linker Command File Commands
}
Listing 14-24. Sample code in the lcf file
MEMORY
{
.............
/* SRAM: 0x40000000 - 0x4000FFFF */
internal_ram: org = 0x40000000, len = 0x0000D000
myram: org = 0x4000D000, len = 0x00001000
...............
}
..........
GROUP : {
.my_ram (VLECODE) : { //VLECODE- if the code is the generated for VLE mode
*(.myCodeInRAM)
}
} > myram
.................
14.11 Linker Command File Commands
The CodeWarrior linker supports these additional commands listed below:
•
•
•
•
•
•
•
•
14.11.1 AGGRESSIVE_MERGE
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
278 Freescale Semiconductor, Inc.
Chapter 14 Linker
Specifies functions that should be considered for aggressive merging when applying the code merging optimization.
Syntax
DO_NOT_MERGE {
symbol [,
symbol ]* }
14.11.2 AGGRESSIVE_MERGE_FILES
Specifies that all functions in object code files should be considered for aggressive merging when applying the code merging optimization.
Syntax
DO_NOT_MERGE_FILES {
file-name [,
file-name ]* }
14.11.3 AGGRESSIVE_MERGE_SECTIONS
Specifies that all functions in object code sections should be considered for aggressive merging when applying the code merging optimization.
Syntax
AGGRESSIVE_MERGE_SECTIONS {
section-name [,
section-name ]* }
14.11.4 DO_NOT_MERGE
Specifies functions that should not be removed when applying the code merging optimization.
Syntax
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
279
Linker Command File Commands
DO_NOT_MERGE {
symbol [,
symbol ]* }
Remarks
This directive specifies functions that the linker should keep in the output file when applying the code merging optimization even if other functions with identical object code exist.
14.11.5 DO_NOT_MERGE_FILES
Specifies that all functions in a file should not be removed when applying the code merging optimization.
Syntax
DO_NOT_MERGE_FILES {
file-name [,
file-name ]* }
14.11.6 DO_NOT_MERGE_SECTIONS
Specifies that all functions in an object code section should not be removed when applying the code merging optimization.
Syntax
DO_NOT_MERGE_SECTIONS {
section-name [,
section-name ]* }
14.11.7 INIT
Defines the initialization entry point for the application.
Syntax
INIT (FunctionName)
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
280 Freescale Semiconductor, Inc.
Chapter 14 Linker
Remarks
This command is mandatory for assembly application and optional otherwise. It cannot be specified more than once in the prm file. When you specify the INIT command in the prm file, the linker uses the specified function as application entry point. This is either the main routine or a startup routine calling the main routine.
When
INIT
is not specified in the prm
file, the linker looks for a function named
__start
and uses it as the application entry point.
Example
INIT (MyGlobStart) /* Specify a global variable as application entry point.*/
14.11.8 FORCEFILES
Specifies that the contents of object code files must not be deadstripped.
Syntax
FORCEFILES { file-name
[,
file-name ]* }
Remarks
Use
FORCEFILES
to list source files, archives, or archive members that you do not want dead-stripped. All objects in each of the files are included in the linker's output file even if the linker has determined that those objects are not referenced by other objects.
If you only have a few symbols that you do not want deadstripped, use
FORCEACTIVE
.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
281
Linker Command File Commands
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
282 Freescale Semiconductor, Inc.
Chapter 15
C Compiler
This chapter explains the CodeWarrior implementation of the C programming language:
•
•
•
15.1 Extensions to Standard C
The CodeWarrior C compiler adds extra features to the C programming language. These extensions make it easier to port source code from other compilers and offer some programming conveniences. Note that some of these extensions do not conform to the
ISO/IEC 9899-1990 C standard ("C90").
•
Controlling Standard C Conformance
•
•
•
Extensions to the Preprocessor
•
•
Declaring Variables by Address
15.1.1 Controlling Standard C Conformance
The compiler offers settings that verify how closely your source code conforms to the
ISO/IEC 9899-1990 C standard ("C90"). Enable these settings to check for possible errors or improve source code portability.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
283
Extensions to Standard C
Some source code is too difficult or time-consuming to change so that it conforms to the
ISO/IEC standard. In this case, disable some or all of these settings.
Table 15-1 shows how to control the compiler's features for ISO conformance.
Table 15-1. Controlling conformance to the ISO/IEC 9899-1990 C language
To control this option from here...
CodeWarrior IDE source code command line
use this setting
ANSI Strict and ANSI Keywords Only in the C/C++
Language Settings panel
#pragma ANSI_strict#pragma only_std_keywords
-ansi
15.1.2 C++-style Comments
When ANSI strictness is off, the C compiler allows C++-style comments.
shows an example.
Listing 15-1. C++ Comments
a = b; // This is a C++-style comment.
c = d; /* This is a regular C-style comment. */
15.1.3 Unnamed Arguments
When ANSI strictness is off, the C compiler allows unnamed arguments in function
definitions. Listing 15-2 on page 284
shows an example.
Listing 15-2. Unnamed Arguments
void f(int) {} /* OK if ANSI Strict is disabled. */ void f(int i) {} /* Always OK. */
15.1.4 Extensions to the Preprocessor
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
284 Freescale Semiconductor, Inc.
Chapter 15 C Compiler
When ANSI strictness is off, the C compiler allows a
#
to prefix an item that is not a macro argument. It also allows an identifier after an
#endif
directive.
285 and Listing 15-4 on page 285
show examples.
Listing 15-3. Using # in Macro Definitions
#define add1(x) #x #1
/* OK, if ANSI_strict is disabled,
but probably not what you wanted:
add1(abc) creates "abc"#1
*/
#define add2(x) #x "2"
/* Always OK: add2(abc) creates "abc2". */
Listing 15-4. Identifiers After #endif
#ifdef __CWCC__
/* . . . */
#endif __CWCC__ /* OK if ANSI_strict is disabled. */
#ifdef __CWCC__
/* . . . */
#endif /*__CWCC__*/ /* Always OK. */
15.1.5 Non-Standard Keywords
When the ANSI keywords setting is off, the C compiler recognizes non-standard keywords that extend the language.
15.1.6 Declaring Variables by Address
The C compiler lets you explicitly specify the address that contains the value of a variable. For example, the following definition states that the variable
MemErr
contains the contents of the address
0x220
: short MemErr:0x220;
You cannot disable this extension, and it has no corresponding pragma or setting in a panel.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
285
C99 Extensions
15.2 C99 Extensions
The CodeWarrior C compiler accepts the enhancements to the C language specified by the ISO/IEC 9899-1999 standard, commonly referred to as "C99."
•
•
Trailing Commas in Enumerations
•
•
•
•
•
Non-constant Static Data Initialization
•
•
•
•
•
•
Hexadecimal Floating-Point Constants
•
•
Unsuffixed Decimal Literal Values
•
15.2.1 Controlling C99 Extensions
Table 15-2 shows how to control C99 extensions.
source code command line
Table 15-2. Controlling C99 extensions to the C language
To control this option from here...
CodeWarrior IDE
use this setting
Enable C99 Extensions in the C/C++ Language Settings panel
#pragma c99
-c99
15.2.2 Trailing Commas in Enumerations
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
286 Freescale Semiconductor, Inc.
Chapter 15 C Compiler
When the C99 extensions setting is on, the compiler allows a comma after the final item
in a list of enumerations. Listing 15-5 on page 287 shows an example.
Listing 15-5. Trailing comma in enumeration example
enum
{
violet,
blue
green,
yellow,
orange,
red, /* OK: accepted if C99 extensions setting is on. */
};
15.2.3 Compound Literal Values
When the C99 extensions setting is on, the compiler allows literal values of structures
and arrays. Listing 15-6 on page 287 shows an example.
Listing 15-6. Example of a Compound Literal
#pragma c99 on struct my_struct {
int i;
char c[2];
} my_var; my_var = ((struct my_struct) {x + y, 'a', 0});
15.2.4 Designated Initializers
When the C99 extensions setting is on, the compiler allows an extended syntax for specifying which structure or array members to initialize.
shows an example.
Listing 15-7. Example of Designated Initializers
#pragma c99 on struct X {
int a,b,c;
} x = { .c = 3, .a = 1, 2 }; union U {
char a;
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
287
C99 Extensions
long b;
} u = { .b = 1234567 }; int arr1[6] = { 1,2, [4] = 3,4 }; int arr2[6] = { 1, [1 ... 4] = 3,4 }; /* GCC only, not part of C99. */
15.2.5 Predefined Symbol __func__
When the C99 extensions setting is on, the compiler offers the
__func__
predefined
variable. Listing 15-8 on page 288
shows an example.
Listing 15-8. Predefined symbol __func__
void abc(void)
{
puts(__func__); /* Output: "abc" */
}
15.2.6 Implicit Return From main()
When the C99 extensions setting is on, the compiler inserts this statement at the end of a program's main() function if the function does not return a value: return 0;
15.2.7 Non-constant Static Data Initialization
When the C99 extensions setting is on, the compiler allows static variables to be initialized with non-constant expressions.
15.2.8 Variable Argument Macros
When the C99 extensions setting is on, the compiler allows macros to have a variable
number of arguments. Listing 15-9 on page 288
shows an example.
Listing 15-9. Variable argument macros example
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
288 Freescale Semiconductor, Inc.
Chapter 15 C Compiler
#define MYLOG(...) fprintf(myfile, __VA_ARGS__)
#define MYVERSION 1
#define MYNAME "SockSorter" int main(void)
{
MYLOG("%d %s\n", MYVERSION, MYNAME);
/* Expands to: fprintf(myfile, "%d %s\n", 1, "SockSorter"); */
return 0;
}
15.2.9 Extra C99 Keywords
When the C99 extensions setting is on, the compiler recognizes extra keywords and the language features they represent.
lists these keywords.
Table 15-3. Extra C99 Keywords
This keyword or combination of keywords...
_Bool long long restrict inline
_Complex
_Imaginary
represents this language feature
boolean data type integer data type type qualifier function qualifier complex number data type imaginary number data type
15.2.10 C++-Style Comments
When the C99 extensions setting is on, the compiler allows C++-style comments as well as regular C comments. A C++-style comment begins with
// and continues until the end of a source code line.
A C-style comment begins with
/* ends with
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
289
C99 Extensions
*/ and may span more than one line.
15.2.11 C++-Style Digraphs
<:
:>
<%
%>
%:
%:%:
When the C99 extensions setting is on, the compiler recognizes C++-style two-character
combinations that represent single-character punctuation. Table 15-4 lists these digraphs.
This digraph
Table 15-4. C++-Style Digraphs is equivalent to this character
{
}
[
]
#
##
15.2.12 Empty Arrays in Structures
When the C99 extensions setting is on, the compiler allows an empty array to be the last
member in a structure definition. Listing 15-10 on page 290 shows an example.
Listing 15-10. Example of an Empty Array as the Last struct Member
struct {
int r;
char arr[];
} s;
15.2.13 Hexadecimal Floating-Point Constants
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
290 Freescale Semiconductor, Inc.
Chapter 15 C Compiler
Precise representations of constants specified in hexadecimal notation to ensure an accurate constant is generated across compilers and on different hosts. The compiler generates a warning message when the mantissa is more precise than the host floating point format. The compiler generates an error message if the exponent is too wide for the host float format.
Examples:
0x2f.3a2p3
0xEp1f
0x1.8p0L
The standard library supports printing values of type float
in this format using the "
%a
" and "
%A
" specifiers.
15.2.14 Variable-Length Arrays
Variable length arrays are supported within local or function prototype scope, as required
by the ISO/IEC 9899-1999 ("C99") standard. Listing 15-11 on page 291 shows an
example.
Listing 15-11. Example of C99 Variable Length Array usage
#pragma c99 on void f(int n) {
int arr[n];
/* ... */
}
While the example shown in
generates an error message.
Listing 15-12. Bad Example of C99 Variable Length Array usage
#pragma c99 on int n; int arr[n];
// ERROR: variable length array
// types can only be used in local or
// function prototype scope.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
291
C99 Extensions
A variable length array cannot be used in a function template's prototype scope or in a local template typedef
, as shown in
.
Listing 15-13. Bad Example of C99 usage in Function Prototype
#pragma c99 on template<typename T> int f(int n, int A[n][n]);
{
};
// ERROR: variable length arrays
// cannot be used in function template prototypes
// or local template variables
15.2.15 Unsuffixed Decimal Literal Values
Listing 15-14 on page 292 shows an example of specifying decimal literal values without
a suffix to specify the literal's type.
Listing 15-14. Examples of C99 Unsuffixed Constants
#pragma c99 on // Note: ULONG_MAX == 4294967295 sizeof(4294967295) == sizeof(long long) sizeof(4294967295u) == sizeof(unsigned long)
#pragma c99 off sizeof(4294967295) == sizeof(unsigned long) sizeof(4294967295u) == sizeof(unsigned long)
15.2.16 C99 Complex Data Types
The compiler supports the C99 complex and imaginary data types when the C99 extensions option is enabled.
Listing 15-15 on page 292 shows an example.
Listing 15-15. C99 Complex Data Type
#include <complex.h> complex double cd = 1 + 2*I;
NOTE
This feature is currently not available for all targets. Use #if
__has_feature(C99_COMPLEX) to check if this feature is available for your target.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
292 Freescale Semiconductor, Inc.
Chapter 15 C Compiler
15.3 GCC Extensions
The CodeWarrior compiler accepts many of the extensions to the C language that the
GCC (Gnu Compiler Collection) tools allow. Source code that uses these extensions does not conform to the ISO/IEC 9899-1990 C ("C90") standard.
•
•
Initializing Automatic Arrays and Structures
•
•
•
•
•
Void and Function Pointer Arithmetic
•
The __builtin_constant_p() Operator
•
Forward Declarations of Static Arrays
•
Omitted Operands in Conditional Expressions
•
The __builtin_expect() Operator
•
•
•
15.3.1 Controlling GCC Extensions
Table 15-5 shows how to turn GCC extensions on or off.
Table 15-5. Controlling GCC extensions to the C language
To control this option from here...
CodeWarrior IDE source code command line
use this setting
Enable GCC Extensions in the C/C++ Language Settings panel
#pragma gcc_extensions
-gcc_extensions
15.3.2 Initializing Automatic Arrays and Structures
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
293
GCC Extensions
When the GCC extensions setting is on, array and structure variables that are local to a function and have the automatic storage class may be initialized with values that do not
need to be constant. Listing 15-16 on page 294 shows an example.
Listing 15-16. Initializing arrays and structures with non-constant values
void f(int i)
{
int j = i * 10; /* Always OK. */
/* These initializations are only accepted when GCC extensions
* are on. */
struct { int x, y; } s = { i + 1, i + 2 };
int a[2] = { i, i + 2 };
}
15.3.3 The sizeof() Operator
When the GCC extensions setting is on, the sizeof()
operator computes the size of function and void types. In both cases, the sizeof()
operator evaluates to 1. The ISO/IEC
9899-1990 C Standard ("C90") does not specify the size of the void
type and functions.
Listing 15-17 on page 294 shows an example.
Listing 15-17. Using the sizeof() operator with void and function types
int f(int a)
{
return a * 10;
} void g(void)
{
size_t voidsize = sizeof(void); /* voidsize contains 1 */
size_t funcsize = sizeof(f); /* funcsize contains 1 */
}
15.3.4 Statements in Expressions
When the GCC extensions setting is on, expressions in function bodies may contain statements and definitions. To use a statement or declaration in an expression, enclose it within braces. The last item in the brace-enclosed expression gives the expression its value.
Listing 15-18 on page 294 shows an example.
Listing 15-18. Using statements and definitions in expressions
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
294 Freescale Semiconductor, Inc.
Chapter 15 C Compiler
#define POW2(n) ({ int i,r; for(r=1,i=n; i>0; --i) r *= 2; r;}) int main()
{
return POW2(4);
}
15.3.5 Redefining Macros
When the GCC extensions setting is on, macros may be redefined with the
#define directive without first undefining them with the
#undef
directive. Listing 15-19 on page
Listing 15-19. Redefining a macro without undefining first
#define SOCK_MAXCOLOR 100
#undef SOCK_MAXCOLOR
#define SOCK_MAXCOLOR 200 /* OK: this macro is previously undefined. */
#define SOCK_MAXCOLOR 300
15.3.6 The typeof() Operator
When the GCC extensions setting is on, the compiler recognizes the typeof()
operator.
This compile-time operator returns the type of an expression. You may use the value returned by this operator in any statement or expression where the compiler expects you to specify a type. The compiler evaluates this operator at compile time. The
__typeof()__ operator is the same as this operator.
shows an example.
Listing 15-20. Using the typeof() operator
int *ip;
/* Variables iptr and jptr have the same type. */ typeof(ip) iptr; int *jptr;
/* Variables i and j have the same type. */ typeof(*ip) i; int j;
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
295
GCC Extensions
15.3.7 Void and Function Pointer Arithmetic
The ISO/IEC 9899-1990 C Standard does not accept arithmetic expressions that use pointers to void
or functions. With GCC extensions on, the compiler accepts arithmetic manipulation of pointers to void
and functions.
15.3.8 The __builtin_constant_p() Operator
When the GCC extensions setting is on, the compiler recognizes the
__builtin_constant_p() operator. This compile-time operator takes a single argument and returns 1 if the argument is a constant expression or 0 if it is not.
15.3.9 Forward Declarations of Static Arrays
When the GCC extensions setting is on, the compiler will not issue an error when you declare a static array without specifying the number of elements in the array if you later declare the array completely.
Listing 15-21 on page 296 shows an example.
Listing 15-21. Forward declaration of an empty array
static int a[]; /* Allowed only when GCC extensions are on. */
/* ... */ static int a[10]; /* Complete declaration. */
15.3.10 Omitted Operands in Conditional Expressions
When the GCC extensions setting is on, you may skip the second expression in a conditional expression. The default value for this expression is the first expression.
Listing 15-22 on page 296 shows an example.
Listing 15-22. Using the shorter form of the conditional expression
void f(int i, int j)
{
int a = i ? i : j;
int b = i ?: j; /* Equivalent to int b = i ? i : j; */
/* Variables a and b are both assigned the same value. */
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
296 Freescale Semiconductor, Inc.
Chapter 15 C Compiler
}
15.3.11 The __builtin_expect() Operator
When the GCC extensions setting is on, the compiler recognizes the
__builtin_expect() operator. Use this compile-time operator in an if
or while
statement to specify to the compiler how to generate instructions for branch prediction.
This compile-time operator takes two arguments:
• the first argument must be an integral expression
• the second argument must be a literal value
The second argument is the most likely result of the first argument.
Listing 15-23. Example for __builtin_expect() operator
void search(int *array, int size, int key)
{
int i;
for (i = 0; i < size; ++i)
{
/* We expect to find the key rarely. */
if (__builtin_expect(array[i] == key, 0))
{
rescue(i);
}
}
}
15.3.12 Void Return Statements
When the GCC extensions setting is on, the compiler allows you to place expressions of type void
in a return
statement. Listing 15-24 on page 297 shows an example.
Listing 15-24. Returning void
void f(int a)
{
/* ... */
return; /* Always OK. */
} void g(int b)
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
297
GCC Extensions
{
/* ... */
return f(b); /* Allowed when GCC extensions are on. */
}
15.3.13 Minimum and Maximum Operators
When the GCC extensions setting is on, the compiler recognizes built-in minimum (
<?
) and maximum (
>?
) operators.
Listing 15-25. Example of minimum and maximum operators
int a = 1 <? 2; // 1 is assigned to a.
int b = 1 >? 2; // 2 is assigned to b.
15.3.14 Local Labels
When the GCC extensions setting is on, the compiler allows labels limited to a block's scope. A label declared with the
__label__
keyword is visible only within the scope of its enclosing block.
Listing 15-26 on page 298 shows an example.
Listing 15-26. Example of using local labels
void f(int i)
{
if (i >= 0)
{
__label__ again; /* First again. */
if (--i > 0)
goto again; /* Jumps to first again. */
}
else
{
__label__ again; /* Second again. */
if (++i < 0)
goto again; /* Jumps to second again. */
}
}
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
298 Freescale Semiconductor, Inc.
Chapter 16
C++ Compiler
This chapter explains the CodeWarrior implementation of the C++ programming language:
•
•
•
Implementation-Defined Behavior
•
16.1 C++ Compiler Performance
Some options affect the C++ compiler's performance. This section explains how to improve compile times when translating C++ source code:
•
•
16.1.1 Precompiling C++ Source Code
The CodeWarrior C++ compiler has these requirements for precompiling source code:
• C source code may not include precompiled C++ header files and C++ source code may not include precompiled C header files.
• C++ source code can contain inline functions
• C++ source code may contain constant variable declarations
• A C++ source code file that will be automatically precompiled must have a
.pch++
file name extension.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
299
Extensions to Standard C++
16.1.2 Using the Instance Manager
The instance manager reduces compile time by generating a single instance of some kinds of functions:
• template functions
• functions declared with the inline
qualifier that the compiler was not able to insert in line
The instance manager reduces the size of object code and debug information but does not affect the linker's output file size, though, since the compiler is effectively doing the same task as the linker in this mode.
Table 16-1 shows how to control the C++ instance manager.
Table 16-1. Controlling the C++ instance manager
To control this option from here...
source code command line
use this setting
#pragma instmgr_file
-instmgr
16.2 Extensions to Standard C++
The CodeWarrior C++ compiler has features and capabilities that are not described in the
ISO/IEC 14882:2003 C++ standard:
•
__PRETTY_FUNCTION__ Identifier
•
Standard and Non-Standard Template Parsing
16.2.1 __PRETTY_FUNCTION__ Identifier
The
__PRETTY_FUNCTION__
predefined identifier represents the qualified (unmangled) C++ name of the function being compiled.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
300 Freescale Semiconductor, Inc.
Chapter 16 C++ Compiler
16.2.2 Standard and Non-Standard Template Parsing
CodeWarrior C++ has options to specify how strictly template declarations and instantiations are translated. When using its strict template parser, the compiler expects the typename
and template
keywords to qualify names, preventing the same name in different scopes or overloaded declarations from being inadvertently used. When using its regular template parser, the compiler makes guesses about names in templates, but may guess incorrectly about which name to use.
A qualified name that refers to a type and that depends on a template parameter must begin with typename
(ISO/IEC 14882:2003 C++, §14.6).
Listing 16-1 on page 301 shows
an example.
Listing 16-1. Using the typename
keyword
template <typename T> void f()
{
T::name *ptr; // ERROR: an attempt to multiply T::name by ptr
typename T::name *ptr; // OK
}
The compiler requires the template
keyword at the end of "
.
" and "
->
" operators, and for qualified identifiers that depend on a template parameter.
shows an example.
Listing 16-2. Using the template
keyword
template <typename T> void f(T* ptr)
{
ptr->f<int>(); // ERROR: f is less than int
ptr->template f<int>(); // OK
}
Names referred to inside a template declaration that are not dependent on the template declaration (that do not rely on template arguments) must be declared before the template's declaration. These names are bound to the template declaration at the point where the template is defined. Bindings are not affected by definitions that are in scope at
the point of instantiation. Listing 16-3 on page 301 shows an example.
Listing 16-3. Binding non-dependent identifiers
void f(char); template <typename T> void tmpl_func()
{
f(1); // Uses f(char); f(int), below, is not defined yet.
g(); // ERROR: g() is not defined yet.
} void g(); void f(int);
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
301
Extensions to Standard C++
Names of template arguments that are dependent in base classes must be explicitly
qualified (ISO/IEC 14882:2003 C++, §14.6.2). See Listing 16-4 on page 302 .
Listing 16-4. Qualifying template arguments in base classes
template <typename T> struct Base
{
void f();
} template <typename T> struct Derive: Base<T>
{
void g()
{
f(); // ERROR: Base<T>::f() is not visible.
Base<T>::f(); // OK
}
}
When a template contains a function call in which at least one of the function's arguments is type-dependent, the compiler uses the name of the function in the context of the template definition (ISO/IEC 14882:2003 C++, §14.6.2.2) and the context of its instantiation (ISO/IEC 14882:2003 C++, §14.6.4.2).
shows an example.
Listing 16-5. Function call with type-dependent argument
void f(char); template <typename T> void type_dep_func()
{
f(1); // Uses f(char), above; f(int) is not declared yet.
f(T()); // f() called with a type-dependent argument.
} void f(int); struct A{}; void f(A); int main()
{
type_dep_func<int>(); // Calls f(char) twice.
type_dep_func<A>(); // Calls f(char) and f(A);
return 0;
}
The compiler only uses external names to look up type-dependent arguments in function calls. See
Listing 16-6. Function call with type-dependent argument and external names
static void f(int); // f() is internal.
template <typename T> void type_dep_fun_ext()
{
f(T()); // f() called with a type-dependent argument.
} int main()
{
type_dep_fun_ext<int>(); // ERROR: f(int) must be external.
}
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
302 Freescale Semiconductor, Inc.
Chapter 16 C++ Compiler
The compiler does not allow expressions in inline assembly statements that depend on template parameters. See
Listing 16-7. Assembly statements cannot depend on template arguments
template <typename T> void asm_tmpl()
{
asm { move #sizeof(T), D0 ); // ERROR: Not supported.
}
The compiler also supports the address of template-id rules. See
Listing 16-8. Address of Template-id Supported
template <typename T> void funcA(T) {} template <typename T> void funcB(T) {}
...
funcA{ &funcB<int> ); // now accepted
16.3 Implementation-Defined Behavior
Annex A of the ISO/IEC 14882:2003 C++ Standard lists compiler behaviors that are beyond the scope of the standard, but which must be documented for a compiler implementation. This annex also lists minimum guidelines for these behaviors, although a conforming compiler is not required to meet these minimums.
The CodeWarrior C++ compiler has these implementation quantities listed in
, based on the ISO/IEC 14882:2003 C++ Standard, Annex A.
NOTE
The term unlimited in
means that a behavior is limited only by the processing speed or memory capacity of the computer on which the CodeWarrior C++ compiler is running.
Table 16-2. Implementation Quantities for C/C++ Compiler (ISO/IEC 14882:2003 C++, §A)
Behavior
Nesting levels of compound statements, iteration control structures, and selection control structures
Nesting levels of conditional inclusion
Pointer, array, and function declarators
(in any combination) modifying an arithmetic, structure, union, or incomplete type in a declaration
Nesting levels of parenthesized expressions within a full expression
256
256
256
256
Standard Minimum Guideline
Table continues on the next page...
Unlimited
256
Unlimited
Unlimited
CodeWarrior Limit
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
303
Implementation-Defined Behavior
Table 16-2. Implementation Quantities for C/C++ Compiler (ISO/IEC 14882:2003 C++, §A)
(continued)
CodeWarrior Limit Behavior Standard Minimum Guideline
Number of initial characters in an internal identifier or macro name
Number of initial characters in an external identifier
1024
1024
External identifiers in one translation unit 65536
Identifiers with block scope declared in one block
1024
Macro identifiers simultaneously defined in one translation unit
65536
Parameters in one function definition
Arguments in one function call
Parameters in one macro definition
256
256
256
Arguments in one macro invocation
Characters in one logical source line
Characters in a character string literal or wide string literal (after concatenation)
Size of an object
Nesting levels for # include files
256
65536
65536
262144
256
16384
Case labels for a switch statement
(excluding those for any nested switch statements)
Data members in a single class, structure, or union
Enumeration constants in a single enumeration
Levels of nested class, structure, or union definitions in a single structdeclaration-list
Functions registered by atexit()
16384
4096
256
Direct and indirect base classes
Direct base classes for a single class
Members declared in a single class
Final overriding virtual functions in a class, accessible or not
Direct and indirect virtual bases of a class
Static members of a class
Friend declarations in a class
Access control declarations in a class
Member initializers in a constructor definition
Scope qualifications of one identifier
32
16384
1024
4096
16384
1024
1024
4096
4096
6144
256
Table continues on the next page...
Unlimited
Unlimited
Unlimited
Unlimited
Unlimited
Unlimited
Unlimited
256
256
Unlimited
Unlimited
2 GB
256
Unlimited
Unlimited
Unlimited
Unlimited
64
Unlimited
Unlimited
Unlimited
Unlimited
Unlimited
Unlimited
Unlimited
Unlimited
Unlimited
Unlimited
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
304 Freescale Semiconductor, Inc.
Chapter 16 C++ Compiler
Table 16-2. Implementation Quantities for C/C++ Compiler (ISO/IEC 14882:2003 C++, §A)
(continued)
Behavior
Nested external specifications
Template arguments in a template declaration
Recursively nested template instantiations
Handlers per try block
Throw specifications on a single function declaration
1024
1024
17
256
256
Standard Minimum Guideline
Unlimited
Unlimited
CodeWarrior Limit
64 (adjustable upto 30000 using
#pragma template_depth(<n>))
Unlimited
Unlimited
16.4 GCC Extensions
The CodeWarrior C++ compiler recognizes some extensions to the ISO/IEC 14882:2003
C++ standard that are also recognized by the GCC (GNU Compiler Collection) C++ compiler.
The compiler allows the use of the :: operator, of the form class
::
member
, in a class declaration.
Listing 16-9. Using the :: operator in class declarations
class MyClass {
int MyClass::getval();
};
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
305
GCC Extensions
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
306 Freescale Semiconductor, Inc.
Chapter 17
Precompiling
Each time you invoke the CodeWarrior compiler to translate a source code file, it
preprocesses
the file to prepare its contents for translation. Preprocessing tasks include expanding macros, removing comments, and including header files. If many source code files include the same large or complicated header file, the compiler must preprocess it each time it is included. Repeatedly preprocessing this header file can take up a large portion of the time that the compiler operates.
To shorten the time spent compiling a project, CodeWarrior compilers can precompile a file once instead of preprocessing it every time it is included in project source files. When it precompiles a header file, the compiler converts the file's contents into internal data structures, then writes this internal data to a precompiled file. Conceptually, precompiling records the compiler's state after the preprocessing step and before the translation step of the compilation process.
This section shows you how to use and create precompiled files:
•
•
•
17.1 What Can be Precompiled
A file to be precompiled does not have to be a header file (
.h
or
.hpp
files, for example), but it must meet these requirements:
• The file must be a source code file in text format.
You cannot precompile libraries or other binary files.
•
• The file must not contain any statements that generate data or executable code.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
307
Using a Precompiled File
However, the file may define static data.
• Precompiled header files for different IDE build targets are not interchangeable.
17.2 Using a Precompiled File
To use a precompiled file, simply include it in your source code files like you would any other header file:
• A source file may include only one precompiled file.
• A file may not define any functions, variables or types before including a precompiled file.
• Typically, a source code file includes a precompiled file before anything else (except comments).
shows an example.
Listing 17-1. Using a precompiled file
/* sock_main.c */
#include "sock.mch" /* Precompiled header file. */
#include "wool.h /* Regular header file. */
/* ... */
17.3 Creating a Precompiled File
This section shows how to create and manage precompiled files:
•
Precompiling a File in the CodeWarrior IDE
•
Precompiling a File on the Command Line
•
Updating a Precompiled File Automatically
•
Preprocessor Scope in Precompiled Files
17.3.1 Precompiling a File in the CodeWarrior IDE
To precompile a file in the CodeWarrior IDE, follow these steps:
1. Start the CodeWarrior IDE.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
308 Freescale Semiconductor, Inc.
2. Open or create a project.
You may now include the new precompiled file in source code files.
Chapter 17 Precompiling
17.3.2 Precompiling a File on the Command Line
To precompile a file on the command line, follow these steps:
1. Start a command line shell.
2. Issue this command
ccaiop h_file -precompile p_file where ccaiop is the name of the CodeWarrior compiler tool, h_file is the name of the header to precompile, and p_file is the name of the resulting precompiled file.
17.3.3 Updating a Precompiled File Automatically
Use the CodeWarrior IDE's project manager to update a precompiled header automatically. The IDE creates a precompiled file from a source code file during a compile, update, or make operation if the source code file meets these criteria:
• The text file name ends with
.pch
.
• The file is in a project's build target.
• The file uses the precompile_target
pragma.
• The file, or files it depends on, have been modified.
.
The IDE uses the build target's settings to preprocess and precompile files.
17.3.4 Preprocessor Scope in Precompiled Files
When precompiling a header file, the compiler preprocesses the file too. In other words, a precompiled file is preprocessed in the context of its precompilation, not in the context of its later compilation.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
309
Creating a Precompiled File
The preprocessor also tracks macros used to guard
#include
files to reduce parsing time. If a file's contents are surrounded with
#ifndef MYHEADER_H
#define MYHEADER_H
/* file contents */
#endif the compiler will not load the file twice, saving some time in the process.
Pragma settings inside a precompiled file affect only the source code within that file. The pragma settings for an item declared in a precompiled file (such as data or a function) are saved then restored when the precompiled header file is included.
For example, the source code in
Listing 17-2 on page 310 specifies that the variable
xxx is a far
variable.
Listing 17-2. Pragma Settings in a Precompiled Header
/* my_pch.pch */
/* Generate a precompiled header named pch.mch. */
#pragma precompile_target "my_pch.mch"
#pragma far_data on extern int xxx;
The source code in Listing 17-3 on page 310 includes the precompiled version of Listing
.
Listing 17-3. Pragma Settings in an Included Precompiled File
/* test.c */
/* Far data is disabled. */
#pragma far_data off
/* This precompiled file sets far_data on. */
#include "my_pch.mch"
/* far_data is still off but xxx is still a far variable. */
The pragma setting in the precompiled file is active within the precompiled file, even though the source file including the precompiled file has a different setting.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
310 Freescale Semiconductor, Inc.
Chapter 18
Intermediate Optimizations
After it translates a program's source code into its intermediate representation, the compiler optionally applies optimizations that reduce the program's size, improve its execution speed, or both. The topics in this chapter explains these optimizations and how to apply them:
•
•
•
18.1 Interprocedural Analysis
Most compiler optimizations are applied only within a function. The compiler analyzes a function's flow of execution and how the function uses variables. It uses this information to find shortcuts in execution and reduce the number of registers and memory that the function uses. These optimizations are useful and effective but are limited to the scope of a function.
The CodeWarrior compiler has a special optimization that it applies at a greater scope.
Widening the scope of an optimization offers the potential to greatly improve performance and reduce memory use. Interprocedural analysis examines the flow of execution and data within entire files and programs to improve performance and reduce size.
•
Invoking Interprocedural Analysis
•
•
•
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
311
Interprocedural Analysis
18.1.1 Invoking Interprocedural Analysis
Table 18-1 explains how to control interprocedural analysis.
Table 18-1. Controlling interprocedural analysis
Turn control this option from here...
CodeWarrior IDE source code command line
use this setting
Choose an item the
IPA option .
#pragma ipa file | on | function | off
-ipa file | function | off
18.1.2 Function-Level Optimization
Interprocedural analysis may be disabled by setting it to either off
or function
. If IPA is disabled, the compiler generates instructions and data as it reads and analyzes each function. This setting is equivalent to the "no deferred codegen" mode of older compilers.
18.1.3 File-Level Optimization
When interprocedural analysis is set to optimize at the file level, the compiler reads and analyzes an entire file before generating instructions and data.
At this level, the compiler generates more efficient code for inline function calls and C++ exception handling than when interprocedural analysis is off. The compiler is also able to increase character string reuse and pooling, reducing the size of object code. This is equivalent to the " deferred inlining" and "deferred codegen" options of older compilers.
The compiler also safely removes static functions and variables that are not referred to within the file, which reduces the amount of object code that the linker must process, resulting in better linker performance.
18.1.4 Program-Level Optimization
When interprocedural analysis is set to optimize at the program level, the compiler reads and analyzes all files in a program before generating instructions and data.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
312 Freescale Semiconductor, Inc.
Chapter 18 Intermediate Optimizations
At this level of interprocedural analysis, the compiler generates the most efficient instructions and data for inline function calls and C++ exception handling compared to other levels. The compiler is also able to increase character string reuse and pooling, reducing the size of object code.
18.1.5 Program-Level Requirements
Program-level interprocedural analysis imposes some requirements and limitations on the source code files that the compiler translates:
•
Dependencies Among Source Files
•
Function and Top-level Variable Declarations
•
•
Unnamed Structures and Enumerations in C
18.1.5.1 Dependencies Among Source Files
A change to even a single source file in a program still requires that the compiler read and analyze all files in the program, even those files that are not dependent on the changed file. This requirement significantly increases compile time.
18.1.5.2 Function and Top-level Variable Declarations
Because the compiler treats all files that compose a program as if they were a single, large source file, you must make sure all non-static declarations for variables or functions with the same name are identical. See
for an example of declarations that prevent the compiler from applying program-level analysis.
Listing 18-2 on page 314 fixes this problem by renaming the conflicting symbols.
Listing 18-1. Declaration conflicts in program-level interprocedural analysis
/* file1.c */ extern int i; extern int f(); int main(void)
{
return i + f();
}
/* file2.c */
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
313
Interprocedural Analysis
short i; /* Conflict with variable i in file1.c. */ extern void f(); /* Conflict with function f() in file1.c */
Listing 18-2. Fixing declaration conflicts for program-level interprocedural analysis
/* file1.c */ extern int i1; extern int f1(); int main(void)
{
return i1 + f1();
}
/* file2.c */ short i2; extern void f2();
18.1.5.3 Type Definitions
Because the compiler examines all source files for a program, make sure all definitions for a type are the same. See
show suggested solutions.
Listing 18-3. Type definitions conflicts in program-level interprocedural analysis
/* fileA.c */ struct a_rec { int i, j; }; a_rec a;
/* fileB.c */ struct a_rec { char c; }; /* Conflict with a_rec in fileA.c */ a_rec b;
Listing 18-4. Fixing type definitions conflicts in C
/* fileA.c */ struct a1_rec { int i, j; }; a1_rec a;
/* fileB.c */ struct a2_rec { char c; }; a2_rec b;
Listing 18-5. Fixing type definitions conflicts in C++
/* fileA.c */ namespace { struct a_rec { int i, j; }; } a_rec a;
/* fileB.c */ namespace { struct a_rec { char c; }; } a_rec b;
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
314 Freescale Semiconductor, Inc.
Chapter 18 Intermediate Optimizations
18.1.5.4 Unnamed Structures and Enumerations in C
The C language allows anonymous struct
and enum
definitions in type definitions. Using such definitions prevents the compiler from properly applying program-level interprocedural analysis. Make sure to give names to structures and enumerations in type
definitions. Listing 18-6 on page 315
shows an example of unnamed structures and enumerations and
shows a suggested solution.
Listing 18-6. Unnamed structures and enumerations in C
/* In C, the types x_rec and y_enum each represent a structure
and an enumeration with no name.
In C++ these same statements define a type x_rec and y_enum,
a structure named x_rec and an enumeration named y_enum.
*/ typedef struct { int a, b, c; } x_rec; typedef enum { Y_FIRST, Y_SECOND, Y_THIRD } y_enum;
Listing 18-7. Naming structures and enumerations in C
typedef struct x_rec { int a, b, c; } x_rec; typedef enum y_enum { Y_FIRST, Y_SECOND, Y_THIRD } y_enum;
18.2 Intermediate Optimizations
After it translates a function into its intermediate representation, the compiler may optionally apply some optimizations. The result of these optimizations on the intermediate representation will either reduce the size of the executable code, improve the executable code's execution speed, or both.
•
•
•
Common Subexpression Elimination
•
•
•
•
•
•
18.2.1 Dead Code Elimination
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
315
Intermediate Optimizations
The dead code elimination optimization removes expressions that are not accessible or are not referred to. This optimization reduces size and increases execution speed.
Table 18-2 explains how to control the optimization for dead code elimination.
Table 18-2. Controlling dead code elimination
Turn control this option from here...
CodeWarrior IDE source code command line
use this setting
#pragma opt_dead_code on | off | reset
-opt [no]deadcode
In
, the call to func1()
will never execute because the if
statement that it is associated with will never be true. Consequently, the compiler can safely eliminate the call to func1()
Figure 18-1. Before dead code elimination
void func_from(void)
{
if (0)
{
func1();
}
func2();
}
Figure 18-2. After dead code elimination
void func_to(void)
{
func2();
}
18.2.2 Expression Simplification
The expression simplification optimization attempts to replace arithmetic expressions with simpler expressions. Additionally, the compiler also looks for operations in expressions that can be avoided completely without affecting the final outcome of the expression. This optimization reduces size and increases speed.
Table 18-3 explains how to control the optimization for expression simplification.
Table 18-3. Controlling expression simplification
Turn control this option from here...
CodeWarrior IDE
Table continues on the next page...
use this setting
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
316 Freescale Semiconductor, Inc.
Chapter 18 Intermediate Optimizations
Table 18-3. Controlling expression simplification (continued)
source code
Turn control this option from here...
command line
use this setting
There is no pragma to control this optimization.
-opt level=1 , -opt level=2 , -opt level=3 , -opt level=4
For example, Figure 18-3 contains a few assignments to some arithmetic expressions:
• addition to zero
• multiplication by a power of 2
• subtraction of a value from itself
• arithmetic expression with two or more literal values
Figure 18-3. Before expression simplification
void func_from(int* result1, int* result2, int* result3, int* result4, int x)
{
*result1 = x + 0;
*result2 = x * 2;
*result3 = x - x;
*result4 = 1 + x + 4;
}
Figure 18-4 shows source code that is equivalent to expression simplification. The
compiler has modified these assignments to:
• remove the addition to zero
• replace the multiplication of a power of 2 with bit-shift operation
• replace a subtraction of x from itself with 0
• consolidate the additions of 1 and 4 into 5
Figure 18-4. After expression simplification
void func_to(int* result1, int* result2, int* result3, int* result4, int x)
{
*result1 = x;
*result2 = x << 1;
*result3 = 0;
*result4 = 5 + x;
}
18.2.3 Common Subexpression Elimination
Common subexpression elimination replaces multiple instances of the same expression with a single instance. This optimization reduces size and increases execution speed.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
317
Intermediate Optimizations
Table 18-4 explains how to control the optimization for common subexpression
elimination.
Table 18-4. Controlling common subexpression elimination
Turn control this option from here...
CodeWarrior IDE source code command line
use this setting
#pragma opt_common_subs on | off | reset
-opt [no]cse
For example, in
, the subexpression x * y
occurs twice.
Figure 18-5. Before common subexepression elimination
void func_from(int* vec, int size, int x, int y, int value)
{
if (x * y < size)
{
vec[x * y - 1] = value;
}
}
Figure 18-6 shows equivalent source code after the compiler applies common
subexpression elimination. The compiler generates instructions to compute x * y
and store it in a hidden, temporary variable. The compiler then replaces each instance of the subexpression with this variable.
Figure 18-6. After common subexpression elimination
void func_to(int* vec, int size, int x, int y, int value)
{
int temp = x * y;
if (temp < size)
{
vec[temp - 1] = value;
}
}
18.2.4 Copy Propagation
Copy propagation replaces variables with their original values if the variables do not change. This optimization reduces runtime stack size and improves execution speed.
Table 18-5 explains how to control the optimization for copy propagation.
Table 18-5. Controlling copy propagation
Turn control this option from here...
CodeWarrior IDE
Table continues on the next page...
use this setting
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
318 Freescale Semiconductor, Inc.
Chapter 18 Intermediate Optimizations
Table 18-5. Controlling copy propagation (continued)
source code
Turn control this option from here...
command line
use this setting
#pragma opt_propagation on | off | reset
-opt [no]prop[agation]
For example, in
, the variable j
is assigned the value of x
. But j
's value is never changed, so the compiler replaces later instances of j
with x
, as shown in
.
By propagating x
, the compiler is able to reduce the number of registers it uses to hold variable values, allowing more variables to be stored in registers instead of slower memory. Also, this optimization reduces the amount of stack memory used during function calls.
Figure 18-7. Before copy propagation
void func_from(int* a, int x)
{
int i;
int j;
j = x;
for (i = 0; i < j; i++)
{
a[i] = j;
}
}
Figure 18-8. After copy propagation
void func_to(int* a, int x)
{
int i;
int j;
j = x;
for (i = 0; i < x; i++)
{
a[i] = x;
}
}
18.2.5 Dead Store Elimination
Dead store elimination removes unused assignment statements. This optimization reduces size and improves speed.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
319
Intermediate Optimizations
Table 18-6 explains how to control the optimization for dead store elimination.
Table 18-6. Controlling dead store elimination
Turn control this option from here...
CodeWarrior IDE source code command line
use this setting
#pragma opt_dead_assignments on | off | reset
-opt [no]deadstore
For example, in
the variable x
is first assigned the value of y * y
. However, this result is not used before x
is assigned the result returned by a call to getresult()
.
In
Figure 18-10 the compiler can safely remove the first assignment to
x
since the result of this assignment is never used.
Figure 18-9. Before dead store elimination
void func_from(int x, int y)
{
x = y * y;
otherfunc1(y);
x = getresult();
otherfunc2(y);
}
Figure 18-10. After dead store elimination
void func_to(int x, int y)
{
otherfunc1(y);
x = getresult();
otherfunc2(y);
}
18.2.6 Live Range Splitting
Live range splitting attempts to reduce the number of variables used in a function. This optimization reduces a function's runtime stack size, requiring fewer instructions to invoke the function. This optimization potentially improves execution speed.
Table 18-7 explains how to control the optimization for live range splitting.
Table 18-7. Controlling live range splitting
Turn control this option from here...
CodeWarrior IDE source code command line
use this setting
There is no pragma to control this optimization.
-opt level=3 , -opt level=4
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
320 Freescale Semiconductor, Inc.
Chapter 18 Intermediate Optimizations
For example, in
a
, b
, and c
, are defined. Although each variable is eventually used, each of their uses is exclusive to the others. In other words, a is not referred to in the same expressions as b
or c
, b
is not referred to with a
or c
, and c
is not used with a
or b
.
In
Figure 18-12 , the compiler has replaced
a
, b
, and c
, with a single variable. This optimization reduces the number of registers that the object code uses to store variables, allowing more variables to be stored in registers instead of slower memory. This optimization also reduces a function's stack memory.
Figure 18-11. Before live range splitting
void func_from(int x, int y)
{
int a;
int b;
int c;
a = x * y;
otherfunc(a);
b = x + y;
otherfunc(b);
c = x - y;
otherfunc(c);
}
Figure 18-12. After live range splitting
void func_to(int x, int y)
{
int a_b_or_c;
a_b_or_c = x * y;
otherfunc(temp);
a_b_or_c = x + y;
otherfunc(temp);
a_b_or_c = x - y;
otherfunc(temp);
}
18.2.7 Loop-Invariant Code Motion
Loop-invariant code motion moves expressions out of a loop if the expressions are not affected by the loop or the loop does not affect the expression. This optimization improves execution speed.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
321
Intermediate Optimizations
Table 18-8 explains how to control the optimization for loop-invariant code motion.
Table 18-8. Controlling loop-invariant code motion
Turn control this option from here...
CodeWarrior IDE source code command line
use this setting
#pragma opt_loop_invariants on | off | reset
-opt [no]loop[invariants]
For example, in
Figure 18-13 , the assignment to the variable
circ
does not refer to the counter variable of the for
loop, i
. But the assignment to circ
will be executed at each loop iteration.
shows source code that is equivalent to how the compiler would rearrange instructions after applying this optimization. The compiler has moved the assignment to circ
outside the for
loop so that it is only executed once instead of each time the for
loop iterates.
Figure 18-13. Before loop-invariant code motion
void func_from(float* vec, int max, float val)
{
float circ;
int i;
for (i = 0; i < max; ++i)
{
circ = val * 2 * PI;
vec[i] = circ;
}
}
Figure 18-14. After loop-invariant code motion
void func_to(float* vec, int max, float val)
{
float circ;
int i;
circ = val * 2 * PI;
for (i = 0; i < max; ++i)
{
vec[i] = circ;
}
}
18.2.8 Strength Reduction
Strength reduction attempts to replace slower multiplication operations with faster addition operations. This optimization improves execution speed but increases code size.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
322 Freescale Semiconductor, Inc.
Chapter 18 Intermediate Optimizations
Table 18-9 explains how to control the optimization for strength reduction.
Table 18-9. Controlling strength reduction
Turn control this option from here...
CodeWarrior IDE source code
use this setting
command line
#pragma opt_strength_reduction on | off | reset
-opt [no]strength
For example, in
Figure 18-15 , the assignment to elements of the
vec
array use a multiplication operation that refers to the for
loop's counter variable, i
.
In
Figure 18-16 , the compiler has replaced the multiplication operation with a hidden
variable that is increased by an equivalent addition operation. Processors execute addition operations faster than multiplication operations.
Figure 18-15. Before strength reduction
void func_from(int* vec, int max, int fac)
{
int i;
for (i = 0; i < max; ++i)
{
vec[i] = fac * i;
}
}
Figure 18-16. After strength reduction
void func_to(int* vec, int max, int fac)
{
int i;
int hidden_strength_red;
hidden_strength_red = 0;
for (i = 0; i < max; ++i)
{
vec[i] = hidden_strength_red;
hidden_strength_red = hidden_strength_red + fac;
}
}
18.2.9 Loop Unrolling
Loop unrolling inserts extra copies of a loop's body in a loop to reduce processor time executing a loop's overhead instructions for each iteration of the loop body. In other words, this optimization attempts to reduce the ratio of time that the processor executes a loop's completion test and branching instructions compared to the time the processor executes the loop's body. This optimization improves execution speed but increases code size.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
323
Inlining
explains how to control the optimization for loop unrolling.
Table 18-10. Controlling loop unrolling
Turn control this option from here...
CodeWarrior IDE source code command line
use this setting
#pragma opt_unroll_loops on | off | reset
-opt level=3 , -opt level=4
For example, in
for
loop's body is a single call to a function, otherfunc()
. For each time the loop's completion test executes for (i = 0; i < MAX; ++i) the function executes the loop body only once.
In
Figure 18-18 , the compiler has inserted another copy of the loop body and rearranged
the loop to ensure that variable i
is incremented properly. With this arrangement, the loop's completion test executes once for every 2 times that the loop body executes.
Figure 18-17. Before loop unrolling
const int MAX = 100; void func_from(int* vec)
{
int i;
for (i = 0; i < MAX; ++i)
{
otherfunc(vec[i]);
}
}
Figure 18-18. After loop unrolling
const int MAX = 100; void func_to(int* vec)
{
int i;
for (i = 0; i < MAX;)
{
otherfunc(vec[i]);
++i;
otherfunc(vec[i]);
++i;
}
}
18.3 Inlining
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
324 Freescale Semiconductor, Inc.
Chapter 18 Intermediate Optimizations
Inlining
replaces instructions that call a function and return from it with the actual instructions of the function being called. Inlining functions makes your program faster because it executes the function code immediately without the overhead of a function call and return. However, inlining can also make your program larger because the compiler may insert the function's instructions many times throughout your program.
The rest of this section explains how to specify which functions to inline and how the compiler performs the inlining:
•
Choosing Which Functions to Inline
•
18.3.1 Choosing Which Functions to Inline
The compiler offers several methods to specify which functions are eligible for inlining.
To specify that a function is eligible to be inlined, precede its definition with the inline
,
__inline__
, or
__inline
keyword. To allow these keywords in C source code, turn off
ANSI Keywords Only in the CodeWarrior IDE's C/C++ Language panel or turn off the only_std_keywords
pragma in your source code.
To verify that an eligible function has been inlined or not, use the Non-Inlined
Functions option in the IDE's Warnings panel or the warn_notinlined
Figure 18-19. Specifying to the compiler that a function may be inlined
#pragma only_std_keywords off inline int attempt_to_inline(void)
{
return 10;
}
To specify that a function must never be inlined, follow its definition's specifier with
__attribute__((never_inline))
.
Figure 18-20 shows an example.
Figure 18-20. Specifying to the compiler that a function must never be inlined
int never_inline(void) __attribute__((never_inline))
{
return 20;
}
To specify that no functions in a file may be inlined, including those that are defined with the inline
,
__inline__
, or
__inline
keywords, use the dont_inline
pragma.
shows an example.
Figure 18-21. Specifying that no functions may be inlined
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
325
Inlining
#pragma dont_inline on
/* Will not be inlined. */ inline int attempt_to_inline(void)
{
return 10;
}
/* Will not be inlined. */ int never_inline(void) __attribute__((never_inline))
{
return 20;
}
#pragma dont_inline off
/* Will be inlined, if possible. */ inline int also_attempt_to_inline(void)
{
return 10;
}
Some kinds of functions are never inlined:
• functions with variable argument lists
• functions defined with
__attribute__((never_inline))
• functions compiled with
#pragma optimize_for_size on
or the Optimize For Size setting in the IDE's Global Optimizations panel
• functions which have their addresses stored in variables
NOTE
The compiler will not inline these functions, even if they are defined with the inline
,
__inline__
, or
__inline keywords.
• functions that return class objects that need destruction
• functions with class arguments that need destruction
The compiler will inline functions that need destruction, without any dependency on the
ISO C++ templates, if the class has a trivial empty constructor. Figure 18-22
shows an example.
Figure 18-22. Inlining function with an empty destructor
struct X {
int n;
X(int a) { n = a; }
~X() {}
}; inline X f(X x) { return X(x.n + 1); }
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
326 Freescale Semiconductor, Inc.
Chapter 18 Intermediate Optimizations
int main()
{
return f(X(1)).n;
}
18.3.2 Inlining Techniques
Normally, the compiler only inlines an eligible function if it has already translated the function's definition. In other words, if an eligible function has not yet been compiled, the compiler has no object code to insert. To overcome this limitation, the compiler can perform interprocedural analysis (IPA) . This lets the compiler evaluate all the functions in a file or even the entire program before inlining the code. The IPA setting in the IDE's
Compiler >C/C++ Language panel.
The compiler normally inlines functions from the first function in a chain of function calls to the last function called. Alternately, the compiler may inline functions from the last function called to the first function in a chain of function calls. The Bottom-up
Inlining option in the IDE's Compiler > Optimization settings panel and the inline_bottom_up
and inline_bottom_up_once
pragmas control this reverse method of inlining.
Some functions that have not been defined with the inline
,
__inline__
, or
__inline keywords may still be good candidates to be inlined. Automatic inlining allows the compiler to inline these functions in addition to the functions that you explicitly specify as eligible for inlining. The Inlining > Auto Inline option in the IDE's Compiler >
Optimization panel and the auto_inline
pragma control this capability.
When inlining, the compiler calculates the complexity of a function by counting the number of statements, operands, and operations in a function to determine whether or not to inline an eligible function. The compiler does not inline functions that exceed a maximum complexity. The compiler uses three settings to control the extent of inlined functions:
• maximum auto-inlining complexity: the threshold for which a function may be autoinlined
• maximum complexity: the threshold for which any eligible function may be inlined
• maximum total complexity: the threshold for all inlining in a function
The inline_max_auto_size
, inline_max_size
, and inline_max_total_size
pragmas control these thresholds, respectively.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
327
Inlining
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
328 Freescale Semiconductor, Inc.
Chapter 19
AIOP Optimizations
This chapter describes optimizations specific to AIOP that the CodeWarrior compiler applies to your object code.
•
19.1 Code Merging
Code merging reduces the size of object code by removing identical functions. Two or more functions are identical when their executable code is identical.
The CodeWarrior build tools can only apply this optimization to object files generated by the CodeWarrior compilers. The CodeWarrior build tools can only apply this optimization to object code translated from C and C++ source code; the tools cannot apply this optimization to object code generated from assembly files.
Tip
For example, the C++ compiler often generates several copies of the same function when it instantiates template functions.
These functions have different names, and these names are considered weak . Under normal circumstances, the linker will issue an error message if it encounters duplicate names. But the linker ignores duplicate names that are marked as weak.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
329
Code Merging
The code merging optimization removes all but one of a group of identical functions.
Table 19-1 shows how to invoke this optimization for all functions. Table 19-2
shows how to invoke this optimization for weak functions.
command line
Table 19-1. Controlling code merging for all identical functions
Control this option from here...
CodeWarrior IDE
use this setting
Choose All Functions from the Code Merging option of the
Linker > General panel.
-code_merging all command line
Table 19-2. Controlling code merging for weak functions only
Control this option from here...
CodeWarrior IDE
use this setting
Choose
Safe Functions from the Code Merging option of the Linker > General panel.
-code_merging safe
The code merging optimization will not remove an identical copy of a function if your program refers to its address. In this case, the compiler keeps this copied function but replaces its executable code with a branch instruction to the original function.
To specify that the compiler and linker must not apply code merging to a function, use this directive in your source code:
__declspec(do_not_merge)
fname ; where fname is the name of a function.
To specify to the linker how to apply code merging to functions, object files, or sections, use these directives in linker command file:
DO_NOT_MERGE
DO_NOT_MERGE_FILES
DO_NOT_MERGE_SECTIONS
AGGRESSIVE_MERGE
AGGRESSIVE_MERGE_FILES
AGGRESSIVE_MERGE_SECTIONS
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
330 Freescale Semiconductor, Inc.
Chapter 20
Inline-Assembly for AIOP Build Tools
This chapter explains how to use the inline assembler built into the CodeWarrior™ C and
C++ compilers for AIOP processors. The compiler's inline assembler allows you to embed assembly language statements in C and C++ functions.
The chapter does not describe the standalone CodeWarrior assembler. For information about this tool, refer to the chapter titled
.
This chapter does not document all the instructions in the AIOP instruction set. For complete documentation of this instruction set, see Programming Environments Manual
for 32-Bit Implementations of the PowerPC™ Architecture,
published by Freescale.
The sections in this chapter are:
•
•
Referring to Assembly, C, and C++ Variables
•
•
20.1 Assembly Syntax
The compiler's inline assembler allows a variety of ways to insert assembly language statements in your C or C++ source code:
•
Specifying Inline Assembly Statements
•
Function-Level Inline Assembly
•
Statement-Level Inline Assembly
•
•
•
•
Normal, Record, and Overflow Forms
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
331
Assembly Syntax
•
•
•
20.1.1 Specifying Inline Assembly Statements
To specify that a block of C or C++ source code should be interpreted as assembly language, use the asm
keyword.
NOTE
To ensure that the C/C++ compiler recognizes the asm
keyword, you must clear the ANSI Keywords Only checkbox in the C/C+
+ Language panel.
As an alternative, the compiler also recognizes the keyword
__asm
even if the ANSI
Keywords Only checkbox is checked.
There are a few ways to use assembly language with the CodeWarrior compilers.
• Function-level assembly language: an entire function is in assembly language.
• Statement-level assembly language: mix assembly language with regular C or C++ statements.
• Intrinsic functions: the compiler makes some assembly instructions available as functions that your program calls as regular C or C++ functions.
Keep these tips in mind as you write inline assembly statements:
• All statements must follow this syntax:
[label :] (instruction | directive ) [operands ]
• Each inline assembly statement must end with a newline or a semicolon (
;
).
• Hexadecimal constants must be in C-style.
For example: li r3, 0xABCDEF
• Assembler directives, instructions, and registers are case-sensitive and must be in lowercase.
20.1.2 Function-Level Inline Assembly
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
332 Freescale Semiconductor, Inc.
Chapter 20 Inline-Assembly for AIOP Build Tools
The compiler accepts function definitions that are composed entirely of assembly statements. Function-level assembly code uses this syntax: asm function-definition
A function that uses function-level assembly must end with a blr
instruction.
Listing 20-1. Example Assembly Language Function
asm void mystrcpy(char *tostr, char *fromstr)
{
addi tostr,tostr,-1
addi fromstr,fromstr,-1
@1 lbzu r5,1(fromstr)
cmpwi r5,0
stbu r5,1(tostr)
bne @1
blr
}
20.1.3 Statement-Level Inline Assembly
The compiler accepts functions that mix regular C/C++ statements with inline assembly.
Statement-level assembly language acts as a block of assembly language that may appear anywhere that the compiler allows a regular C or C++ statement. It has this syntax: asm { one or more instructions }
Listing 20-2. Example of statement-level inline assembly
void g(void)
{
asm { add r2,r3,r4 ; }
}
NOTE
If you check the Inlined Assembler is Volatile checkbox in the
Processor panel, functions that contain an asm
block are only partially optimized. The optimizer optimizes the function, but skips any asm
blocks of code. If the Inlined Assembler is
Volatile checkbox is clear, the compiler also optimizes asm statements.
20.1.4 GCC - Style Inline Assembly
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
333
Assembly Syntax
The CodeWarrior compiler accepts GCC (Gnu Compiler Collection) syntax for inline assembly statements: asm("assembly-statements") where assembly-statements represents inline assembly statements that follow the syntax recognized by the GCC C/C++ compiler.
Figure 20-1. Example of GCC-style inline assembly
void g(void)
{
asm ("add r2,r3,r4\n\t");
}
20.1.5 Branch Prediction
To set the branch prediction ( y
) bit for those branch instructions that can use it, use plus
(+) or minus (-). For example:
@1 bne+ @2
@2 bne- @1
20.1.6 PC-Relative Addressing
The compiler does not accept references to addresses that are relative to the program counter. For example, the following is not supported: asm(b *+8);
Instead, use one of the following:
1. Use labels to specify an address in executable code.
Listing 20-3. Using a label instead if PC-relative addressing
asm(b next); asm(next:);
/* OR */ asm{
b next1
next1:
}
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
334 Freescale Semiconductor, Inc.
Chapter 20 Inline-Assembly for AIOP Build Tools
2. Use relative branch in the function-level assembly instead of statement level.
Listing 20-4. Using relative branching in the function-level assembly
asm void functionLevel(); asm void functionLevel(){
b *+8
nop
blr
}
20.1.7 Normal, Record, and Overflow Forms
Most integer instructions have four forms:
• normal form - add r3,r4,r5
• record form - add. r3,r4,r5
This form ends in a period. This form sets register cr0
to whether the result is less, than, equal to, or greater than zero.
• overflow - addo r3,r4,r5
This form ends in the letter ( o
). This form sets the
SO
and
OV
bits in the
XER
if the result overflows.
• overflow and record - addo. r3,r4,r5
This form ends in ( o.
). This form sets both registers.
Some instructions only have a record form (with a period). Always make sure to include the period. For example:
andi. r3,r4,7
andis. r3,r4,7
stwcx. r3,r4,r5
20.1.8 Creating Statement Labels
The name of an inline assembly language statement label must follow these rules:
• A label name cannot be the same as the identifier of any local variables of the function in which the label name appears.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
335
Assembly Syntax
• A label name does not have to start in the first column of the function in which it appears; a label name can be preceded by white space.
• A label name can begin with an "at-sign" character (
@)
unless the label immediately follows a local variable declaration.
• A label name must end with a colon character (
:
) unless it begins with an at-sign character (
@
).
For example, red:
and
@red
are valid, but red
is not valid.
• A label name can be the same as an assembly language statement mnemonic.
For example, this statement is valid: add: add r3, r4, r5
Examples: asm void func1(){
int i;
@x: li r0,1 //Invalid !!!
} asm void func2(){
int i;
x: li r0,1 //OK
@y: add r3, r4, r5 //OK
}
This is an example of a complete inline assembly language function: asm void red(void){ x1: add r3,r4,r5 @x2: add r6,r7,r8 }
20.1.9 Using Comments
You cannot begin comments with a pound sign (
#
) because the preprocessor uses the pound sign. For example, this format is invalid:
add r3,r4,r5 #
Comment
Use C and C++ comments in this format:
add r3,r4,r5 //
Comment
add r3,r4,r5 /*
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
336 Freescale Semiconductor, Inc.
Chapter 20 Inline-Assembly for AIOP Build Tools
Comment
*/
20.1.10 Using Preprocessor
You can use all preprocessor features, such as comments and macros in the assembler. In multi-line macros, you must end each assembly statement with a semicolon (
;
) because the (
\
) operator removes newlines. For example:
#define remainder(x,y,z) \ divw z,x,y; \ mullw z,z,y; \ subf z,z,x
asm void newPointlessMath(void) { remainder(r3,r4,r5) blr }
20.2 Referring to Assembly, C, and C++ Variables
The compiler's inline assembler has access to the local and global variables declared in C or C++ source code. The assembler also allows access to members of array, struct
, and class
objects:
•
Using Local Variables and Arguments
•
•
Referring to Variables in Instruction Operands
•
Referring to Variables in Memory
•
•
•
Using Variable Names as Memory Locations
•
20.2.1 Using Local Variables and Arguments
To refer to a memory location, you can use the name of a local variable or argument.
The rule for assigning arguments to registers or memory depends on whether the function has a stack frame.
If function has a stack frame, the inline assembler assigns:
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
337
Referring to Assembly, C, and C++ Variables
• scalar arguments declared as register
to general purpose registers r14
to r31
• floating-point arguments declared as register
to floating point fp14
to fp31
• other arguments to memory locations
If a function has no stack frame, the inline assembler assigns arguments that are declared register
and kept in registers. If you have variable or non-register arguments, the compiler will warn you that you should use frfree
NOTE
Some op-codes require registers, and others require objects. For example, if you use nofralloc
with function arguments, you may run into difficulties.
20.2.2 Creating a Stack Frame
You need to create a stack frame for a function if the function:
• calls other functions.
• declares non-register arguments or local variables.
To create a stack frame, use the fralloc
directive at the beginning of your function and the frfree
directive just before the blr
statement. The directive fralloc
automatically allocates
(while ffree
automatically de-allocates) memory for local variables, and saves and restores the register contents.
Listing 20-5. Example of creating a stack frame
asm void red ()
{
fralloc
// Your code here
frfree
blr
}
The fralloc
directive has an optional argument, number, that lets you specify the size, in bytes, of the parameter area of the stack frame. The stack frame is an area for storing parameters used by the assembly code. The compiler creates a 0-byte parameter area for you to pass variables into your assembly language functions.
Function arguments are passed using registers. If your assembly-language routine calls any function that requires more parameters than will fit into registers r3
to r10
and fp1
to fp8
, you need to pass that size to fralloc
. In the case of integer values, registers r3
- r10 are used. For floating-point values, registers fp1
- fp8 are used.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
338 Freescale Semiconductor, Inc.
Chapter 20 Inline-Assembly for AIOP Build Tools
As an example, if you pass 12 values of type long integer
to your assembly function, this would consume 16 bytes of the parameter area. Registers r3
- r10
will hold eight integers, leaving 4 byte integers in the parameter area.
20.2.3 Referring to Variables in Instruction Operands
For instructions that require register operands, (such as the add
instruction), global variables, function parameters, and local variables must be declared with the keyword register
.
shows inline assembly language statements that correctly use
C-language variables as operands in instructions that require register operands.
Listing 20-6. Using C Variables with Instructions that Require Register Operands
int my_global asm("r14")= 25; /* global variable */ int my_loc asm("r15")= 1; asm void red(register int *my_param)
{ register int result; fralloc add result, r4, my_global /* line 10 */ add result, my_global, my_param /* line 11 */ add result, my_param, my_loc /* line 12 */ frfree blr
}
In
Listing 20-6 on page 339 , the statement on line 10, 11, and 12 are all correct because
their operands are all declared with the register
keyword.
20.2.4 Referring to Variables in Memory
For instructions that take a memory operand (such as the lwz
instruction), follow these rules when using a C-language variable as an operand:
• Global variables and function parameters must:
Be declared/assigned with the register.
• Adhere to the syntax below when used as operands, so they are treated as an offset from zero.
instrName regName , 0(globalVarName ) or
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
339
Referring to Assembly, C, and C++ Variables
instrName regName , 0(parameterName )
shows inline assembly language statements that correctly use
C-language variables as operands in instructions that take a memory operand.
Listing 20-7. Using C Variables with Instructions that Take a Memory Operand
int my_global asm("r14")= 25; /* global variable */ int my_loc = 1; /* my_loc is on memory, not in a register */ asm void red(register int *my_param)
{ lwz r4, 0(my_global) /* line 9 */ lwz r4, 0(my_param) /* line 10 */ lwz r4, my_loc /* line 11 */ lwz r4, my_loc(SP) blr
}
In
• The statement on line 9 is correct.
• The operand is fully expressed (because it is an offset from zero).
• The argument my_global
is in a register.
• The statement on line 10 is correct for the same reasons as stated above.
• The statement on line 11 is correct.
The CodeWarrior inline assembler automatically adds the contents of the
SP
register to local variable my_loc
.
• The statement on line 12 is correct.
Note that statements 11 and 12 are equivalent.
As mentioned above, the inline assembler automatically adds the
SP
register to local variable my_loc
, so explicitly including
(SP)
is redundant.
20.2.5 Referring to Registers
For a register operand, you must use one of the register names of the appropriate kind for the instruction. The register names are case-sensitive. You also can use a symbolic name for an argument or local variable that was assigned to a register.
The general registers are
SP
, r0
to r31
, and gpr0
to gpr31
. The floating-point registers are fp0
to fp31
and f0
to f31
. The condition registers are cr0
to cr7
.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
340 Freescale Semiconductor, Inc.
Chapter 20 Inline-Assembly for AIOP Build Tools
20.2.6 Referring to Labels
For a label operand, you can use the name of a label. For long branches (such as b
and bl instructions) you can also use function names. For bla
and la
instructions, use absolute addresses.
For other branches, you must use the name of a label. For example,
• b @3
- correct syntax for branching to a local label
• b red
- correct syntax for branching to external function red
• bl @3
- correct syntax for calling a local label
• bl red
- correct syntax for calling external function red
• bne red
- incorrect syntax; short branch outside function red
NOTE
You cannot use local labels that have already been declared in other functions.
20.2.7 Using Variable Names as Memory Locations
Whenever an instruction, such as a load instruction, a store instruction, or la
, requires a memory location, you can use a local or global variable name. You can modify local variable names with struct member references, class member references, array subscripts, or constant displacements. For example, all the local variable references in
Listing 20-8 on page 341 are valid.
Listing 20-8. Example of referring to variables stored in memory locations
asm void red(void){
long myVar;
long myArray[1];
Rect myRectArray[3];
fralloc
lwz r3,myVar(SP)
la r3,myVar(SP)
lwz r3,myRect.top
lwz r3,myArray[2](SP)
lwz r3,myRectArray[2].top
lbz r3,myRectArray[2].top+1(SP)
frfree
blr
}
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
341
Referring to Assembly, C, and C++ Variables
You can also use a register variable that is a pointer to a struct
or class
to access a
member of the object, shown in Listing 20-9 on page 342
.
Listing 20-9. Example of referring to a struct or class member
void red(void){
Rect q;
register Rect *p = &q;
asm {
lwz r3,p->top;
}
}
You can use the
@hiword
and
@loword
directives to access the high and low four bytes of 8
byte long longs and software floating-point doubles ( Listing 20-10 on page 342
).
Listing 20-10. Example of referring to high and low words
long long gTheLongLong = 5; asm void Red(void); asm void Red(void)
{
fralloc
lwz r5, gTheLongLong@hiword
lwz r6, gTheLongLong@loword
frfree
blr
}
20.2.8 Using Immediate Operands
For an immediate operand, you can use an integer or enum constant, sizeof
expression, and any constant expression using any of the C dyadic and monadic arithmetic operators.
These expressions follow the same precedence and associativity rules as normal C expressions. The inline assembler carries out all arithmetic with 32-bit signed integers.
An immediate operand can also be a reference to a member of a struct or class type. You can use any struct or class name from a typedef
statement, followed by any number of member references. This evaluates to the offset of the member from the start of the struct.
For example:
lwz r4,Rect.top(r3)
addi r6,r6,Rect.left
As a side note, la rD,d(rA)
is the same as addi rD,rA,d
.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
342 Freescale Semiconductor, Inc.
Chapter 20 Inline-Assembly for AIOP Build Tools
You also can use the top or bottom half-word of an immediate word value as an immediate operand by using one of the @ modifiers (
).
Listing 20-11. Example of referring to immediate operands
long gTheLong; asm void red(void)
{
fralloc
lis r6, gTheLong@ha
addi r6, r6, gTheLong@h
lis r7, gTheLong@h
ori r7, br7, gTheLong@l
frfree
blr
}
The access patterns are:
lis x,var@ha
la x,var@l(x) or
lis x,var@h
ori x,x,var@l
In this example, la
is the simplified form of addi
to load an address. The instruction las
is similar to la
but shifted.
Using
@ha
is preferred since you can write:
lis x,var@ha
lwz v,var@l(x)
You cannot do this with
@h
because it requires that you use the ori
instruction.
20.3 Assembler Directives
This section describes some special assembler directives that the AIOP built-in assembler accepts. These directives are:
•
•
•
•
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
343
Assembler Directives
•
•
20.3.1 entry
Defines an entry point into the current function.
entry [ extern | static ]
name
Use the extern
qualifier to declare a global entry point; use the static
qualifier to declare a local entry point. If you leave out the qualifier, extern
is assumed.
NOTE
Inline-assembly directive entry
can be used only with Functionlevel assembly code.
Listing 20-12 on page 344 shows how to use the
entry
directive.
Listing 20-12. Using the entry directive
void __save_fpr_15(void); void __save_fpr_16(void); asm void __save_fpr_14(void)
{
stfd fp14,-144(SP)
entry __save_fpr_15
stfd fp15,-136(SP)
entry __save_fpr_16
stfd fp16,-128(SP)
// ...
blr
}
20.3.2 fralloc
Creates a stack frame for a function and reserves registers for local register variables.
fralloc [ number ]
You need to create a stack frame for a function if the function:
• calls other functions.
• uses more arguments than will fit in the designated parameters
(r3
- r10
, fp1
- fp8
).
• declares local registers.
• declares non-registered parameters.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
344 Freescale Semiconductor, Inc.
Chapter 20 Inline-Assembly for AIOP Build Tools
The fralloc
directive has an optional argument number that lets you specify the size in bytes of the parameter area of the stack frame. The compiler creates a 0-byte parameter area. If your assembly language routine calls any function that requires more parameters than will fit in r3
- r10
and fp1
- fp8
, you must specify a larger amount.
20.3.3 frfree
Frees a function's stack frame and restores local register variables.
frfree
This directive frees the stack frame and restores the registers that fralloc
reserved.
The frfree
directive does not generate a blr
instruction. If your function uses functionlevel inline assembly, you must explicitly terminate it with this instruction.
20.3.4 machine
Specifies the processor that the assembly language targets.
machine
number
The value of number must be one of those listed in the table below.
505 all
509 generic
Table 20-1. CPU Identifiers
555 56x
If you use generic
, the compiler supports the core instructions for the 603, 604, 740, and
750 processors. In addition, the compiler supports all optional instructions.
If you use all
, the compiler recognizes assembly instructions for all core and optional instructions for all AIOP processors.
If you do not use the machine
directive, the compiler uses the settings you selected from the Processor listbox of the Processor settings panel.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
345
Assembler Directives
20.3.5 nofralloc
Specifies that the function will build a stack frame explicitly.
nofralloc
Use the nofralloc
directive so that an inline assembly function does not build a stack frame. When you use nofralloc
, if you have local variables, parameters or make function calls, you are responsible for creating and deleting your own stack frame. For an example of nofralloc
, see the file
__start.c
in the directory:
InstallDir
LS\AIOP_Support\ewl\EWL_Runtime\Runtime_AIOP\Source where InstallDir is the name of the directory on your host computer where you installed your CodeWarrior development tools.
20.3.6 opword
Inserts raw bytes into the object code.
opword
value
This directive inserts value into the object code. For example opword 0x7C0802A6 is equivalent to mflr r0
The compiler does not check the validity of value ; the compiler simply copies it into the object code that it generates.
20.3.7 equ
Sets the value of symbol to expression.
equ
symbol, expression
For example
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
346 Freescale Semiconductor, Inc.
Chapter 20 Inline-Assembly for AIOP Build Tools
foo()
{ asm(".equ VAL, 24"); asm volatile("se_li r5, VAL");
} fun()
{ asm volatile("li VAL, 12");
}
20.4 Intrinsic Functions
Intrinsic functions are a mechanism you can use to get assembly language into your source code without using the asm
keyword. Intrinsic functions are not part of the
ISO/IEC C or C++ standards. They are an extension provided by the CodeWarrior compilers.
There is an intrinsic function for several common processor op-codes (instructions).
Rather than using inline assembly syntax and specifying the op-code in an asm
block, you call the intrinsic function that matches the op-code.
When the compiler encounters the intrinsic function call in your source code, it does not actually make a function call. The compiler substitutes the assembly instruction that matches your function call. As a result, no function call occurs in the final object code.
The final code is the assembly language instructions that correspond to the intrinsic functions.
20.4.1 Low-Level Processor Synchronization
These functions perform low-level processor synchronization.
• void __eieio(void)
- Enforce in-order execution of I/O
• void __sync(void)
- Synchronize
• void __isync(void)
- Instruction synchronize
For more information on these functions, see the instructions eieio
, sync
, and isync
in
PowerPC Microprocessor Family: The Programming Environments by Freescale.
20.4.2 Absolute Value Functions
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
347
Intrinsic Functions
These functions generate inline instructions that take the absolute value of a number.
• int __abs(int)
- Absolute value of an integer
• float __fabs(float)
- Absolute value of a float
• float __fnabs(float)
- Negative absolute value of a float
• long __labs(long)
- Absolute value of a long int
__fabs(float)
and
__fnabs(float)
are not available if the Hardware option button is cleared in the Processor settings panel.
20.4.3 Byte-Reversing Functions
These functions generate inline instructions that can dramatically speed up certain code sequences, especially byte-reversal operations.
• int __lhbrx(const void *, int)
- Load halfword byte; reverse indexed
• int __lwbrx(const void *, int)
- Load word byte; reverse indexed
• void __sthbrx(unsigned short, const void *, int)
- Store halfword byte; reverse indexed
• void __stwbrx(unsigned int, const void *, int)
- Store word byte; reverse indexed
20.4.4 Setting the Floating-Point Environment
This function lets you change the Floating Point Status and Control Register ( FPSCR). It sets the FPSCR to its argument and returns the original value of the FPSCR.
This function is not available if you select the None option button in the Processor settings panel.
float
__setflm(float); shows how to set and restore the FPSCR.
Listing 20-13. Example of setting the FPSCR
double old_fpscr;
/* Clear flag/exception/mode bits, save original settings */ oldfpscr = __setflm(0.0);
/* Peform some floating-point operations */
__setflm(old_fpscr); /* Restores the FPSCR */
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
348 Freescale Semiconductor, Inc.
Chapter 20 Inline-Assembly for AIOP Build Tools
20.4.5 Manipulating the Contents of a Variable or Register
These functions rotate the contents of a variable to the left:
• int __rlwinm(int, int, int, int)
- Rotate left word (immediate), then AND with mask
• int __rlwnm(int, int, int, int)
- Rotate left word, then AND with mask
• int __rlwimi(int, int, int, int, int)
- Rotate Left word (immediate), then mask insert
The first argument to
__rlwimi
is overwritten. However, if the first parameter is a local variable allocated to a register, it is both an input and output parameter. For this reason, this intrinsic should always be written to put the result in the same variable as the first parameter as shown here: ra = __rlwimi( ra, rs, sh, mb, me );
You can count the leading zeros in a register using this intrinsic: int
__cntlzw(int);
You can use inline assembly for a complete assembly language function, as well as individual assembly language statements.
20.4.6 Data Cache Manipulation
The intrinsics shown in the table below, map directly to AIOP assembly instructions
Table 20-2. Data Cache Intrinsics
Intrinsic Prototype
void __dcbf(const void *, int); void __dcbt(const void *, int); void __dcbst(const void *, int); void __dcbtst(const void *, int); void __dcbz(const void *, int); void __dcba(const void *, int); dcbf dcbt dcbst dcbtst dcbz dcba
AIOP Instruction
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
349
Intrinsic Functions
20.4.7 Math Functions
The table below lists intrinsic functions for mathematical operations.
Table 20-3. Math Intrinsics
Intrinsic Prototype
int __mulhw(int, int); uint __mulhwu(uint, uint); double __fmadd(double, double, double); double __fmsub(double, double, double); double __fnmadd(double, double, double); double __fnmsub(double, double, double); float __fmadds(float, float, float); float __fmsubs(float, float, float); float __fnmadds(float, float, float); float __fnmsubs(float, float, float); double __mffs(void); float __fabsf(float); float __fnabsf(float); mulhw mulhwu fmadd fmsub fnmadd fnmsub fmadds fmsubs fnmadds fnmsubs mffs fabsf fnabsf
AIOP Instruction
20.4.8 Buffer Manipulation
Some intrinsics allow control over areas of memory, so you can manipulate memory blocks.
void *__alloca(ulong);
__alloca
implements alloca()
in the compiler. char *__strcpy(char *, const char *);
__strcpy()
detects copies of constant size and calls
__memcpy()
. This intrinsic requires that a
__strcpy
function be implemented because if the string is not a constant it will call
__strcpy
to do the copy.
void *__memcpy(void *, const void *, size_t);
__memcpy()
provides access to the block move in the code generator to do the block move inline.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
350 Freescale Semiconductor, Inc.
Chapter 20 Inline-Assembly for AIOP Build Tools
20.4.9 AIOP Intrinsics
This chapter explains the following AIOP intrinsics, supported by the current release:
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
351
Intrinsic Functions
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
20.4.9.1 __ldw
This function loads double word.
Syntax
void __ldw(unsigned int &arg1, unsigned int &arg2, unsigned int displ, void * base)
Remarks
•
&arg1
and
&arg2
are loaded from location (displ + base). Where
&
serves as a notational convenience to indicate that the argument is defined by the intrinsic function call.
The compiler will treat the argument as a register object and retain the object in a register.
• Registers for arg1
and arg2
are allocated to even-odd consecutive register by the compiler (excluding R0).
• displ
is a word aligned constant value between 0-1020.
• If base
is a literal 0, the rA
field in the instruction is encoded as 0.
20.4.9.2 __llldw
This function loads double word using a unsigned long long argument.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
352 Freescale Semiconductor, Inc.
Chapter 20 Inline-Assembly for AIOP Build Tools
Syntax
void __llldw(unsigned long long &arg1, unsigned int displ, void * base);
20.4.9.3 __ldwbrw
This function loads byte reversed double word.
Syntax
void __ldwbrw(unsigned int &arg1, unsigned int &arg2, unsigned int displ, void * base)
Remarks
• arg1
and arg2
are loaded from location (displ + base). Where
&
serves as a notational convenience to indicate that the argument is defined by the intrinsic function call.
The compiler will treat the argument as a register object and retain the object in a register.
• Registers for arg1
and arg2
are allocated to even-odd consecutive register by the compiler (excluding R0).
• Bytes within the words are byte reversed.
• displ must be 0.
• If base
is a literal 0, the rA
field in the instruction is encoded as 0.
20.4.9.4 __llldwbrw
This function loads byte reversed double word using a unsigned long long argument.
Syntax
void __llldwbrw(unsigned long long expr, unsigned int displ, void * base);
20.4.9.5 __stdw
This function stores double word.
Syntax
void __stdw(unsigned int arg1, unsigned int arg2, unsigned int displ, void * base)
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
353
Intrinsic Functions
• arg1
and arg2
are stored to location (displ + base).
• Registers for arg1
and arg2
are allocated to even-odd consecutive register by the compiler (excluding R0).
• displ is a word aligned constant value between 0-1020.
• If base
is a literal 0, the rA
field in the instruction is encoded as 0.
20.4.9.6 __llstdw
This function stores double word using an unsigned long long argument.
Syntax
void __llstdw(unsigned long long expr, unsigned int displ, void * base);
20.4.9.7 __stdwbrw
This function stores byte reversed double word.
Syntax
void __stdwbrw(unsigned int arg1, unsigned int arg2, unsigned int displ, void * base)
Remarks
• arg1
and arg2
are stored to location (displ + base).
• Registers for arg1
and arg2
are allocated to even-odd consecutive register by the compiler (excluding R0).
• Bytes are reversed within words.
• displ
must be 0.
• If base
is a literal 0, the rA
field in the instruction is encoded as 0.
20.4.9.8 __llstdbrw
This function stores a single long long expression to memory, byte reversed.
Syntax
void __llstdbrw(unsigned long long expr, unsigned int displ, void *base)
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
354 Freescale Semiconductor, Inc.
Chapter 20 Inline-Assembly for AIOP Build Tools
20.4.9.9 __lqw
This function loads quad word.
Syntax
void __lqw(unsigned int &arg1, unsigned int &arg2, unsigned int &arg3, unsigned int
&arg4,unsigned int displ, void * base)
Remarks
• arg1-4
are loaded from location (displ + base). Where
&
serves as a notational convenience to indicate that the argument is defined by the intrinsic function call.
The compiler will treat the argument as a register object and retain the object in a register.
• Registers for arg1-4
are allocated to even/odd/even or odd consecutive register by the compiler (excluding R0).
• displ
is a word aligned constant value between 0-1020.
• If base
is a literal 0, the rA field in the instruction is encoded as 0.
20.4.9.10 __lqdbrw
This function loads byte-reversed quad word.
Syntax
void __lqdbrw(unsigned int &arg1, unsigned int &arg2, unsigned int &arg3, unsigned int
&arg4, unsigned int displ, void * base)
Remarks
• arg1-4
are loaded from location (displ + base). Where
&
serves as a notational convenience to indicate that the argument is defined by the intrinsic function call.
The compiler will treat the argument as a register object and retain the object in a register.
• Registers for arg1-4
are allocated to even/odd/even/odd consecutive register by the compiler (excluding R0).
• Bytes within the double words are byte reversed.
• If base
is a literal 0, the rA field in the instruction is encoded as 0.
20.4.9.11 __stqw
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
355
Intrinsic Functions
This function stores quad word.
Syntax
void __stqw(unsigned int arg1, unsigned int arg2, unsigned int arg3, unsigned int arg4, unsigned int displ, void * base)
Remarks
• arg1-4
are stored to location (displ + base).
• Registers for arg1-4
are allocated to even/odd/even/odd consecutive register by the compiler (excluding R0).
• displ
is a word aligned constant value between 0-1020.
• If base
is a literal 0, the rA field in the instruction is encoded as 0.
20.4.9.12 __stqdbrw
This function stores byte-reversed quad word.
Syntax
void __stqdbrw(unsigned int arg1, unsigned int arg2, unsigned int arg3, unsigned int arg4, unsigned int displ, void * base)
Remarks
• arg1-4
are stored to location (displ + base).
• Registers for arg1-4
are allocated to even/odd/even/odd consecutive register by the compiler (excluding R0).
• Bytes within a double word are byte reversed.
• displ
must be 0. If base
is a literal 0, the rA
field in the instruction is encoded as 0.
20.4.9.13 __e_osmcmd
Orders scope management command request.
Syntax
int __e_osmcmd(unsigned int osm_op, unsigned int scope_expr) int __e_osmcmd_(unsigned int osm_op, unsigned int scope_expr) -- recording form
Remarks
• osm_op
must be a literal constant C expression.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
356 Freescale Semiconductor, Inc.
Chapter 20 Inline-Assembly for AIOP Build Tools
• scope_expr
is a C expression whose results are representable in a register.
• If scope_expr
is a literal zero value, then no scope specifier is provided from the instruction directly.
20.4.9.14 __e_hwaccel
Hardware acceleration request.
Syntax
int __e_hwaccel(unsigned int accel_id) int __e_hwaccel_(unsigned int accel_id) -- recording form
Remarks
• accel_id
is a C expression whose results are representable in a register.
• accel_id
must not be a literal zero value.
20.4.9.15 __e_hwacceli
Hardware acceleration request
Syntax
int __e_hwacceli(unsigned int accel_id) int __e_hwacceli_(unsigned int accel_id) -- recording form
Remarks
• accel_id
is a literal constant C expression representing the accelerator id.
• Legal values are 0-31. A zero literal value indicates the
Null
accelerator.
20.4.9.16 __e_ordhwaccel
Ordered hardware acceleration request
Syntax
int __e_ordhwaccel(unsigned int accel_id, unsigned int osm_op) int __e_ordhwaccel_(unsigned int accel_id,unsigned int osm_op) -- recording form
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
357
Intrinsic Functions
• accel_id
is a C expression representable in a register. The expression must not be a literal zero expression.
• osm_op
is a literal constant C expression between 0 and 31.
20.4.9.17 __e_ordhwacceli
Ordered hardware acceleration request
Syntax
int __e_ordhwacceli(unsigned int accel_id, unsigned int osm_op, unsigned int rA_expr) int __e_ordhwacceli_(unsigned int accel_id,unsigned int osm_op, unsigned int rA_expr) -- recording form
Remarks
• accel_id
is a literal constant C expression between 0 and 31.
• osm_op
is a literal constant C expression between 0 and 31.
• rA_expr
is a C expression representable in a register.
20.4.9.18 __e_rlwimi
Rotates Left Word Immediate then Mask Insert.
Syntax
uint32 __e_rlwimi(uint32 ra, uint32 rs, uint32 sh, uint32 mb, uint32 me)
Remarks
The contents of symbol rs
are rotated left sh
bits. A mask is generated having 1-bits from bit mb+32
through bit me+32
and 0-bits elsewhere. The rotated data is inserted into symbol ra under control of the generated mask. The intrinsic also returns an unsigned int
value containing the ra
value.
NOTE
The syntax of this instrinsic does not reflect that the ra
symbol is both read and written. Therefore, it is mandatory that the above intrinsic be written as: ra = __e_rlwimi(ra, rs, sh, mb, me);
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
358 Freescale Semiconductor, Inc.
Chapter 20 Inline-Assembly for AIOP Build Tools
This will ensure that the ra
symbol is properly written with the new value resulting from the operation.
20.4.9.19 __e_rlwinm
Rotate Left Word Immediate then AND with Mask
Syntax
uint32 __e_rlwinm(uint32 ra, uint32 rs, uint32 sh, uint32 mb, uint32 me)
Remarks
The contents of symbol rs
are rotated left sh
bits. A mask is generated having 1-bits from bit mb+32
through bit me+32
and 0-bits elsewhere. The rotated data is ANDed with the generated mask and the result is placed into symbol ra
. The intrinsic also returns an unsigned int
value containing the ra
value.
NOTE
The syntax of this instrinsic does not reflect that the ra
symbol is both read and written. Therefore, it is mandatory that the above intrinsic be written as: ra = __e_rlwinm(ra, rs, sh, mb, me);
This will ensure that the ra
symbol is properly written with the new value resulting from the operation.
20.4.9.20 __addb
This function adds two signed char
expressions to obtain the signed result in an int
. Suffix the intrinsic with
-
to indicate the recording form of the operation.
Syntax
signed int __addb(signed char expr, signed char expr) signed int __addb_(signed char expr, signed char expr)
20.4.9.21 __addbu
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
359
Intrinsic Functions
This function adds two unsigned char
expressions to obtain an unsigned result in an int
.
Suffix the intrinsic with
-
to indicate the recording form of the operation.
Syntax
unsigned int __addbu(unsigned char expr, unsigned char expr) unsigned int __addbu_(unsigned char expr, unsigned char expr)
20.4.9.22 __addh
This function adds two signed half-word
expressions to obtain an signed result in an int
.
Suffix the intrinsic with
-
to indicate the recording form of the operation.
Syntax
signed int __addh(signed short expr, signed short expr) signed int __addh_(signed short expr, signed short expr)
20.4.9.23 __addhu
This function adds two unsigned half-word
expressions to obtain an unsigned result in an int
. Suffix the intrinsic with
-
to indicate the recording form of the operation.
Syntax
unsigned int __addhu(unsigned short expr, unsigned short expr) unsigned int __addhu_(unsigned short expr, unsigned short expr)
20.4.9.24 __subfb
This function calculates the sum of one's complement of the signed first char
, the signed second char
expression and
1
to obtain an signed result in an int
. Suffix the intrinsic with
-
to indicate the recording form of the operation.
Syntax
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
360 Freescale Semiconductor, Inc.
signed int __subfb(signed char expr, signed char expr) signed int __subfb_(signed char expr, signed char expr)
Chapter 20 Inline-Assembly for AIOP Build Tools
20.4.9.25 __subfbu
This function calculates the sum of one's complement of the unsigned first char
, the unsigned second char
expression and
1
to obtain an unsigned result in an int
. Suffix the intrinsic with
-
to indicate the recording form of the operation.
Syntax
unsigned int __subfbu(unsigned char expr, unsigned char expr) unsigned int __subfbu_(unsigned char expr, unsigned char expr)
20.4.9.26 __subfh
This function calculates the sum of one's complement of the signed first half-word expression, the signed second half-word
expression and
1
to obtain a signed result in an int
. Suffix the intrinsic with
-
to indicate the recording form of the operation.
Syntax
signed int __subfh(signed short expr, signed short expr) signed int __subfh_(signed short expr, signed short expr)
20.4.9.27 __subfhu
This function calculates the sum of one's complement of the unsigned first half-word expression, the unsigned second half-word
expression and
1
to obtain an unsigned result in an int
. Suffix the intrinsic with
-
to indicate the recording form of the operation.
Syntax
unsigned int __subfhu(unsigned short expr, unsigned short expr) unsigned int __subfhu_(unsigned short expr, unsigned short expr)
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
361
Intrinsic Functions
20.4.9.28 __byterevw
This function reverses the four bytes of unsigned int
expression to return an unsigned int
.
Syntax
unsigned int __byterevw(unsigned int expr)
20.4.9.29 __byterevh
This function reverses bytes of each halfword in unsigned int
expression to return an unsigned int
.
Syntax
unsigned int __byterevh(unsigned int expr)
20.4.9.30 __ldwar
This function loads double word into two unsigned expressions. The instruction creates a reservation.
Syntax
void __ldwar(unsigned int &arg1, unsigned int &arg2, unsigned int displ, void *base)
20.4.9.31 __llldwar
This function loads double word into a single unsigned long long expression. The instruction creates a reservation.
Syntax
void __llldwar(unsigned long long &arg1, unsigned int displ, void *base)
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
362 Freescale Semiconductor, Inc.
Chapter 20 Inline-Assembly for AIOP Build Tools
20.4.9.32 __ldwcb
This function loads double word as two unsigned expressions. The instruction bypasses the cache.
Syntax
void __ldwcb(unsigned int &arg1, unsigned int &arg2, unsigned int displ, void *base)
20.4.9.33 __llldwcb
This function loads double word into a single unsigned long long expression. The instruction bypasses the cache.
Syntax
void __llldwcb(unsigned long long &arg1, unsigned int displ, void *base)
20.4.9.34 __ldbrw
This function loads double word into two unsigned expressions. The loaded double word is byte-reversed.
Syntax
void __ldbrw(unsigned int &arg1, unsigned int &arg2, unsigned int displ, void *base)
20.4.9.35 __llldbrw
This function loads double word into a single unsigned long long expression. The loaded double word is byte-reversed.
Syntax
void __llldbrw(unsigned long long &arg1, unsigned int displ, void *base)
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
363
Intrinsic Functions
20.4.9.36 __lhbr
This function loads a half-word from memory into an int
expression. The half-word is byte-reversed.
Syntax
int __lhbr(unsigned int displ, void *base)
20.4.9.37 __lwbr
This function loads a word from memory into an int
expression. The word is bytereversed.
Syntax
int __lwbr(unsigned int displ, void *base)
20.4.9.38 __stdwc
This function stores two unsigned int
expressions to memory. The instruction is conditional on reservation status.
Syntax
void __stdwc(unsigned int expr, unsigned int expr, unsigned int displ, void *base)
20.4.9.39 __llstdwc
This function stores a single unsigned long long expression to memory. The instruction is conditional on reservation status.
Syntax
void __llstdwc(unsigned long long expr, unsigned int displ, void *base)
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
364 Freescale Semiconductor, Inc.
Chapter 20 Inline-Assembly for AIOP Build Tools
20.4.9.40 __stdwwt
This function stores two unsigned int
expressions to memory. The instruction is cache bypassed with write-through.
Syntax
void __stdwwt(unsigned int expr, unsigned int expr, unsigned int displ, void *base)
20.4.9.41 __llstdwwt
This function stores a single unsigned long long expression to memory. The instruction is cache bypassed with write-through.
Syntax
void __llstdwwt(unsigned long long expr, unsigned int displ, void *base)
20.4.9.42 __stdbrw
This function stores two unsigned int
expressions to memory, byte reversed.
Syntax
void __stdbrw(unsigned int expr, unsigned int expr, unsigned int displ, void *base)
20.4.9.43 __llstdbrw
This function stores a single long long expression to memory, byte reversed.
Syntax
void __llstdbrw(unsigned long long expr, unsigned int displ, void *base)
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
365
Intrinsic Functions
20.4.9.44 __sthbr
This function stores a half-word expression to memory, byte reversed.
Syntax
void __sthbr(unsigned short, unsigned int displ, void *base)
20.4.9.45 __stwbr
This function stores a word expression to memory, byte reversed.
Syntax
void __stwbr(unsigned int expr, unsigned int displ, void *base)
20.4.9.46 __lbdx
This function loads a char
byte addressed by base and decorated by expression and returns a signed char
. Suffix the intrinsic with
-
to indicate the recording form of the operation.
Syntax
char __lbdx(unsigned int expr, void *base) char __lbdx_(unsigned int expr, void *base)
20.4.9.47 __lhdx
This function loads a half-word
addressed by base and decorated by expression and returns a signed short
. Suffix the intrinsic with
-
to indicate the recording form of the operation.
Syntax
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
366 Freescale Semiconductor, Inc.
Chapter 20 Inline-Assembly for AIOP Build Tools
short __lhdx(unsigned int expr, void *base) short __lhdx_(unsigned int expr, void *base)
20.4.9.48 __lwdx
This function loads a word
addressed by base and decorated by expression and returns a signed int
. Suffix the intrinsic with
-
to indicate the recording form of the operation.
Syntax
int __lwdx(unsigned int expr, void *base) int __lwdx_(unsigned int expr, void *base)
20.4.9.49 __stbdcbx
This function stores a char
to a location addressed by base and decorated by an expression, cache bypassed.
Syntax
void __stbdcbx(char, unsigned int expr, void *base)
20.4.9.50 __sthdcbx
This function stores a short
to a location addressed by base and decorated by an expression, cache bypassed.
Syntax
void __sthdcbx(short, unsigned int expr, void *base)
20.4.9.51 __stwdcbx
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
367
Intrinsic Functions
This function stores an int
to a location addressed by base and decorated by an expression, cache bypassed.
Syntax
void __stwdcbx(int, unsigned int expr, void *base)
20.4.9.52 __dsn
Decorated storage notify of address base and decoration expression.
Syntax
void __dsn(unsigned int expr, void *base)
20.4.9.53 __dsncb
Decorated storage notify of address base and decoration expression, cache bypassed.
Syntax
void __dsncb(unsigned int expr, void *base)
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
368 Freescale Semiconductor, Inc.
Chapter 21
AIOP Code Generation
This chapter describes the conventions that the C/C++ compiler and linker follow to generate object code, the data types that the compiler recognizes, and how to specify to the compiler the byte-alignment of data in object code.
•
•
•
•
•
Small Data Area PIC/PID Support
•
•
•
21.1 ABI Conformance
The CodeWarrior compiler for AIOP follows the Embedded application binary interface
(EABI) specified by PowerPC Embedded Binary Interface, 32-Bit Implementation.
21.2 Data Representation
The compiler recognizes ISO standard data types and some AIOP-specific types:
•
•
•
•
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
369
Data Representation
21.2.1 Boolean Type
Table 21-1 lists the name, size, and range of the boolean data type. The compiler
recognizes this data type when compiling C99 (ISO/IEC 9899-1999) source code.
_Bool
This type
Table 21-1. C99 boolean data type has this size
8 bits when pragma uchar_bool is on ,
32 bits when pragma uchar_bool is off
and holds this range of values
0 ("false") and 1 ("true")
Table 21-2 lists the name, size, and range of the C++ boolean data type. The C++
compiler does not recognize the C99
_Bool
type.
bool
This type
Table 21-2. Boolean data type has this size
8 bits when pragma uchar_bool is on ,
32 bits when pragma uchar_bool is off
and holds this range of values
true , false
21.2.2 Character Types
Table 21-3 lists the name, size, and range of the character data types.
This type
char unsigned char signed char
Table 21-3. Character data types has this size
8 bits
8 bits
8 bits
and holds this range of values
either -128 to 127 or 0 to 255
0 to 255
-128 to 127
Table 21-4 lists the name, size, and range of the C++
wchar_t
data types.
wchar_t
This type
Table 21-4. Character data types has this size
16 bits
Table continues on the next page...
and holds this range of values
either -32768 to 32767 or 0 to 65535
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
370 Freescale Semiconductor, Inc.
This type
unsigned wchar_t signed wchar_t
Chapter 21 AIOP Code Generation
Table 21-4. Character data types (continued)
16 bits
16 bits
has this size and holds this range of values
0 to 65535
-32768 to 32767
The pragma unsigned_char controls whether or not the compiler treats the wchar_t
and char
types as signed or unsigned.
21.2.3 Integer Types
Table 21-5 lists the name, size, and range of the integer data types.
This type
short unsigned short int unsigned int long unsigned long long long unsigned long long
16 bits
16 bits
32 bits
32 bits
32 bits
32 bits
64 bits
Table 21-5. Integer data type has this size
64 bits
and holds this range of values
-32,768 to 32,767
0 to 65,535
-2,147,483,648 to 2,147,483,647
0 to 4,294,967,295
-2,147,483,648 to 2,147,483,647
0 to 4,294,967,295
-9,223,372,036,854,775,808 to
9,223,372,036,854,775,807
0 to 18,446,744,073,709,551,615
The compiler recognizes the long long
data type when pragma longlong
is on or when compiling C99 source code (ISO/IEC 9899-1999 standard).
21.2.4 Floating-Point
Table 21-6 lists the floating point data types
float double long double
Type
Table 21-6. AIOP Floating Point Types
32 bits
64 bits
64 bits
Size Range
1.17549e-38 to 3.40282e+38
2.22507e-308 to 1.79769e+308
2.22507e-308 to 1.79769e+308
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
371
Data Addressing
21.3 Data Addressing
In absolute addressing, the compiler generates two instructions to fetch the address of a
variable. For example the compiler translates Listing 21-1 on page 372
into the instructions in
Listing 21-1. Source Code
int red; int redsky; void sky()
{
red = 1;
redsky = 2;
}
Listing 21-2. Generated Code
li r3,1 lis r4,red@ha addi r4,r4,red@l stw r3,0(r4) li r5,2 lis r6,redsky@ha addi r6,r6,redsky@l stw r5,0(r6)
Each variable access takes two instructions and a total of four bytes to make a simple assignment. If you set the small data threshold to be at least the size of an int
data type, the compiler generates instructions to fetch variables with one instruction (
Listing 21-3. Fetching variables with one instruction
li r3,1 stw r3,red li r4,2 stw r4,redsky
Because small data sections are limited in size you might not be able to put all of your application data into the small data and small data2 sections. We recommend that you make the threshold as high as possible until the linker reports that you have exceeded the size of the section.
If you do exceed the available small data space, consider using pooled data.
Because the linker can not deadstrip unused pooled data, you should:
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
372 Freescale Semiconductor, Inc.
Chapter 21 AIOP Code Generation
1. Check the Generate Link Map and List Unused Objects checkboxes in the
CodeWarrior IDE's Linker Properties > C/C++ Build > Settings > Tool Settings >
Linker > Output panel.
2. Link and examine the map for data objects that are reported unused.
3. Delete or comment out those used definitions in your source.
4. Select Properties > C/C++ Build > Settings > Tool Settings > Compiler >
Processor, check theCheck the Pool Data checkbox.
The code in Listing 21-4 on page 373 has a zero small data threshold.
Listing 21-4. Zero Small Data Threshold
lis r3,...bss.0@ha addi r3,r3,...bss.0@l li r0,1 stw r0,0(r3) li r0,2 stw r0,4(r3)
When pooled data is implemented, the first used variable of either the
.data
,
.bss
or
.rodata
section gets a two-instruction fetch of the first variable in that section.
Subsequent fetches in that function use the register containing the already-loaded section address with a calculated offset.
NOTE
You can access small data in assembly files with the twoinstruction fetch used with large data, because any data on your board can be accessed as if it were large data. The opposite is not true; large data can never be accessed with small data relocations (the linker issues an error if you try to do so).
External declarations of empty arrays (for example, extern int red []
;) are always treated as if they were large data. If you know that the size of the array fits into a small data section, specify the size in the brackets.
21.4 Aligning Data
This section contains these topics:
•
•
Aligning a Variable Declaration
•
Alignment in a Structure Definition
•
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
373
Aligning Data
•
•
21.4.1 Alignment Attribute Syntax
Use
__attribute__ ((aligned(...))) directive to specify to the compiler on what memory boundary to store data objects. This directive specifies which multiple of bytes to store an object.
The format of this directive is
__attribute__ ((aligned( x )) where x is a decimal number of a power of 2 from 1 to 8192.
21.4.2 Aligning a Variable Declaration
Use the alignment attribute to specify a variable's alignment. For example, the following variable declaration aligns
V1
on a 16-byte boundary.
int V1[4] __attribute__ ((aligned (16)));
The following variable declaration aligns
V2
on a 2-byte boundary.
int V2[4] __attribute__ ((aligned (2)));
21.4.3 Alignment in a Structure Definition
Use the alignment attribute to specify how instances of a structure should be aligned. You must specify a minimum alignment of at least 4 bytes for structures. Specifying a lower number might cause alignment exceptions at runtime.
For example, this definition aligns all definitions of struct S1
on an 8-byte boundary.
struct S1 { short f[3]; }
__attribute__ ((aligned (8)));
struct S1 s1;
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
374 Freescale Semiconductor, Inc.
Chapter 21 AIOP Code Generation
The following definition aligns all definitions of struct S2
on a 4-byte boundary.
struct S2 { short f[3]; }
__attribute__ ((aligned (1)));
struct S2 s2;
21.4.4 Structure Member
Use the alignment attribute to specify how to align a member in a structure.
NOTE
Local memory based objects that have __attribute__((align(N))) annotations are dependent on the stack pointer alignment. Since the initial stack pointer is a user defined value (via the linker control file), it is the user's responsibility to ensure that the stack pointer is aligned to at least the maximum value of the required local object alignments.
For example, the following structure member definition aligns all definitions of struct S3 on an 8-byte boundary, where a
is at offset 0 and b
is at offset 8.
struct S3 {
char a;
int b __attribute__ ((aligned (8)));
};
struct S3 s3;
The following struct member definition aligns all definitions of struct S4
on a 4-byte boundary, where a
is at offset 0 and b
is at offset 4.
struct S4 {
char a;
int b __attribute__ ((aligned (2)));
};
struct S4 s4;
NOTE
Specifying
__attribute__ ((aligned (2)))
does not affect the alignment of
S4
because 2 is less than the natural alignment of int
.
21.4.5 Typedef Declaration
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
375
Small Data Area PIC/PID Support
Use the alignment attribute to specify how objects of a specific type should be aligned.
For example, the following typedef declaration aligns all definitions of
T1
on an 8-byte boundary.
typedef int T1 __attribute__ ((aligned (8)));
T1 t1;
The following typedef declaration aligns all definitions of
T2
on an 1-byte boundary.
typedef int T2 __attribute__ ((aligned (1)));
T2 t2;
21.4.6 Bitfields
If your program's structure has bitfields and the alignment does not give you as small a structure as you desire, double-check that you are specifying the smallest integer size for your bitfields.
For example, Listing 21-5 on page 376
would be smaller if it were written as shown in
.
Listing 21-5. Before
typedef struct red {
unsigned a: 1;
unsigned b: 1;
unsigned c: 1;
} red;
Listing 21-6. After
typedef struct red {
unsigned char a: 1;
unsigned char b: 1;
unsigned char c: 1;
} red;
21.5 Small Data Area PIC/PID Support
The basic requirement for position independent code and data in the small data area is, at runtime, maintaining the link time address relationships between the startup code (
.init
) and the
.sdata
and
.sdata2
segments. For example, if the link time addresses are:
.init = 0x00002000
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
376 Freescale Semiconductor, Inc.
Chapter 21 AIOP Code Generation
.sdata2 = 0x00003000
.sdata = 0x00004000 but
.init
somehow is executed at 0x00002500, then those link time addresses must all increment by 0x00000500 for their runtime addresses.
Any segment that does not maintain the address relationship at runtime is considered external and must be addressed with absolute addresses. Segments that do maintain their link time address relationship at runtime are considered internal and must be addressed with PC-relative and SDA-relative addressing.
•
Internal and External Segments and References
•
PIC/PID Linker Command File Directives
•
•
•
Building an SDA PIC/PID Application
•
Internal and External Addressing Modes
21.5.1 Internal and External Segments and References
The linker determines at link time whether code and data segments are external or internal. Internal segments reference their data as far or near offsets of the small data registers r2
and r13
. Their code references are normally PC-relative, but if far code references are needed, they also use offsets of the small data registers.
Internal segments can also reference code and data in other internal segments with the same addressing that they would use for their own code and data.
By default, the linker considers all segments in your application to be internal with the exception of segments that are at absolute addresses. Segments with names such as
.abs.xxxxxxxx
, where xxxxxxxx
is a hex address, are considered external.
External segments reference their data with absolute addressing and code references within the segment may be either PC-relative or absolute. Any other segment must use absolute references to reference code or data in external segments. External segments must reference an internal segment with small data registers for code and data.
Related to external segments are external symbol references. These are symbols, usually linker-generated, that are determined not to be within any segment in your application.
They are referenced with absolute addressing. All symbols in an external segment are considered to be external symbol references.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
377
Small Data Area PIC/PID Support
21.5.2 PIC/PID Linker Command File Directives
A few linker command file directives override PIC/PID related linker default settings:
•
MEMORY
•
INTERNAL_SYMBOL
•
EXTERNAL_SYMBOL
21.5.3 Linker-defined Symbols
The linker-generated start and end symbols that are automatically generated for loadable segments are internal if they are addresses into internal segments, and external if they are for external segments. All other linker defined symbols you create in a LCF are considered external unless you redefine them with
INTERNAL_SYMBOL
. The linker also defines some linker defined symbols for its own use (
Table 21-7. Linker-defined Symbols
Symbol Name
_stack_addr
_stack_end
_heap_addr
_heap_end
_SDA_BASE_
_SDA2_BASE_
_ABS_SDA_BASE_
_ABS_SDA2_BASE_
_nbfunctions
SIZEOF_HEADERS top of the stack top of the heap
Value
bottom of the stack bottom of the heap
.sdata + 0x00008000
.sdata2 + 0x00008000
.sdata + 0x00008000
.sdata2 + 0x00008000 number of functions in program size of the segment headers
Description
External. Comes from settings panel settings.
External. Comes from settings panel settings.
External. Comes from settings panel settings.
External. Comes from settings panel settings.
Internal per EABI requirement. May not be redefined.
Internal per EABI requirement. May not be redefined.
External version of _SDA_BASE_ that can be used as an absolute. May not be redefined.
External version of _SDA2_BASE_ that can be used as an absolute. May not be redefined.
Deprecated. External. This is a number, not an address. May not be redefined.
External. This is a number, not an address. May not be redefined.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
378 Freescale Semiconductor, Inc.
Chapter 21 AIOP Code Generation
NOTE
The symbols
_SDA_BASE_
and
_SDA2_BASE
are not accessible until the small data registers are properly initialized before being accessible. The symbols
_ABS_SDA_BASE
and
_ABS_SDA2_BASE
allow you to access those pointers as absolutes addresses, as it is difficult to initialize those pointers without accessing them as absolute addresses.
NOTE
The stack and heap linker generated symbols are external. It may be more practical in a SDA PIC/PID application to make the heap and stack be contiguous with an internal segment and define them as internal.
21.5.4 Uses for SDA PIC/PID
The PIC/PID runtime can be used for different scenarios:
1. All code and data segments are internal. The simplest case would be for all segments to use the same
MEMORY
directive and to have all of the
.bss
type segments at the end.
In such a simple case, the application could be converted to a binary file and linked into another application which could copy it to RAM and jump to its entry point.
2. All of the essential segments are internal and therefore moveable. But, there may be some external segments which are absolute. This situation is probably difficult to test but we can download the entire application to the chip and at least debug it at its link time addresses.
3. There are internal and external segments, but the application is linked as a ROM image (the application does not need to be flashed to ROM, however). It is possible to change the ROM Image Address to be an address into RAM and have the debugger download the image to the RAM address. Alternatively, we could have the
ROM image converted to a binary file and linked into another application as in 1, above. The structures used in
__init_data()
,
_rom_copy_info
and
__bss_init_info
, have been modified for SDA PIC/PID to have an extra field which tells the runtime where the segment is internal or external so that the internal segments are copied to position-relative addresses and the external segments copied to absolute addresses.
21.5.5 Building an SDA PIC/PID Application
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
379
Small Data Area PIC/PID Support
To build a SDA PIC/PID application, select SDA PIC/PID in the ABI list box in the
CodeWarrior IDE's Properties > C/C++ Build > Settings > Tool Settings > CPU panel.
The compiler defines a simple variable that we can use to guard PIC/PID source.
#if __option(sda_pic_pid) // is true if we have chosen SDA
PIC/PID ABI
If our application contains absolute addressing relocations, you will receive linker warnings telling us that those relocations may cause a problem. To resolve these warnings, either:
• change the Code Model listbox in the CodeWarrior IDE's Properties > C/C++
Build > Settings > Tool Settings > CPU panel to be SDA Based PIC/PID
Addressing for all of our sources and libraries
• check the Tune Relocations checkbox in the Properties > C/C++ Build > Settings
> Tool Settings > CPU panel. This new option is only available for the EABI and
SDA PIC/PID ABIs. For EABI, it changes 14-bit branch relocations to 24-bit branch relocations, but only if they can not reach the calling site from the original relocation.
For SDA PIC/PID, this option changes absolute-addressed references of data from code to use a small data register instead of r0
and changes absolute code-to-code references to use the PC-relative relocations.
21.5.5.1 Linking Assembly Files
It is always possible to link in an assembly file that does not behave in a standard way.
For example, taking the address of a variable with: addis rx,r0,object@h ori rx,rx,objec@l generally can not be converted by the linker to SDA PIC/PID Addressing and the linker will warn us if it finds an occurrence.
The following will work with Absolute Addressing as well as allow the linker to convert the instructions to SDA PIC/PID Addressing: addis rx,r0,object@ha addi rx,rx,objec@l
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
380 Freescale Semiconductor, Inc.
Chapter 21 AIOP Code Generation
Another possible problem may arise if we put constant initialized pointers into a readonly section, thereby not letting the runtime convert the addresses.
21.5.5.2 Modifications to the Section Pragma
The pragma
#pragma section
has been modified to accept far_sda_rel
for the data_mode
and code_mode
options, even if we are not using Code Model SDA Based PIC/PID Addressing.
If we omit these options, the compiler uses the Code Model to determine the appropriate modes.
• Absolute Addressing data_mode = far_abs code_mode = pc_rel
• SDA Based PIC/PID Addressing data_mode = far_sda_rel code_mode = pc_rel
21.5.6 Internal and External Addressing Modes
An address mode is applied to a memory segment as a part of the ROM image or at the executing (or logical) address of the segment. Following address modes can be applied to a memory segment:
• Internal-the segment executes from an address not specified at link time.
• External- the segment must execute from the address specified at the link time.
Consider an example where the segment .foo is a part of ROM Image and will be copied to a RAM location. The link time addresses are:
• ROM = 0x00100000
• RAM = 0x00002000
NOTE
Both the link time addresses can be external or internal.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
381
Small Data Area PIC/PID Support
Also assume that the real time (physical) ROM address is 0x00200000 instead of the link
time specified address 0x00100000. Table 21-8 lists the possible address mode scenarios.
A
B
C
D
Scenario
Table 21-8. Possible addr_mode Scenarios
ROM addr_mode
internal
RAM addr_mode
external internal external external internal external internal
Description
Runtime correctly figures out that the ROM address is
0x00200000 and copies it to
0x00002000
Runtime correctly figures out that the ROM address is
0x00200000 and copies it to
0x00102000
Runtime incorrectly assumes that the ROM address is
0x00100000 and copies it to
0x00002000
Runtime incorrectly assumes that the ROM address is
0x00100000 and copies it to
0x00102000
In the above possible scenarios only A and B are correct. The difference between scenario A and B is that in A, the executing (logical) address of .foo is absolute and that in B, the executing (logical) address of.foo is relative.
Scenario C and D are possible if .foo is flashed to ROM at its correct ROM address and all other segments are at an offset from their link time ROM addresses.
NOTE
.init segment determines the correct address of an application.
If .init is at its link time ROM address, then all the segments in the application will be treated as external.
21.5.6.1 Specifying ROM addr_mode
Use the following directives to specify ROM addr_mode:
• LOAD- To specify an external ROM addr_mode.
• INTERNAL_LOAD-To specify an internal ROM addr_mode.
By default the ROM addresses are external.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
382 Freescale Semiconductor, Inc.
Chapter 21 AIOP Code Generation
21.5.6.2 Specifying RAM addr_mode
Use MEMORY directive and any of the following parameters to specify the RAM addr_mode.
addr_mode = external-To specify an external RAM addr_mode.
addr_mode = internal-To specify an internal RAM addr_mode.
By default the RAM addresses are internal.
For example, RAM : org = 0x000e0000, addr_mode = external will make sections defined in the RAM external.
NOTE
addr_mode is ignored if SDA PIC/PID in the ABI list box in the CodeWarrior IDE's Properties > C/C++ Build > Settings
> Tool Settings > CPU panel is not selected.
21.6 Variable Length Encoding
The Variable Length Encoding (VLE) instruction set architecture is an extension to the instruction set specified in Freescale Semiconductor's Book E Implementation Standard
(EIS) for AIOP. This instruction set adds a few identically operating counterparts to the regular EIS instruction set. But where regular EIS instructions occupy 32 bits and must be aligned to 32-bit boundaries, VLE instructions are either 16 or 32 bits long and can be aligned to 16-bit boundaries. This extra flexibility in instruction encoding and alignment allows the compiler and linker to greatly compress the size of runtime object code with only a small penalty in execution performance.
These topics describe how and when to configure the build tools to generate VLE object code:
•
Processors With VLE Capability
•
•
•
21.6.1 Processors With VLE Capability
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
383
Variable Length Encoding
The VLE (Variable Length Encoding) instruction set is an extension to the instruction set specified in the Freescale Book E Implementation Standard (EIS).
21.6.2 Compiling VLE Instructions
Table 21-9 shows how to control VLE (Variable Length Encoding) code generation.
Table 21-9. Controlling VLE code generation
To control this option from here...
CodeWarrior IDE
C/C++ source code command line
use this setting
__declspec(vle_on) __declspec(vle_off)
-vle
21.6.3 Assembling VLE Instructions
The mnemonics for VLE (Variable Length Encoding) instructions begin with "se_" or
"e_". The compiler's inline assembler recognizes these mnemonics when the compiler is configured to generate VLE object code.
Only a subset of EIS instructions have equivalent VLE instructions. To save you time and effort, the inline assembler can convert regular EIS instructions to equivalent VLE instructions automatically. In other words, the inline assembler can generate VLE object
code from inline assembly statements that use only regular mnemonics. Table 21-10
shows how to control VLE code generation for inline assembly statements.
Table 21-10. Controlling VLE inline assembly
To control this option from here...
CodeWarrior IDE command line -ppc_asm_to_vle
use this setting
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
384 Freescale Semiconductor, Inc.
Chapter 21 AIOP Code Generation
The standalone assembler also recognizes and generates VLE instructions. Table 21-11
shows how to control VLE code generation with the standalone assembler.
Table 21-11. Controlling VLE code generation for the standalone assembler use this setting To control this option from here...
CodeWarrior IDE command line
-ppc_asm_to_vle
To specify that a section containing executable code should be executed in the processor's
VLE mode, use the text_vle
identifier with the
.section
examples.
Figure 21-1. Examples of specifying VLE sections in standalone assembly
.section .text_vle # Section name is .text_vle
.section .text,text_vle # Section name is .text
.section .littletext,text_vle # Section name is .littletext
Unlike the inline assembler, the standalone assembler does not offer the option to convert regular instruction mnemonics to VLE instructions. To perform this conversion automatically, copy and paste standalone assembly source code into a C or C++ source
Figure 21-2. Using automatic VLE instruction conversion
extern asm void my_func(void)
{
nofralloc /* No stack frame. */
/* Paste standalone assembly source code here. */
}
21.6.4 Linking VLE Object Code
A processor capable of executing VLE (Variable Length Encoding) instructions must use separate memory pages for VLE and regular instructions. The compiler and linker ensure this separation by placing executable code that uses VLE instructions and regular instructions in separate object code sections.
To maintain this separation in your own linker command file, specify output sections for
VLE and regular instructions.
shows an example. This linker control file specifies that output sections named
.init_vle
and
.text_vle
should only contain object code that the compiler has tagged with
VLECODE
.
Listing 21-7. Separating VLE and regular object code in the linker's output file
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
385
Building a ROM Image
.init : { } > code
.init_vle (VLECODE) : {
*(.init)
*(.init_vle)
} > code
.text : { } > code
.text_vle (VLECODE) : {
*(.text)
*(.text_vle)
} > code
To save memory space, the linker compresses VLE object code by shortening the gaps between functions. A VLE function must meet these criteria to be re-aligned:
• The VLE function is referred to only by other VLE functions.
The linker will not re-align a function if it is referred to by a non-VLE function.
• The VLE function's alignment is 4 bytes.
The linker will not re-align a function if the compiler's function alignment settings specify an explicit alignment value.
• The object code was generated by the CodeWarrior compiler.
21.7 Building a ROM Image
The CodeWarrior compiler and linker can generate a program image that may be stored in and started from ROM (read-only memory). This section uses the term ROM to mean any kind of persistent main storage, including ROM and flash memory.
To create an image for read-only memory, you must configure the compiler and linker:
•
•
•
•
•
Specifying Jump Table Location
•
Specifying Constant Data Location
21.7.1 Linking a ROM Image
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
386 Freescale Semiconductor, Inc.
Chapter 21 AIOP Code Generation
The table below compares the differences between the linker's default RAM image layout and how you should configure your program for loading into and running from ROM.
Table 21-12. Comparing RAM and ROM images
RAM image properties
The S record file contains executable code, constants, and initialization values for data.
Executable code, constants, and data are loaded for execution by the debugger or the program loader.
ROM image properties
One or more memory areas defined in the linker's .lcf file specifies where store the program image in the target system's memory map.
Initialization values for data are copied from the ROM image to RAM at program startup. Executable code and constant data may also be copied to RAM to improve performance
(while requiring more RAM space).
The linker's output ELF file for a ROM image contains a segment named .PPC.EMB.seginfo. This segment describes which segments in the image will be copied from ROM to RAM at runtime. The linker uses this non-loadable segment to generate a data structure named
_rom_copy_info
. At startup, the program uses the
_rom_copy_info
structure to determine which segments to move from ROM to RAM.
Figure 21-3 shows the part of an example disassembly that lists the contents of
segment .PPC.EMB.seginfo. When is_rom_copy is set to 1, the corresponding segment is copied from ROM to its final destination during startup. In this example, these sections will be copied from ROM to RAM at startup:
.bss, .data, .sdata, .sbss, .sdata2.
Figure 21-3. Example of segments to copy to RAM at startup
entry is_rom_copy name ram index
[ 0] 0 .abs.00010000 0
[ 1] 0 .reset 0
[ 2] 0 .init 0
[ 3] 0 .text 0
[ 4] 0 .rodata 0
[ 5] 0 .dtors 0
[ 6] 1 .bss 7
[ 7] 0 .bss 0
[ 8] 1 .data 9
[ 9] 0 .data 0
[ 10] 1 .sdata 11
[ 11] 0 .sdata 0
[ 12] 1 .sbss 13
[ 13] 0 .sbss 0
[ 14] 1 .sdata2 15
[ 15] 0 .sdata2 0
21.7.2 ROM Image Addresses
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
387
Building a ROM Image
The program ROM image is usually the whole image of the program. The ROM image allocates RAM space for its variables, and optionally for its executable code, at application startup. A ROM image is defined by these addresses:
• ROM image address: is the address where you want the ROM image to be allocated.
Usually it is the start address of one of the memory blocks defined in the linker
.lcf
file.
• RAM buffer address: specifies the address in RAM that is to be used as a buffer for the flash image programmer.
NOTE
ROM Image address needs to be synchronized with the LCF specified ROM address to avoid conflict between execution address (start address) and load address (ROM address). The linker will pick LCF specified address as the execution address
(start address).
To specify these addresses in the CodeWarrior IDE, turn on Generate ROM Image in the Linker settings panel, then enter addresses in the RAM Buffer Address of ROM
Image and ROM Image Address fields on the Properties > C/C++ Build > Settings >
Tool Settings > Linker > Output panel.
21.7.3 Specifying A Single ROM Block
When specifying a single ROM memory block in a
.lcf
file, the start address of this memory block can be used as ROM image address. All executable code and constant sections will be allocated in ROM and all variables initialization values will be copied from ROM to RAM during startup.
shows an example
.lcf
file for a single ROM block.
Listing 21-8. Configuring a linker file for a ROM image
MEMORY {
ram : org = 0x00c02000
rom : org = 0x00000000 // desired ROM address (boot
// address for 555)
}
SECTIONS {
.reset : {} > rom
.init : {} > rom
GROUP : {
.text (TEXT) ALIGN(0x1000) : {}
.rodata (CONST) : {
*(.rdata)
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
388 Freescale Semiconductor, Inc.
Chapter 21 AIOP Code Generation
*(.rodata)
}
.ctors : {}
.dtors : {}
extab : {}
extabindex : {}
} > rom // for ROM images, this can be 'rom' if you want
// to execute in ROM or 'ram' if you want to
// execute in RAM
GROUP : {
.data : {}
.sdata : {}
.sbss : {}
.sdata2 : {}
.sbss2 : {}
.bss : {}
.PPC.EMB.sdata0 : {}
.PPC.EMB.sbss0 : {}
} > ram
}
21.7.4 Specifying Several ROM Blocks
To specify several ROM blocks in a
.lcf
file, the start address of the main memory block must be the ROM image address.
To prevent all executable code or constants allocated in other ROM blocks to be copied during startup, use the
LOAD
linker directive. To prevent a specific executable code or constant section from being copied to its runtime RAM destination, specify the final destination address in the
LOAD
directive.
shows an example
.lcf
file that specifies several ROM blocks.
Listing 21-9. Configuring linker file for an image with several ROM blocks
MEMORY
{
APPL_INT_VECT : org= 0x00010000, len= 0x000000FF
// If org is changed, make sure to adjust start address in
// .applexctbl LOAD (0x00010000): {} > APPL_INT_VECT
// accordingly
CST_DATA : org= 0x00010100, len= 0x000000FF
APPL_CODE_FLASH : org= 0x00010200, len= 0x000EFE00
// APPL_CODE_FLASH= int. flash area for application
// external RAM
EXT_RAM_A : org= 0x00800000, len= 0x00100000
}
SECTIONS {
.applexctbl LOAD (0x0001000): {} > APPL_INT_VECT
.syscall: {} > APPL_CODE_FLASH
.reset : {} > APPL_CODE_FLASH
.init: {} > APPL_CODE_FLASH
GROUP : {
.text (TEXT) : {}
.rodata (CONST) : {
*(.rdata)
*(.rodata)
}
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
389
Building a ROM Image
.ctors : {}
.dtors : {}
extab : {}
extabindex : {}
} > APPL_CODE_FLASH
GROUP : {
.bss : {}
.data : {}
.sdata : {}
.sbss : {}
.sdata2 : {}
.sbss2 : {}
.PPC.EMB.sdata0: {}
.PPC.EMB.sbss0 : {}
} > EXT_RAM_A //DPTRAM_AB
GROUP:{
.CstData LOAD (0x00010100): {}
} > CST_DATA
}
If several sections must be allocated in one of the secondary memory areas, use the linker's
ROMADDR
directive to evaluate the final destination address of the sections.
21-10 on page 390 shows an example.
Listing 21-10. Placing several sections in a secondary memory area
.applexctbl LOAD (0x0010000): {} > APPL_INT_VECT
.syscall LOAD (ROMADDR(.applexctbl)+SIZEOF(.applexctbl)):{}
> APPL_INT_VECT
If the program contains an absolute code section, a section which contains object code that must not be copied at startup, the section must also be specified in the
.lcf
file with the
LOAD
directive.
shows example C source code that generates an interrupt service routine that must be placed at a specific address at runtime.
Listing 21-12 on page 390 shows the linker directives that ensure that this routine's
object code will be loaded at a specific address at runtime.
Listing 21-11. Absolute code example
#pragma push
#pragma section code_type ".abs.00010000" code_mode=pc_rel asm void _ISRVectorTable(void)
{
b InterruptHandler
nop
nop
b InterruptHandler
}
#pragma pop
Listing 21-12. Linker commands for absolute code in ROM
MEMORY
{
//internal Flash
APPL_INT_VECT : org= 0x00010000, len= 0x000000FF;
// If org is changed, make sure to adjust start
// address in .abs.00010000 LOAD (0x00010000): {} >
// APPL_INT_VECT accordingly
// ...
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
390 Freescale Semiconductor, Inc.
Chapter 21 AIOP Code Generation
}
SECTIONS {
.abs.00010000 LOAD (0x00010000): {} > APPL_INT_VECT
<...>
}
21.7.5 Specifying Jump Table Location
By default the CodeWarrior compiler and linker allocate jump tables for switch
statements in RAM. When the application executes from ROM, it is sometimes better to have the switch table allocated in ROM to reduce RAM requirements.
To tell the compiler and linker to place jump tables in an object code section that will be placed in ROM, use this directive in your C or C++ source code:
#pragma read_only_switch_tables on
Alternately, to tell the compiler to generate a branch tree in executable code instead of a jump table in a data section, use this directive in C or C++:
#pragma switch_tables off
21.7.6 Specifying Constant Data Location
By default, the CodeWarrior compiler allocates all constant values of a size greater than 8 bytes in the
.rodata
section. There are two solutions for storing constants with sizes smaller than 8 bytes to be allocated in this section:
Solution 1: Define the variable in section
.rodata
using the __declspec(section) directive in C or C++ source code.
shows an example.
Listing 21-13. Using __declspec(section) to store small constants in .rodata
#define SMALL_ROM_CONST __declspec(section ".rodata")
SMALL_ROM_CONST const unsigned int MyInt2 = 0x4534
21.8 Embedded C++
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
391
Embedded C++
Embedded C++ (EC++) is a subset of the ISO/IEC 14882:2003 C++ language that is intended to compile into smaller, faster executable code suitable for embedded systems.
Embedded C++ source code is upwardly compatible with ISO/IEC C++ source code.
•
•
Differences Between ISO C++ and EC++
•
21.8.1 Activating EC++
The table below describes the settings used to control Embedded C++ conformance.
Table 21-13. Controlling Embedded C++ conformance
To control this option from here...
CodeWarrior IDE source code command line
use this setting
EC++ Compatibility Mode in the C/C++ Language Settings panel
#pragma ecplusplus
-dialect ec++
To test for EC++ compatibility mode at compile time, use the
__embedded_cplusplus predefined symbol.
21.8.2 Differences Between ISO C++ and EC++
The EC++ proposal does not support the following ISO/IEC 14882:2003 C++ features:
•
•
•
•
•
•
21.8.2.1 Templates
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
392 Freescale Semiconductor, Inc.
Chapter 21 AIOP Code Generation
ISO/IEC C++ specifies templates. The EC++ proposal does not include template support for class or functions.
21.8.2.2 Libraries
The EC++ proposal supports the
<string>
,
<complex>
,
<ios>
,
<streambuf>
,
<istream>
, and
<ostream>
classes, but only in a non-template form. The EC++ specifications do not support any other ISO/IEC C++ libraries, including the STL-type algorithm libraries.
21.8.2.3 File Operations
The EC++ proposal does not support any file operations except simple console input and output file types.
21.8.2.4 Localization
The EC++ proposal does not contain localization libraries because of the excessive memory requirements.
21.8.2.5 Exception Handling
The EC++ proposal does not support exception handling.
21.8.2.6 Unsupported Language Features
The EC++ proposal does not support the following language features:
• mutable specified
• RTTI
• namespace
• multiple inheritance
• virtual inheritance
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
393
Embedded C++
21.8.3 EC++ Specifications
Topics in this section describe how to design software that adhere to the EC++ proposal:
•
•
21.8.3.1 Language Related Issues
To make sure your source code complies with both ISO/IEC 14882:2003 C++ and EC++ standards, follow these guidelines:
• Do not use RTTI (Run Time Type Identification).
• Do not use exception handling, namespaces, or other unsupported features.
• Do not use multiple or virtual inheritance.
21.8.3.2 Library-Related Issues
Do not refer to routines, data structures, and classes in the Main Standard Library (MSL) for C++.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
394 Freescale Semiconductor, Inc.
Chapter 22
Libraries and Support Code
CodeWarrior software includes libraries and support files you can add to your project.
The sections of this chapter are:
•
Third Party Standard Libraries
•
•
22.1 Third Party Standard Libraries
You might be able to use a third-party C standard library with your CodeWarrior tools.
To determine if the CodeWarrior tools will generate object code that is compatible with a third-party library, compare the file stdarg.h
from the third-party library with stdarg.h
from the MSL library. The CodeWarrior C/C++ compiler for processors uses the MSL files stdarg.h
and runtime library file
__va_arg.c
to generate variable-length parameter functions. Your third-party library must be compatible with these files.
You cannot use a third-party standard C++ library with your CodeWarrior product.
22.2 Embedded Warrior Library
This section explains how to use the AIOP version of the Embedded Warrior Library
(EWL).
•
Using the Embedded Warrior Libraries
•
•
How to Rebuild the EWL Libraries
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
395
Embedded Warrior Library
For more information refer to the EWL C Reference and the EWL C++ Reference.
22.2.1 Using the Embedded Warrior Libraries
The Embedded Warrior Libraries (EWL) are a complete, configurable set of C and C++ standard libraries. The libraries also include EWL Extras, which extends the standard library and adds compatibility with common Linux libraries. All of the source files required to build EWL are included in your CodeWarrior product, along with makefiles to rebuild the libraries, if desired.
A reference implementation of a runtime library is also provided, although the amount of runtime support required for your AIOP project will vary.
22.2.2 EWL Naming Convention
Each archive name has 3 pieces: prefix, core and flags. Following are the details of each piece:
• The prefix is one of the following:
• libm_
- mean that the library has math features
• librt_
- mean that the library has runtime features
• libc_
- mean that the library has reduced code size C features
• libc99_
- mean that the library has faster and increased C99 conformant C features
• libstdc++_
- mean that the library has latest C++ features
• libc++_
- mean that the library has reduced code size C++ features
• The core starts with the processor family (like z490 or z4201) and optionally ends with the core name.
• Current flags are VLE, Soft (software floating point).
NOTE
EWL can only build unsigned char libraries. CodeWarrior no longer provides signed char library in EWL as it is not
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
396 Freescale Semiconductor, Inc.
Chapter 22 Libraries and Support Code
compliant with EABI. If required, users must build their own signed char libraries.
EWL Library Name
libc_z490_VLE_Soft.a
libc_z4201_VLE_Soft.a
libc99_z490_VLE_Soft.a
libc99_z4201_VLE_Soft.a
libstdc++_z490_VLE_Soft.a
libstdc++_z4201_VLE_Soft.a
libm_z490_VLE_Soft.a
libm_z4201_VLE_Soft.a
librt_z490_VLE_Soft.a
librt_z4201_VLE_Soft.a
libconsole.a
_start_z490.a
_start_z4201.a
Table 22-1. EWL AIOP Libraries
Purpose
EWL C library for z490 (AIOP) device
EWL C library for z4201 device
EWL C99 library for z490 (AIOP) device
EWL C99 library for z4201 device
EWL C++ library for z490 (AIOP) device
EWL C++ library for z4201 device
EWL math library for z490 (AIOP) device
EWL math library for z4201 device
EWL Runtime library for z490 (AIOP) device
EWL Runtime library for z4201 device
IO Support library (simulator only)
_start symbol and hardware init for z490
_start symbol and hardware init for z4201
22.2.3 How to Rebuild the EWL Libraries
The EWL library files are present in the ewl\lib
folder. To rebuild the EWL library files, perform the following steps:
NOTE
Ensure that you have access to a make
utility, before rebuilding the EWL libraries. A make utility is provided in the installation at: installDir/CW_APP/gnu/bin.
22.3 Runtime Initialization
The
__start_z490.c
and
__start_z4201.c
provide a reference implementation for program startup. These files provide the first entry point,
__start
which is used by the linker as a program entry point. It initializes the program state (registers, stack pointer, data pointers) and other initializations.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
397
Runtime Initialization
void __start(void)
Main entry point for the EABI program. It will call main()
. The
__start_z490.c
and
__start_z4201.c
files allow for board-level initialization and user-level initializations.
If your project does not use exceptions, then the initialization code for them is stripped out. If exceptions are used, the linker will add them automatically.
Interface for board-level initialization and user-level initialization
There are 3 main methods to customize the startup for your needs:
• Modify this file and rebuild the
__start_xxx.a
library or
• Copy this C file to your project and customize it to your needs
• Override the functions in your sources (interface are declared weak meaning a nonweak definition will override the implementation here)
The code in
__start_z490.c
or
__start_z4201.c
is by default placed in the ".init" section. The
Linker control file (LCF) must place this section in order to use the default implementations.
Note that
__init_hardware
should not write on the stack until the memory controller is properly configured if applicable to your target. The following interfaces are defined:
• Stack pointer, small data pointer, and target specific register initializations void __init_registers(void)
• Target specific hardware initializations void __init_hardware(void)
• BSS initialization and ROM-to-RAM copy if applicable void __init_data(void)
• CodeWarrior C++ initialization before calling main(). Calls constructors of which the cpp exception handling initialization is the first constructor, if exceptions are used void __init_cpp(void)
• Allows you to perform initialization before calling main() void __init_user(void)
• Local implementation of ANSI abort void abort(void)
• Local implementation of ANSI exit void exit(void)
Other interfaces (cannot be overridden):
• C++ static initialization void __init_cpp(void)
• Calls destructors
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
398 Freescale Semiconductor, Inc.
Chapter 22 Libraries and Support Code
void __fini_cpp(void)
• This function simply stalls the debugger. You may want to rewrite this function if you are using an OS void _ExitProcess(void)
Abort and Exit
In order to correctly implement the required startup/termination sequence for C and C++ programs, we need to have an exit()
routine that can be called by the program startup code. The exit()
routine is supposed to:
1. call any functions registered via atexit()
2. call destructors for any global objects
3. flush any unwritten buffers, close any open files, etc.
4. terminates the program
You don't, however, want to require the ANSI C library for every CodeWarrior program, since it drags in lots of code that may not be needed. Instead we provide a dummy exit() function which simply calls the destructors and terminates the program. You can assume that any program which uses atexit()
or
<stdio.h>
and requires those cleanup behaviors will have linked with the ANSI C library, whose definition of exit()
will override the one here.
Similarly, define a dummy abort()
function (which is called by the default terminate() handler).
Programs which rely on the proper ANSI C/C++ behavior must use the ANSI C library, and order it in the CodeWarrior project or command-line so that its definitions supersede these definitions in the runtime support library.
NOTE
The .PPC.EMB.sdata0 and .PPC.EMB.sbss section pointers are not currently supported. Target specific notes
__start_z490.c
specifically copies r2, r13 to DCR469, DCR470.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
399
Runtime Initialization
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
400 Freescale Semiconductor, Inc.
Chapter 23
Declaration Specifications
Declaration specifications describe special properties to associate with a function or variable at compile time. You insert these specifications in the object's declaration.
•
Syntax for Declaration Specifications
•
Syntax for Attribute Specifications
23.1 Syntax for Declaration Specifications
The syntax for a declaration specification is
__declspec(
spec [
options ])
function-declaration ; where spec is the declaration specification, options represents possible arguments for the declaration specification, and function-declaration represents the declaration of the function. Unless otherwise specified in the declaration specification's description, a function's definition does not require a declaration specification.
23.1.1 __declspec(never_inline)
Specifies that a function must not be inlined.
Syntax
__declspec (never_inline)
function_prototype ;
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
401
Syntax for Attribute Specifications
Remarks
Declaring a function's prototype with this declaration specification tells the compiler not to inline the function, even if the function is later defined with the inline
,
__inline__
, or
__inline
keywords.
23.2 Syntax for Attribute Specifications
The syntax for an attribute specification is
__attribute__((
list-of-attributes )) where list-of-attributes is a comma-separated list of zero or more attributes to associate with the object. Place an attribute specification at the end of the delcaration and definition
of a function, function parameter, or variable. Listing 23-1 on page 402 shows an
example.
Listing 23-1. Example of an attribute specification
int f(int x __attribute__((unused))) __attribute__((never_inline)); int f(int x __attribute__((unused))) __attribute__((never_inline))
{
return 20;
}
23.2.1 __attribute__((noreturn))
Specifies that a function will never return if it is called. This allows the compiler to make certain modifications to the control flow graph to improve the code generation. An example of an improvement would be to not conserve the LR or non-volatiles on the stack.
23.2.2 __attribute__((deprecated))
Specifies that the compiler must issue a warning when a program refers to an object.
Syntax
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
402 Freescale Semiconductor, Inc.
Chapter 23 Declaration Specifications
variable-declaration __attribute__((deprecated));
variable-definition __attribute__((deprecated));
function-declaration __attribute__((deprecated));
function-definition __attribute__((deprecated));
Remarks
This attribute instructs the compiler to issue a warning when a program refers to a function or variable. Use this attribute to discourage programmers from using functions and variables that are obsolete or will soon be obsolete.
Listing 23-2. Example of deprecated attribute
int velocipede(int speed) __attribute__((deprecated)); int bicycle(int speed); int f(int speed)
{
return velocipede(speed); /* Warning. */
} int g(int speed)
{
return bicycle(speed * 2); /* OK */
}
23.2.3 __attribute__((force_export))
Prevents a function or static variable from being dead-stripped.
Syntax
function-declaration __attribute__((force_export));
function-definition __attribute__((force_export));
variable-declaration __attribute__((force_export));
variable-definition __attribute__((force_export));
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
403
Syntax for Attribute Specifications
Remarks
This attribute specifies that the linker must not dead-strip a function or static variable even if the linker determines that the rest of the program does not refer to the object.
23.2.4 __attribute__((malloc))
Specifies that the pointers returned by a function will not point to objects that are already referred to by other variables.
Syntax
function-declaration __attribute__((malloc));
function-definition __attribute__((malloc));
Remarks
This attribute specification gives the compiler extra knowledge about pointer aliasing so that it can apply stronger optimizations to the object code it generates.
23.2.5 __attribute__((noalias))
Prevents access of data object through an indirect pointer access.
Syntax
function-parameter __attribute__((noalias));
variable-declaration __attribute__((noalias));
variable-definition __attribute__((noalias));
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
404 Freescale Semiconductor, Inc.
Chapter 23 Declaration Specifications
This attribute specifies to the compiler that a data object is only accessed directly, helping the optimizer to generate a better code. The sample code in
not return a correct result if ip
is pointed to a
.
Listing 23-3. Example of the noalias attribute
extern int a __attribute__((noalias)); int f(int *ip)
{
a = 1;
*ip = 0;
return a; // optimized to return 1;
}
23.2.6 __attribute__((returns_twice))
Specifies that a function may return more than one time because of multithreaded or nonlinear execution.
Syntax
function-declaration __attribute__((returns_twice));
function-definition __attribute__((returns_twice));
Remarks
This attribute specifies to the compiler that the program's flow of execution might enter and leave a function without explicit function calls and returns. For example, the standard library's setjmp()
function allows a program to change its execution flow arbitrarily.
With this information, the compiler limits optimizations that require explicit program flow.
23.2.7 __attribute__((unused))
Specifies that the programmer is aware that a variable or function parameter is not referred to.
Syntax
function-parameter __attribute__((unused));
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
405
Syntax for Attribute Specifications
variable-declaration __attribute__((unused));
variable-definition __attribute__((unused));
Remarks
This attribute specifies that the compiler should not issue a warning for an object if the object is not referred to. This attribute specification has no effect if the compiler's unused warning setting is off.
Listing 23-4. Example of the unused attribute
void f(int a __attribute__((unused))) /* No warning for a. */
{
int b __attribute__((unused)); /* No warning for b. */
int c; /* Possible warning for c. */
return 20;
}
23.2.8 __attribute__((used))
Prevents a function or static variable from being dead-stripped.
Syntax
function-declaration __attribute__((used));
function-definition __attribute__((used));
variable-declaration __attribute__((used));
variable-definition __attribute__((used));
Remarks
This attribute specifies that the linker must not dead-strip a function or static variable even if the linker determines that the rest of the program does not refer to the object.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
406 Freescale Semiconductor, Inc.
Chapter 23 Declaration Specifications
23.2.9 __attribute__((never_inline))
Specifies that a function must not be inlined.
Syntax
__attribute__((never_inline)) function_prototype ;
Remarks
Declaring a function's prototype with this attribute tells the compiler not to inline the function, even if the function is later defined with the inline
,
__inline__
, or
__inline keywords.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
407
Syntax for Attribute Specifications
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
408 Freescale Semiconductor, Inc.
Chapter 24
Declaration Specifications for AIOP Code
Declaration specifications describe special attributes to associate with a function at compile time. For example, these attributes might change how the compiler translates the function, describe what properties it has, or pass other information to the compiler.
•
Syntax for Declaration Specifications
•
24.1 Syntax for Declaration Specifications
The syntax for a declaration specification is
__declspec(
spec [
options ])
function-prototype ; where spec is the declaration specification, options represents possible arguments for the declaration specification, and function-prototype represents the declaration of the function. Unless otherwise specified in the declaration specification's description, a function's definition does not require a matching declaration specification.
24.2 Declaration Specifications
This section explains declaration specifications supported by the current release.
•
•
•
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
409
Declaration Specifications
•
•
•
•
•
•
24.2.1 __declspec(do_not_merge)
Specifies that a function must not be removed from object code during code merging optimization.
Syntax
__declspec(do_not_merge)
function-declaration ;
Remarks
When you declare or define a function with this declaration specification the linker does not consider this function when applying the code merging optimization. This declaration specification ensures that the linker will not remove a function from object code during code merging if another function with identical object code exists.
Use this declaration specification for functions that your program refers to with function pointers.
24.2.2 __declspec(final)
Provides Java-style function override checking.
Syntax
__declspec(final)
declaration
Example
struct A { virtual __declspec(final) void vf1(); }; struct B : A { void vf1(); /*
Error : the final function A::vf() is overridden by B::vf1() */ };
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
410 Freescale Semiconductor, Inc.
Chapter 24 Declaration Specifications for AIOP Code
24.2.3 __declspec(force_export)
Specifies that a function or variable must not be dead-stripped.
Syntax
__declspec(force_export)
function-declaration ;
__declspec(force_export)
variable-declaration ;
Remarks
When the linker determines that a function or static variable is not referred to by the rest of the program, the linker removes the object from the final executable image. The linker does not remove a static variable or function declared with this specification even if this object is not referred to.
24.2.4 __declspec(interrupt)
Controls the compilation of object code for interrupt service routines.
Syntax
__declspec (interrupt [
option [
optionN ]]) void
__InterruptHandler__(void);
__declspec (interrupt [
option [
optionN ]]) void
__InterruptHandler__(void)
{
/* ... */
} where option and optionN are zero or more of the following choices:
NOTE
If no choice is specified, save_spe
is enabled.
enable
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
411
Declaration Specifications
Enables interrupts while this interrupt handler is executing.
SRR
Saves the appropriate save/restore register (SRR0 or SRR1) on the stack.
DAR
Saves the data address register on the stack.
DSISR
Saves the DSI status register on the stack.
save_fprs
Saves the floating-point registers on the stack.
nowarn
Does not issue a warning message if the function being compiled as an interrupt service routine is larger than the processor's interrupt vector area (256 bytes/64 instructions).
noncritical | critical | machine | debug
Specify the type of interrupt service routine to generate. Specifically, the compiler choose a return instruction based on this choice. The default is noncritical.
vle_multiple
Enables generation of new VLE instructions in the interrupt prolog/epilog in C-functions by the compiler. Compile error will be returned if this option is used for any other cores.
When enabled, the compiler:
• emits e_lmvgprw and e_stmvgprw if any of [r3,r12] needs to be saved to stack
• emits e_lmvsprw and e_stmvsprw if any of the CR, LR, CTR and XER needs to be saved to stack
• emits e_lmvsrrw and e_stmvsrr if any of the SRR0 and SRR1 needs to be saved to stack
• emits e_lmvcsrrw and e_stmvcsrrw if any of the CSRR0 and CSRR1 needs to be saved to stack
• emits e_lmvdsrrw and e_stmvdsrrw if any of the DSRR0 and DSRR1 needs to be saved to stack.
NOTE
Enabling vle_multiple could increase the space being taken up by the stack, especially when saving the volatile gprs.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
412 Freescale Semiconductor, Inc.
Chapter 24 Declaration Specifications for AIOP Code
Remarks
When you declare or define a function with this declaration specification the compiler generates a special prologue and epilogue for the function so that it can respond to a processor interrupt. For convenience, the compiler also marks interrupt functions so that the linker does not dead-strip them.
24.2.5 __declspec(no_linker_opts)
Specifies that the linker must not apply its optimizations to a function.
Syntax
__declspec(no_linker_opts)
function-declaration
24.2.6 __declspec(section name)
Specifies where to store a variable or function in a section that has been predefined or defined with the
#pragma section
directive.
Syntax
__declspec(section <section_name>) declaration
__declspec(section <section_name>) definition
Parameters
section_name
Specifies the name of an initialized data section.
NOTE
The name of a section must be enclosed in double-quotes ("").
To use a user defined section, create the section using #pragma section directive before using __declspec(section
<section_name>).
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
413
Declaration Specifications
• When you specify an uninitialized section name while declaring or defining a function or variable with this declaration specification, the compiler generates an error.
For example, if you use __declspec (section ".bss") extern int my_var;, where .bss is an uninitialized section you will get a descriptive error. In this case, use __declspec
(section ".data") extern int my_var; as .data is normally paired with .bss and .data is an initialized section. Assuming the variable you are attaching this __declspec to is an uninitialized object (which is the case with my_var), the object will go into .bss.
• When you use __declspec on the definition, __declspec on a declaration is ignored.
The listing below shows an example of __declspec used to put data and code objects into specific sections.
Figure 24-1. Example of a declaration specification
__declspec (section ".init") extern void cache_init1(void);
__declspec (section ".text") extern void cache_init2(void); extern void cache_init3(void); void cache_init1(){} // goes into .init if the prototype is visible for this definition
__declspec (section ".init") void cache_init2(){} // ignores previous section .text and goes into .init
__declspec (section ".init") void cache_init3(){} // by default the declaration implies .text but the definition forces it to .init
Predefined sections and default sections
The predefined sections set with an object type become the default section for that type.
The compiler predefines the sections listed in the table below.
Table 24-1. Predefined sections
Type
code_type data_type const_type sdata_type sconst_type
Name
".text"
".data"
".rodata"
".sdata"
".sdata2" ".sbss2"
".PPC.EMB.sdata0"
".PPC.EMB.sbss0"
Data mode
data_mode=far_abs data_mode=far_abs data_mode=far_abs data_mode=sda_rel data_mode=sda_rel data_mode=sda_rel
Code mode
code_mode=pc_rel code_mode=pc_rel code_mode=pc_rel code_mode=pc_rel code_mode=pc_rel code_mode=pc_rel
NOTE
The "
.PPC.EMB.sdata0
" and "
.PPC.EMB.sbss0
" sections are predefined as an alternative to the sdata_type
object type.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
414 Freescale Semiconductor, Inc.
Chapter 24 Declaration Specifications for AIOP Code
24.2.7 __declspec(vle_off)
Forces the compiler to use the regular instruction set instead of the Variable Length
Encoded (VLE) instruction set.
Syntax
__declspec (vle_off)
function_prototype ;
Remarks
Declaring a function's prototype with this declaration specification tells the compiler to use the regular instruction set. The compiler and linker must arrange such instructions more rigidly than VLE instructions, resulting in larger object code.
For information on the availability of VLE instructions, refer to your processor's documentation.
This declaration specification overrides other compiler settings for VLE code generation.
24.2.8 __declspec(vle_on)
Forces the compiler to use the VLE (Variable Length Encoded) instruction set for a function.
Syntax
__declspec (vle_on)
function _prototype;
Remarks
Declaring a function's prototype with this declaration specification tells the compiler to use the VLE instruction set when generating object code for the function. The compiler and linker can arrange VLE instructions more compactly than regular instructions.
The VLE instruction set extends the regular instruction set defined by the Freescale Book
E Implementation Standard (EIS). For information on the availability of VLE instructions, refer to your processor's documentation.
This declaration specification overrides other compiler settings for VLE code generation.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
415
Declaration Specifications
24.2.9 __declspec(weak)
Makes a global code or data deinition "weak", so that it can be replaced by another definition, in another translation unit.
Syntax
__declspec(weak)
function _declaration;
Remarks
This declaration replaces
#pragma overload
.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
416 Freescale Semiconductor, Inc.
Chapter 25
Predefined Macros
The compiler preprocessor has predefined macros (some refer to these as predefined symbols). The compiler simulates variable definitions that describe the compile-time environment and properties of the target processor.
This chapter lists the predefined macros that all CodeWarrior compilers make available.
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
417
__COMPILER__
25.1 __ASSEMBLER__
Preprocessor macro defined by the CodeWarrior assembler only.
Syntax
__ASSEMBLER__
Remarks
This macro can be used to determine if the tool under which a file is processed is the
CodeWarrior assembler.
25.2 __COMPILER__
Preprocessor macro defined by the CodeWarrior compiler only.
Syntax
__COMPILER__
Remarks
This macro can be used to determine if the tool under which a file is processed is the
CodeWarrior compiler.
25.3 __COUNTER__
Preprocessor macro that expands to an integer.
Syntax
__COUNTER__
Remarks
The compiler defines this macro as an integer that has an initial value of 0 incrementing by 1 every time the macro is used in the translation unit.
The value of this macro is stored in a precompiled header and is restored when the precompiled header is used by a translation unit.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
418 Freescale Semiconductor, Inc.
Chapter 25 Predefined Macros
25.4 __TARGETARCH__
Preprocessor macro defined as the eRISC architecture mode.
Syntax
__TARGETARCH__
Remarks
This macro can be used to determine the assembly instructions to use in the eRISC architecture.
25.5 __VERSION_NUM__
Preprocessor macro defined as the CodeWarrior tool version number.
Syntax
__VERSION_NUM__
Remarks
This macro can be used to determine in which tool version we are using.
Both Compiler and Assembler tools support this macro (the value returned from this macro depends on the tool compiling the file in which the macro lies).
This macro is comparable, and returns a numeric comparable value (i.e. version 10.0.0.1
returns the value: 0x10000001).
25.6 __VERSION__
Preprocessor macro defined as the CodeWarrior compiler version string.
Syntax
__VERSION__
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
419
__cplusplus
Remarks
This macro can be used to determine in which compiler version we are using.
This macro returns a non-comparable string value (i.e. version 10.0.0.1 returns the value:
10.0.0.1).
25.7 __cplusplus
Preprocessor macro defined if compiling C++ source code.
Syntax
__cplusplus
Remarks
The compiler defines this macro when compiling C++ source code. This macro is undefined otherwise.
25.8 __CWBUILD__
Preprocessor macro defined as the build number of the CodeWarrior compiler.
Syntax
__CWBUILD__
Remarks
The ISO standards do not specify this symbol.
25.9 __CWCC__
Preprocessor macro defined as the version of the CodeWarrior compiler frontend.
Syntax
__CWCC__
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
420 Freescale Semiconductor, Inc.
Chapter 25 Predefined Macros
Remarks
CodeWarrior compilers issued after 2006 define this macro with the compiler's frontend version. For example, if the compiler frontend version is 4.2.0, the value of
__CWCC__
is
0x4200
.
CodeWarrior compilers issued prior to 2006 used the pre-defined macro
__MWERKS__
. The
__MWERKS__
predefined macro is still functional as an alias for
__CWCC__
.
The ISO standards do not specify this symbol.
25.10 __DATE__
Preprocessor macro defined as the date of compilation.
Syntax
__DATE__
Remarks
The compiler defines this macro as a character string representation of the date of compilation. The format of this string is
"
Mmm dd yyyy " where Mmm is the a three-letter abbreviation of the month, dd is the day of the month, and yyyy is the year.
25.11 __embedded_cplusplus
Defined as 1 when compiling embedded C++ source code, undefined otherwise.
Syntax
__embedded_cplusplus
Remarks
The compiler defines this macro as 1 when the compiler's settings are configured to restrict the compiler to translate source code that conforms to the Embedded C++ proposed standard. The compiler does not define this macro otherwise.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
421
__FILE__
25.12 __FILE__
Preprocessor macro of the name of the source code file being compiled.
Syntax
__FILE__
Remarks
The compiler defines this macro as a character string literal value of the name of the file being compiled, or the name specified in the last instance of a
#line
directive.
25.13 __func__
Predefined variable of the name of the function being compiled.
Prototype
static const char __func__[] = "
function-name ";
Remarks
The compiler implicitly defines this variable at the beginning of each function if the function refers to
__func__
. The character string contained by this array, function-name , is the name of the function being compiled.
This implicit variable is undefined outside of a function body. This variable is also undefined when C99 (ISO/IEC 9899-1999) or GCC (GNU Compiler Collection) extension settings are off.
25.14 __FUNCTION__
Predefined variable of the name of the function being compiled.
Prototype
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
422 Freescale Semiconductor, Inc.
Chapter 25 Predefined Macros
static const char __FUNCTION__[] = "
function-name ";
Remarks
The compiler implicitly defines this variable at the beginning of each function if the function refers to
__FUNCTION__
. The character string contained by this array, function-
name
, is the name of the function being compiled.
This implicit variable is undefined outside of a function body.
25.15 __ide_target()
Preprocessor operator for querying the IDE about the active build target.
Syntax
__ide_target("
target_name ") target-name
The name of a build target in the active project in the CodeWarrior IDE.
Remarks
Expands to
1
if target_name is the same as the active build target in the CodeWarrior
IDE's active project. Expands to
0
otherwise. The ISO standards do not specify this symbol.
25.16 __LINE__
Preprocessor macro of the number of the line of the source code file being compiled.
Syntax
__LINE__
Remarks
The compiler defines this macro as a integer value of the number of the line of the source code file that the compiler is translating. The
#line
directive also affects the value that this macro expands to.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
423
__MWERKS__
25.17 __MWERKS__
Deprecated. Preprocessor macro defined as the version of the CodeWarrior compiler.
Syntax
__MWERKS__
Remarks
Replaced by the built-in preprocessor macro
__CWCC__
.
CodeWarrior compilers issued after 1995 define this macro with the compiler's version.
For example, if the compiler version is 4.0, the value of
__MWERKS__
is
0x4000
.
This macro is defined as
1
if the compiler was issued before the CodeWarrior CW7 that was released in 1995.
The ISO standards do not specify this symbol.
25.18 __optlevel
Optimization level exported as a predefined macro.
Syntax
__optlevel0
__optlevel1
__optlevel2
__optlevel3
__optlevel4
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
424 Freescale Semiconductor, Inc.
Chapter 25 Predefined Macros
Using these macros, user can conditionally compile code for a particular optimization level. The following table lists the level of optimization provided by the
__optlevelx macro.
__optlevel0
__optlevel1
__optlevel2
__optlevel3
__optlevel4
Macro
Table 25-1. Optimization Levels
Optimization Level
O0
O1
O2
O3
O4
Example
The listing below shows an example of
__optlevelx
macro usage.
Listing 25-1. Example usage of
__optlevel
macro
int main()
{
#if __optlevel0
... // This code compiles only if this code compiled with Optimization level 0
#elif __optlevel1
... // This code compiles only if this code compiled with Optimization level 1
#elif __optlevel2
... // This code compiles only if this code compiled with Optimization level 2
#elif __optlevel3
... // This code compiles only if this code compiled with Optimization level 3
#elif __optlevel4
... // This code compiles only if this code compiled with Optimization level 4
#endif
}
25.19 __PRETTY_FUNCTION__
Predefined variable containing a character string of the "unmangled" name of the C++ function being compiled.
Syntax
Prototype
static const char __PRETTY_FUNCTION__[] = "
function-name ";
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
425
__profile__
Remarks
The compiler implicitly defines this variable at the beginning of each function if the function refers to
__PRETTY_FUNCTION__
. This name, function-name , is the same identifier that appears in source code, not the "mangled" identifier that the compiler and linker use.
The C++ compiler "mangles" a function name by appending extra characters to the function's identifier to denote the function's return type and the types of its parameters.
The ISO/IEC 14882:2003 C++ standard does not specify this symbol.
25.20 __profile__
Preprocessor macro that specifies whether or not the compiler is generating object code for a profiler.
Syntax
__profile__
Remarks
Defined as 1 when generating object code that works with a profiler. Undefined otherwise. The ISO standards does not specify this symbol.
25.21 __STDC__
Defined as 1 when compiling ISO/IEC Standard C source code, undefined otherwise.
Syntax
__STDC__
Remarks
The compiler defines this macro as 1 when the compiler's settings are configured to restrict the compiler to translate source code that conforms to the ISO/IEC 9899-1990 and ISO/IEC 9899-1999 standards. The compiler does not define this macro otherwise.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
426 Freescale Semiconductor, Inc.
Chapter 25 Predefined Macros
25.22 __TIME__
Preprocessor macro defined as a character string representation of the time of compilation.
Syntax
__TIME__
Remarks
The compiler defines this macro as a character string representation of the time of compilation. The format of this string is
"
hh :
mm :
ss " where hh is a 2-digit hour of the day, mm is a 2-digit minute of the hour, and ss is a 2digit second of the minute.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
427
__TIME__
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
428 Freescale Semiconductor, Inc.
Chapter 26
Predefined Macros for AIOP Compiler
As listed in the previous chapter, the compiler preprocessor has predefined macros. The macros in the previous chapter are available to all architecture. This chapter describes the predefined macros available specifically for the AIOP compiler.
•
•
•
•
•
Processor Core Specific Macros
26.1 __ALTIVEC__
Preprocessor macro defined if language extensions for using AIOP AltiVec™ instructions are available.
Syntax
__ALTIVEC__
Remarks
The compiler defines this macro when pragma altivec_model
is on
. This macro is undefined otherwise.
26.2 __POWERPC__
Preprocessor macro defined to describe the target powerpc processor.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
429
__PPCBROADWAY__
Syntax
__POWERPC__
Remarks
The compiler defines this macro to describe powerpc processor that the compiler is generating code for.
26.3 __PPCBROADWAY__
Preprocessor macro defined if the compiler is generating object code for the AIOP
"Broadway" processor.
Syntax
__PPCBROADWAY__
Remarks
The compiler defines this macro as
1
when generating object code for the "Broadway" processor. The
__PPCGECKO__
macro is also defined.
This macro is undefined otherwise.
26.4 __PPCGECKO__
Preprocessor macro defined if the compiler is generating object code for the AIOP
"Gecko" processor.
Syntax
__PPCGECKO__
Remarks
The compiler defines this macro as
1
when generating object code for the "Gecko" processor. This macro is undefined otherwise.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
430 Freescale Semiconductor, Inc.
Chapter 26 Predefined Macros for AIOP Compiler
26.5 Processor Core Specific Macros
824x
8260
827x
850/85x
860/86x
87x
88x
8280
7450/745x
74xx
750/75x
801
821
823
8240
82xx
5100
5200 e300[c1]
401
403
405
505
509
555
56x
601
602
603
603e
604
604e
7400
740/74x
The compiler defines the macros, discussed in the following table, to describe powerpc processor core that the compiler is generating code for.
Table 26-1. Processor core specific macros
Processor Core Compiler Macro
__PPC7450__
__PPC74xx__
__PPC750__
__PPC801__
__PPC821__
__PPC823__
__PPC8240__
__PPC82xx__
__PPC8240__
__PPC8260__
__PPC827x__
__PPC850__
__PPC860__
__PPC87x__
__PPC88x__
__PPC8280__
__PPC5100__
__PPC5200__
__PPCe300__
__PPC401__
__PPC403__
__PPC405__
__PPC505__
__PPC509__
__PPC555__
__PPC56x__
__PPC601__
__PPC602__
__PPC603__
__PPC603e__
__PPC604__
__PPC604e__
__PPC7400__
__PPC740__
Table continues on the next page...
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
431
Processor Core Specific Macros
Table 26-1. Processor core specific macros (continued)
Processor Core Compiler Macro
e300c2 e300c3 e300c4 e500[v1] e500v2 e600 e700
Zen generic
__PPCe300c2__
__PPCe300c3__
__PPCe300c4__
__PPCe500__
__PPCe500v2__
__PPCe600__
__PPCe700__
__PPCZen__
__PPCGENERIC__
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
432 Freescale Semiconductor, Inc.
Chapter 27
Using Pragmas
The #pragma preprocessor directive specifies option settings to the compiler to control the compiler and linker's code generation.
•
•
Saving and Restoring Pragma Settings
•
Determining Which Settings Are Saved and Restored
•
•
27.1 Checking Pragma Settings
The preprocessor function
__option()
returns the state of pragma settings at compile-time.
The syntax is
__option(
setting-name ) where setting-name is the name of a pragma that accepts the on
, off
, and reset arguments.
If setting-name is on
,
__option(
setting-name
)
returns 1. If setting-name is off
,
__option(
setting-name
)
returns 0. If setting-name is not the name of a pragma,
__option(
setting-name
)
returns false. If setting-name is the name of a pragma that does not accept the on
, off
, and reset
arguments, the compiler issues a warning message.
shows an example.
Listing 27-1. Using the __option() preprocessor function
#if __option(ANSI_strict)
#include "portable.h" /* Use the portable declarations. */
#else
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
433
Saving and Restoring Pragma Settings
#include "custom.h" /* Use the specialized declarations. */
#endif
27.2 Saving and Restoring Pragma Settings
There are some occasions when you would like to apply pragma settings to a piece of source code independently from the settings in the rest of the source file. For example, a function might require unique optimization settings that should not be used in the rest of the function's source file.
Remembering which pragmas to save and restore is tedious and error-prone. Fortunately, the compiler has mechanisms that save and restore pragma settings at compile time.
Pragma settings may be saved and restored at two levels:
• all pragma settings
• some individual pragma settings
Settings may be saved at one point in a compilation unit (a source code file and the files that it includes), changed, then restored later in the same compilation unit. Pragma settings cannot be saved in one source code file then restored in another unless both source code files are included in the same compilation unit.
Pragmas push
and pop
save and restore, respectively, most pragma settings in a compilation unit. Pragmas push
and pop
may be nested to unlimited depth. Listing 27-2 on page 434
shows an example.
Listing 27-2. Using push and pop to save and restore pragma settings
/* Settings for this file. */
#pragma opt_unroll_loops on
#pragma optimize_for_size off void fast_func_A(void)
{
/* ... */
}
/* Settings for slow_func(). */
#pragma push /* Save file settings. */
#pragma optimization_size 0 void slow_func(void)
{
/* ... */
}
#pragma pop /* Restore file settings. */ void fast_func_B(void)
{
/* ... */
}
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
434 Freescale Semiconductor, Inc.
Chapter 27 Using Pragmas
Pragmas that accept the reset
argument perform the same actions as pragmas push
and pop
, but apply to a single pragma. A pragma's on
and off
arguments save the pragma's current setting before changing it to the new setting. A pragma's reset
argument restores the pragma's setting. The on
, off
, and reset
arguments may be nested to an unlimited
depth. Listing 27-3 on page 435
shows an example.
Listing 27-3. Using the reset option to save and restore a pragma setting
/* Setting for this file. */
#pragma opt_unroll_loops on void fast_func_A(void)
{
/* ... */
}
/* Setting for smallslowfunc(). */
#pragma opt_unroll_loops off void small_func(void)
{
/* ... */
}
/* Restore previous setting. */
#pragma opt_unroll_loops reset void fast_func_B(void)
{
/* ... */
}
27.3 Determining Which Settings Are Saved and Restored
Not all pragma settings are saved and restored by pragmas push
and pop
. Pragmas that do not change compiler settings are not affected by push
and pop
. For example, pragma message
cannot be saved and restored.
shows an example that checks if the
ANSI_strict
pragma setting is saved and restored by pragmas push
and pop
.
Listing 27-4. Testing if pragmas push and pop save and restore a setting
/* Preprocess this source code. */
#pragma ANSI_strict on
#pragma push
#pragma ANSI_strict off
#pragma pop
#if __option(ANSI_strict)
#error "Saved and restored by push and pop."
#else
#error "Not affected by push and pop."
#endif
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
435
Pragma Scope
27.4 Invalid Pragmas
If you enable the compiler's setting for reporting invalid pragmas, the compiler issues a warning when it encounters a pragma it does not recognize. For example, the pragma statements in the listing below, generates warning with the invalid pragmas setting enabled.
Listing 27-5. Invalid Pragmas
#pragma silly_data off // WARNING: silly_data is not a pragma.
#pragma ANSI_strict select // WARNING: select is not defined
#pragma ANSI_strict on // OK
The table below lists various options on how to control the recognition of invalid pragmas.
Table 27-1. Controlling invalid pragmas
To control this option from here...
CodeWarrior IDE source code command line
use this setting
Illegal Pragmas in the C/C++ Compiler > Warnings panel
#pragma warn_illpragma
-warnings illpragmas
27.5 Pragma Scope
The scope of a pragma setting is limited to a compilation unit (a source code file and the files that it includes).
At the beginning of compilation unit, the compiler uses its default settings. The compiler then uses the settings specified by the CodeWarrior IDE's build target or in commandline options.
The compiler uses the setting in a pragma beginning at the pragma's location in the compilation unit. The compilers continues using this setting:
• until another instance of the same pragma appears later in the source code
• until an instance of pragma pop
appears later in the source code
• until the compiler finishes translating the compilation unit
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
436 Freescale Semiconductor, Inc.
Chapter 28
Pragmas for Standard C Conformance
Pragmas for Standard C Conformance should be placed outside the functions to which it is desired that they be applied.
This chapter explains the following pragmas:
•
•
•
•
•
•
28.1 ANSI_strict
Controls the use of non-standard language features.
Syntax
#pragma ANSI_strict on | off | reset
Remarks
If you enable the pragma
ANSI_strict
, the compiler generates an error message if it encounters some CodeWarrior extensions to the C language defined by the ISO/IEC
9899-1990 ("C90") standard:
• C++-style comments
• unnamed arguments in function definitions
• non-standard keywords
By default, this pragma is off
.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
437
c99
28.2 c99
Controls the use of a subset of ISO/IEC 9899-1999 ("C99") language features.
Syntax
#pragma c99 on | off | reset
Remarks
If you enable this pragma, the compiler accepts many of the language features described by the ISO/IEC 9899-1999 standard:
• More rigid type checking.
• Trailing commas in enumerations.
• GCC/C99-style compound literal values.
• Designated initializers.
•
__func__
predefined symbol.
• Implicit return 0;
in main()
.
• Non- const
static data initializations.
• Variable argument macros (
__VA_ARGS__
).
• bool
and
_Bool
support.
• long long
support (separate switch).
• restrict
support.
•
//
comments.
• inline
support.
• Digraphs.
•
_Complex
and
_Imaginary
(treated as keywords but not supported).
• Empty arrays as last struct members.
• Designated initializers
• Hexadecimal floating-point constants.
• Variable length arrays are supported within local or function prototype scope (as required by the C99 standard).
• Unsuffixed decimal constant rules.
•
++bool--
expressions.
•
(T) (int-list)
are handled/parsed as cast-expressions and as literals.
•
__STDC_HOSTED__
is
1
.
By default, this pragma is disabled.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
438 Freescale Semiconductor, Inc.
Chapter 28 Pragmas for Standard C Conformance
28.3 c9x
Equivalent to
#pragma c99
.
28.4 ignore_oldstyle
Controls the recognition of function declarations that follow the syntax conventions used before ISO/IEC standard C (in other words, "K&R" style).
Syntax
#pragma ignore_oldstyle on | off | reset
Remarks
If you enable this pragma, the compiler ignores old-style function declarations and lets you prototype a function any way you want. In old-style declarations, you specify the types of arguments on separate lines instead of the function's argument list. For example,
the code in Listing 28-1 on page 439 defines a prototype for a function with an old-style
definition.
Listing 28-1. Mixing Old-style and Prototype Function Declarations
int f(char x, short y, float z);
#pragma ignore_oldstyle on f(x, y, z) char x; short y; float z;
{
return (int)x+y+z;
}
#pragma ignore_oldstyle reset
This pragma does not correspond to any panel setting. By default, this setting is disabled.
28.5 only_std_keywords
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
439
require_prototypes
Controls the use of ISO/IEC keywords.
Syntax
#pragma only_std_keywords on | off | reset
Remarks
The compiler recognizes additional reserved keywords. If you are writing source code that must follow the ISO/IEC C standards strictly, enable the pragma only_std_keywords
.
By default, this pragma is disabled.
28.6 require_prototypes
Controls whether or not the compiler should expect function prototypes.
Syntax
#pragma require_prototypes on | off | reset
Remarks
This pragma only affects non-static functions.
If you enable this pragma, the compiler generates an error message if you use a function that does not have a preceding prototype. Use this pragma to prevent error messages caused by referring to a function before you define it. For example, without a function prototype, you might pass data of the wrong type. As a result, your code might not work as you expect even though it compiles without error.
In
Listing 28-2 on page 440 , function
main()
calls
PrintNum()
with an integer argument even though
PrintNum()
takes an argument of type float
.
Listing 28-2. Unnoticed Type-mismatch
#include <stdio.h> void main(void)
{
PrintNum(1); /* PrintNum() tries to interpret the
integer as a float. Prints 0.000000. */
} void PrintNum(float x)
{
printf("%f\n", x);
}
When you run this program, you could get this result:
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
440 Freescale Semiconductor, Inc.
Chapter 28 Pragmas for Standard C Conformance
0.000000
Although the compiler does not complain about the type mismatch, the function does not give the result you intended. Since
PrintNum()
does not have a prototype, the compiler does not know to generate instructions to convert the integer to a floating-point number before calling
PrintNum()
. Consequently, the function interprets the bits it received as a floating-point number and prints nonsense.
A prototype for
PrintNum()
, as in
, gives the compiler sufficient information about the function to generate instructions to properly convert its argument to a floating-point number. The function prints what you expected.
Listing 28-3. Using a Prototype to Avoid Type-mismatch
#include <stdio.h> void PrintNum(float x); /* Function prototype. */ void main(void)
{
PrintNum(1); /* Compiler converts int to float.
} Prints 1.000000. */ void PrintNum(float x)
{
printf("%f\n", x);
}
In other situations where automatic conversion is not possible, the compiler generates an error message if an argument does not match the data type required by a function prototype. Such a mismatched data type error is easier to locate at compile time than at runtime.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
441
require_prototypes
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
442 Freescale Semiconductor, Inc.
Chapter 29
Pragmas for C++
Pragmas for C++ should be placed outside the functions to which it is desired that they be placed.
This section explains the following pragmas:
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
443
access_errors
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
29.1 access_errors
Controls whether or not to change invalid access errors to warnings.
Syntax
#pragma access_errors on | off | reset
Remarks
If you enable this pragma, the compiler issues an error message instead of a warning when it detects invalid access to protected or private class members.
This pragma does not correspond to any IDE panel setting. By default, this pragma is on
.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
444 Freescale Semiconductor, Inc.
Chapter 29 Pragmas for C++
29.2 always_inline
Controls the use of inlined functions.
Syntax
#pragma always_inline on | off | reset
Remarks
This pragma is deprecated. We recommend that you use the inline_depth()
pragma instead.
29.3 arg_dep_lookup
Controls C++ argument-dependent name lookup.
Syntax
#pragma arg_dep_lookup on | off | reset
Remarks
If you enable this pragma, the C++ compiler uses argument-dependent name lookup.
This pragma does not correspond to any IDE panel setting. By default, this setting is on
.
29.4 ARM_conform
This pragma is no longer available. Use
ARM_scoping
instead.
29.5 ARM_scoping
Controls the scope of variables declared in the expression parts of if
, while
, do
, and for statements.
Syntax
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
445
array_new_delete
#pragma ARM_scoping on | off | reset
Remarks
If you enable this pragma, any variables you define in the conditional expression of an if
, while
, do
, or for
statement remain in scope until the end of the block that contains the statement. Otherwise, the variables only remain in scope until the end of that statement.
shows an example.
This pragma corresponds to the Legacy for-scoping setting in the CodeWarrior IDE's
Properties > C/C++ Build > Settings > Tool Settings > Compiler > C/C++ Language
settings panel. By default, this pragma is off
.
Listing 29-1. Example of Using Variables Declared in for
Statement
for(int i=1; i<1000; i++) { /* . . . */ } return i; // OK if ARM_scoping is on, error if ARM_scoping is off.
29.6 array_new_delete
Enables the operator new[]
and delete[]
in array allocation and deallocation operations, respectively.
Syntax
#pragma array_new_delete on | off | reset
Remarks
By default, this pragma is on
.
29.7 auto_inline
Controls which functions to inline.
Syntax
#pragma auto_inline on | off | reset
Remarks
If you enable this pragma, the compiler automatically chooses functions to inline for you, in addition to functions declared with the inline
keyword.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
446 Freescale Semiconductor, Inc.
Chapter 29 Pragmas for C++
Note that if you enable either the Do not Inline setting or the dont_inline
pragma, the compiler ignores the setting of the auto_inline
pragma and does not inline any functions.
This pragma corresponds to the Auto-Inline setting in the CodeWarrior IDE's Properties
> C/C++ Build > Settings > Tool Settings > Compiler > C/C++ Language settings panel. By default, this pragma is disabled.
29.8 bool
Determines whether or not bool
, true
, and false
are treated as keywords in C++ source code.
Syntax
#pragma bool on | off | reset
Remarks
If you enable this pragma, you can use the standard C++ bool
type to represent true
and false
. Disable this pragma if bool
, true
, or false
are defined in your source code.
Enabling the bool
data type and its true
and false
values is not equivalent to defining them in source code with typedef
, enum
, or
#define
. The C++ bool
type is a distinct type defined by the ISO/IEC 14882:2003 C++ Standard. Source code that does not treat bool
as a distinct type might not compile properly.
This pragma corresponds to the Enable bool Support setting in the CodeWarrior IDE's
Properties > C/C++ Build > Settings > Tool Settings > Compiler> C/C++ Language
settings panel. By default, this setting is on
.
29.9 cplusplus
Controls whether or not to translate subsequent source code as C or C++ source code.
Syntax
#pragma cplusplus on | off | reset
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
447
cpp1x
If you enable this pragma, the compiler translates the source code that follows as C++ code. Otherwise, the compiler uses the suffix of the filename to determine how to compile it. If a file name ends in
.c
,
.h
, or
.pch
, the compiler automatically compiles it as C code, otherwise as C++. Use this pragma only if a file contains both C and C++ code.
NOTE
The CodeWarrior C/C++ compilers do not distinguish between uppercase and lowercase letters in file names and file name extensions except on UNIX-based systems.
This pragma corresponds to the Force C++ Compilation setting in the CodeWarrior
IDE's Properties > C/C++ Build > Settings > Tool Settings > Compiler > C/C++
Language settings panel. By default, this pragma is disabled.
29.10 cpp1x
Controls whether or not to enable support to experimental features made available in the
1x version of C++ standard.
Syntax
#pragma cpp1x on | off | reset
Remarks
If you enable this pragma, you can use the following extensions to the 1x or 05 version of the C++ standard that would otherwise be invalid:
• Enables support for
__alignof__
.
• Enables support for
__decltype__
, which is a reference type preserving typeof.
• Enables support for nullptr
.
• Enables support to allow
>>
to terminate nested template argument lists.
• Enables support for static_assert
.
NOTE
This pragma enables support to experimental and unvalidated implementations of features that may or may not be available in the final version of the C++ standard.
The features should not be used for critical or production code.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
448 Freescale Semiconductor, Inc.
Chapter 29 Pragmas for C++
29.11 cpp_extensions
Controls language extensions to ISO/IEC 14882:2003 C++.
Syntax
#pragma cpp_extensions on | off | reset
Remarks
If you enable this pragma, you can use the following extensions to the ISO/IEC
14882:2003 C++ standard that would otherwise be invalid:
• Anonymous struct
& union
objects. Listing 29-2 on page 449
shows an example.
Listing 29-2. Example of Anonymous struct
& union
Objects
#pragma cpp_extensions on void func()
{
union {
long hilo;
struct { short hi, lo; }; // anonymous struct
};
hi=0x1234;
lo=0x5678; // hilo==0x12345678
}
• Unqualified pointer to a member function. Listing 29-3 on page 449
shows an example.
Listing 29-3. Example of an Unqualified Pointer to a Member Function
#pragma cpp_extensions on struct RecA { void f(); } void RecA::f()
{
void (RecA::*ptmf1)() = &RecA::f; // ALWAYS OK
void (RecA::*ptmf2)() = f; // OK if you enable cpp_extensions.
}
• Inclusion of const
data in precompiled headers.
By default, this pragma is disabled.
29.12 debuginline
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
449
def_inherited
Controls whether the compiler emits debugging information for expanded inline function calls.
Syntax
#pragma debuginline on | off | reset
Remarks
If the compiler emits debugging information for inline function calls, then the debugger can step to the body of the inlined function. This behavior more closely resembles the debugging experience for un-inlined code.
NOTE
Since the actual "call" and "return" instructions are no longer present when stepping through inline code, the debugger will immediately jump to the body of an inlined function and
"return" before reaching the return statement for the function.
Thus, the debugging experience of inlined functions may not be as smooth as debugging un-inlined code.
This pragma does not correspond to any panel setting. By default, this pragma is on
.
29.13 def_inherited
Controls the use of inherited
.
Syntax
#pragma def_inherited on | off | reset
Remarks
The use of this pragma is deprecated. It lets you use the non-standard inherited
symbol in
C++ programming by implicitly adding typedef base inherited; as the first member in classes with a single base class.
NOTE
The ISO/IEC 14882:2003 C++ standard does not support the inherited
symbol. Only the CodeWarrior C++ language implements the inherited
symbol for single inheritance.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
450 Freescale Semiconductor, Inc.
By default, this pragma is off
.
29.14 defer_codegen
Obsolete pragma. Replaced by interprocedural analysis options.
Chapter 29 Pragmas for C++
29.15 defer_defarg_parsing
Defers the parsing of default arguments in member functions.
Syntax
#pragma defer_defarg_parsing on | off
Remarks
To be accepted as valid, some default expressions with template arguments will require
additional parenthesis. For example, Listing 29-4 on page 451
results in an error message.
Listing 29-4. Deferring parsing of default arguments
template<typename T,typename U> struct X { T t; U u; }; struct Y {
// The following line is not accepted, and generates
// an error message with defer_defarg_parsing on.
void f(X<int,int> = X<int,int>());
};
does not generate an error message.
Listing 29-5. Correct default argument deferral
template<typename T,typename U> struct X { T t; U u; }; struct Y {
// The following line is OK if the default
// argument is parenthesized.
void f(X<int,int> = (X<int,int>()) );
};
This pragma does not correspond to any panel setting. By default, this pragma is on
.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
451
direct_to_som
29.16 direct_destruction
This pragma is obsolete. It is no longer available.
29.17 direct_to_som
This pragma is obsolete. It is no longer available.
29.18 dont_inline
Controls the generation of inline functions.
Syntax
#pragma dont_inline on | off | reset
Remarks
If you enable this pragma, the compiler does not inline any function calls, even those declared with the inline
keyword or within a class declaration. Also, it does not automatically inline functions, regardless of the setting of the auto_inline
pragma,
described in "auto_inline" on page 195.
If you disable this pragma, the compiler expands all inline function calls, within the limits you set through other inlining-related pragmas.
This pragma corresponds to the Do not Inline setting in the CodeWarrior IDE's C/C++
Language settings panel. By default, this pragma is off
.
29.19 ecplusplus
Controls the use of embedded C++ features.
Syntax
#pragma ecplusplus on | off | reset
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
452 Freescale Semiconductor, Inc.
Chapter 29 Pragmas for C++
If you enable this pragma, the C++ compiler disables the non-EC++ features of ISO/IEC
14882:2003 C++ such as templates, multiple inheritance, and so on.
This pragma corresponds to the EC++ Compatibility Mode setting in the CodeWarrior
IDE's Properties > C/C++ Build > Settings > Tool Settings > Compiler> C/C++
Language settings panel. By default, this pragma is off
.
29.20 exceptions
Controls the availability of C++ exception handling.
Syntax
#pragma exceptions on | off | reset
Remarks
If you enable this pragma, you can use the try
and catch
statements in C++ to perform exception handling. If your program does not use exception handling, disable this setting to make your program smaller.
You can throw exceptions across any code compiled by the CodeWarrior C/C++ compiler with
#pragma exceptions o
n.
You cannot throw exceptions across libraries compiled with
#pragma exceptions off
. If you throw an exception across such a library, the code calls terminate()
and exits.
This pragma corresponds to the Enable C++ Exceptions setting in the CodeWarrior IDE's
Properties > C/C++ Build > Settings > Tool Settings > Compiler > C/C++ Language
settings panel. By default, this pragma is on
.
29.21 inline_bottom_up
Controls the bottom-up function inlining method.
Syntax
#pragma inline_bottom_up on | off | reset
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
453
inline_bottom_up_once
Bottom-up function inlining tries to expand up to eight levels of inline leaf functions. The maximum size of an expanded inline function and the caller of an inline function can be controlled by the pragmas shown in
Listing 29-6 on page 454 and Listing 29-7 on page
Listing 29-6. Maximum Complexity of an Inlined Function
// Maximum complexity of an inlined function
#pragma inline_max_size( max ) // default max == 256
Listing 29-7. Maximum Complexity of a Function that Calls Inlined Functions
// Maximum complexity of a function that calls inlined functions
#pragma inline_max_total_size( max ) // default max == 10000 where max loosely corresponds to the number of instructions in a function.
If you enable this pragma, the compiler calculates inline depth from the last function in the call chain up to the first function that starts the call chain. The number of functions the compiler inlines from the bottom depends on the values of inline_depth
, inline_max_size
, and inline_max_total_size
. This method generates faster and smaller source code for some (but not all) programs with many nested inline function calls.
If you disable this pragma, top-down inlining is selected, and the inline_depth
setting determines the limits for top-down inlining. The inline_max_size
and inline_max_total_size pragmas do not affect the compiler in top-down mode.
29.22 inline_bottom_up_once
Performs a single bottom-up function inlining operation.
Syntax
#pragma inline_bottom_up_once on | off | reset
Remarks
By default, this pragma is off
.
29.23 inline_depth
Controls how many passes are used to expand inline function calls.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
454 Freescale Semiconductor, Inc.
Chapter 29 Pragmas for C++
Syntax
#pragma inline_depth(
n )
#pragma inline_depth(smart)
Parameters
n
Sets the number of passes used to expand inline function calls. The number n is an integer from
0
to
1024
or the smart
specifier. It also represents the distance allowed in the call chain from the last function up. For example, if d
is the total depth of a call chain, then functions below a depth of d-n
are inlined if they do not exceed the following size settings:
#pragma inline_max_size(n);
#pragma inline_max_total_size(n);
The first pragma sets the maximum function size to be considered for inlining; the second sets the maximum size to which a function is allowed to grow after the functions it calls are inlined. Here, n is the number of statements, operands, and operators in the function, which turns out to be roughly twice the number of instructions generated by the function.
However, this number can vary from function to function. For the inline_max_size
pragma, the default value of n is 256; for the inline_max_total_size
pragma, the default value of n is
10000.
smart
The smart
specifier is the default mode, with four passes where the passes 2-4 are limited to small inline functions. All inlineable functions are expanded if inline_depth
is set to
1-1024.
Remarks
The pragmas dont_inline
and always_inline
override this pragma. This pragma corresponds to the Inline Depth setting in the CodeWarrior IDE's C/C++ Language settings panel.
Setting the Inline Depth option to "Do not Inline" in the settings panel will also override this pragma. By default, this pragma is disabled.
29.24 inline_max_auto_size
Determines the maximum complexity for an auto-inlined function.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
455
inline_max_size
Syntax
#pragma inline_max_auto_size ( complex )
Parameters
complex
The complex
value is an approximation of the number of statements in a function, the current default value is 15. Selecting a higher value will inline more functions, but can lead to excessive code bloat.
Remarks
This pragma does not correspond to any panel setting.
29.25 inline_max_size
Sets the maximum number of statements, operands, and operators used to consider the function for inlining.
Syntax
#pragma inline_max_size ( size )
Parameters
size
The maximum number of statements, operands, and operators in the function to consider it for inlining, up to a maximum of 256.
Remarks
This pragma does not correspond to any panel setting.
29.26 inline_max_total_size
Sets the maximum total size a function can grow to when the function it calls is inlined.
Syntax
#pragma inline_max_total_size ( max_size )
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
456 Freescale Semiconductor, Inc.
Chapter 29 Pragmas for C++
Parameters
max_size
The maximum number of statements, operands, and operators the inlined function calls that are also inlined, up to a maximum of 7000.
Remarks
This pragma does not correspond to any panel setting.
29.27 internal
Controls the internalization of data or functions.
Syntax
#pragma internal on | off | reset
#pragma internal list name1 [, name2 ]*
Remarks
When using the
#pragma internal on
format, all data and functions are automatically internalized.
Use the
#pragma internal list
format to tag specific data or functions for internalization. It applies to all names if it is used on an overloaded function. You cannot use this pragma for C++ member functions or static class members.
shows an example:
Listing 29-8. Example of an Internalized List
extern int f(), g;
#pragma internal list f,g
This pragma does not correspond to any panel setting. By default, this pragma is disabled.
29.28 iso_templates
Controls whether or not to use the new parser supported by the CodeWarrior 2.5 C++ compiler and issue warning messages for missing typenames.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
457
new_mangler
Syntax
#pragma iso_templates on | off | reset
Remarks
This pragma combines the functionality of pragmas parse_func_templ ,
and
This pragma ensures that your C++ source code is compiled using the newest version of the parser, which is stricter than earlier versions. The compiler issues a warning message if a typenames required by the C++ standard is missing but can still be determined by the compiler based on the context of the surrounding C++ syntax.
By default, this pragma is on.
29.29 new_mangler
Controls the inclusion or exclusion of a template instance's function return type to the mangled name of the instance.
Syntax
#pragma new_mangler on | off | reset
Remarks
The C++ standard requires that the function return type of a template instance to be included in the mangled name, which can cause incompatibilities. Enabling this pragma within a prefix file resolves those incompatibilities.
This pragma does not correspond to any panel setting. By default, this pragma is on
.
29.30 no_conststringconv
Disables the deprecated implicit const string literal conversion (ISO/IEC 14882:2003 C+
+, §4.2).
Syntax
#pragma no_conststringconv on | off | reset
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
458 Freescale Semiconductor, Inc.
Chapter 29 Pragmas for C++
Remarks
When enabled, the compiler generates an error message when it encounters an implicit const string conversion.
Listing 29-9. Example of const string conversion
#pragma no_conststringconv on char *cp = "Hello World"; /* Generates an error message. */
This pragma does not correspond to any panel setting. By default, this pragma is off
.
29.31 no_static_dtors
Controls the generation of static destructors in C++.
Syntax
#pragma no_static_dtors on | off | reset
Remarks
If you enable this pragma, the compiler does not generate destructor calls for static data objects. Use this pragma to generate smaller object code for C++ programs that never exit
(and consequently never need to call destructors for static objects).
This pragma does not correspond to any panel setting. By default, this setting is disabled.
29.32 nosyminline
Controls whether debug information is gathered for inline/template functions.
Syntax
#pragma nosyminline on | off | reset
Remarks
When on, debug information is not gathered for inline/template functions.
This pragma does not correspond to any panel setting. By default, this pragma is disabled.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
459
old_friend_lookup
29.33 old_friend_lookup
Implements non-standard C++ friend declaration behavior that allows friend declarations to be visible in the enclosing scope.
#pragma old_friend_lookup on | off | reset
Example
This example shows friend declarations that are invalid without #pragma old_friend_lookup.
Listing 29-10. Valid and invalid declarations without #pragma old_friend_lookup
class C2; void f2(); struct S {
friend class C1;
friend class C2;
friend void f1();
friend void f2();
};
C1 *cp1; // error, C1 is not visible without namespace declaration
C2 *cp2; // OK int main()
{
f1(); // error, f1() is not visible without namespace declaration
f2(); // OK
}
29.34 old_pods
Permits non-standard handling of classes, structs, and unions containing pointer-topointer members
Syntax
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
460 Freescale Semiconductor, Inc.
Chapter 29 Pragmas for C++
#pragma old_pods on | off | reset
Remarks
According to the ISO/IEC 14882:2003 C++ Standard, classes/structs/unions that contain pointer-to-pointer members are now considered to be plain old data (POD) types.
This pragma can be used to get the old behavior.
29.35 old_vtable
This pragma is no longer available.
29.36 opt_classresults
Controls the omission of the copy constructor call for class return types if all return statements in a function return the same local class object.
Syntax
#pragma opt_classresults on | off | reset
Remarks
Listing 29-11 on page 461 shows an example.
Listing 29-11. Example #pragma opt_classresults
#pragma opt_classresults on struct X {
X();
X(const X&);
// ...
};
X f() {
X x; // Object x will be constructed in function result buffer.
// ...
return x; // Copy constructor is not called.
}
This pragma does not correspond to any panel setting. By default, this pragma is on
.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
461
parse_mfunc_templ
29.37 parse_func_templ
Controls whether or not to use the new parser supported by the CodeWarrior 2.5 C++ compiler.
Syntax
#pragma parse_func_templ on | off | reset
Remarks
If you enable this pragma, your C++ source code is compiled using the newest version of the parser, which is stricter than earlier versions.
This option actually corresponds to the ISO C++ Template Parser option (together with
and
warn_no_typename ). By default, this pragma is disabled.
29.38 parse_mfunc_templ
Controls whether or not to use the new parser supported by the CodeWarrior 2.5 C++ compiler for member function bodies.
Syntax
#pragma parse_mfunc_templ on | off | reset
Remarks
If you enable this pragma, member function bodies within your C++ source code is compiled using the newest version of the parser, which is stricter than earlier versions.
This pragma does not correspond to any panel setting. By default, this pragma is disabled.
29.39 RTTI
Controls the availability of runtime type information.
Syntax
#pragma RTTI on | off | reset
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
462 Freescale Semiconductor, Inc.
Chapter 29 Pragmas for C++
Remarks
If you enable this pragma, you can use runtime type information (or RTTI) features such as dynamic_cast
and typeid
. The other RTTI expressions are available even if you disable the Enable RTTI setting. Note that
*type_info::before(const type_info&) is not implemented.
This pragma corresponds to the Enable RTTI setting in the CodeWarrior IDE's
Properties > C/C++ Build > Settings > Tool Settings > Compiler> C/C++ Language
settings panel.
29.40 suppress_init_code
Controls the suppression of static initialization object code.
Syntax
#pragma suppress_init_code on | off | reset
Remarks
If you enable this pragma, the compiler does not generate any code for static data initialization such as C++ constructors.
NOTE
Using this pragma because it can produce erratic or unpredictable behavior in your program.
This pragma does not correspond to any panel setting. By default, this pragma is disabled.
29.41 template_depth
Controls how many nested or recursive class templates you can instantiate.
#pragma template_depth(n)
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
463
thread_safe_init
This pragma lets you increase the number of nested or recursive class template instantiations allowed. By default, n equals 64; it can be set from 1 to 30000. You should always use the default value unless you receive the error message template too complex or recursive
This pragma does not correspond to any panel setting.
29.42 thread_safe_init
Controls the addition of extra code in the binary to ensure that multiple threads cannot enter a static local initialization at the same time.
Syntax
#pragma thread_safe_init on | off | reset
Remarks
A C++ program that uses multiple threads and static local initializations introduces the possibility of contention over which thread initializes static local variable first. When the pragma is on
, the compiler inserts calls to mutex functions around each static local initialization to avoid this problem. The C++ runtime library provides these mutex functions.
Listing 29-12. Static local initialization example
int func(void) {
// There may be synchronization problems if this function is
// called by multiple threads.
static int countdown = 20;
return countdown--;
}
NOTE
This pragma requires runtime library functions which may not be implemented on all platforms, due to the possible need for operating system support.
Listing 29-13 on page 464 shows another example.
Listing 29-13. Example thread_safe_init
#pragma thread_safe_init on void thread_heavy_func()
{
// Multiple threads can now safely call this function:
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
464 Freescale Semiconductor, Inc.
Chapter 29 Pragmas for C++
// the static local variable will be constructed only once.
static std::string localstring = thread_unsafe_func();
}
NOTE
When an exception is thrown from a static local initializer, the initializer is retried by the next client that enters the scope of the local.
This pragma does not correspond to any panel setting. By default, this pragma is off
.
29.43 warn_hidevirtual
Controls the recognition of a non-virtual member function that hides a virtual function in a superclass.
Syntax
#pragma warn_hidevirtual on | off | reset
Remarks
If you enable this pragma, the compiler issues a warning message if you declare a nonvirtual member function that hides a virtual function in a superclass. One function hides another if it has the same name but a different argument type.
shows an example.
Listing 29-14. Hidden Virtual Functions
class A {
public:
virtual void f(int);
virtual void g(int);
}; class B: public A {
public:
void f(char); // WARNING: Hides A::f(int)
virtual void g(int); // OK: Overrides A::g(int)
};
The ISO/IEC 14882:2003 C++ Standard does not require this pragma.
NOTE
A warning message normally indicates that the pragma name is not recognized, but an error indicates either a syntax problem or that the pragma is not valid in the given context.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
465
warn_no_explicit_virtual
This pragma corresponds to the Hidden Virtual Functions setting in the CodeWarrior
IDE's Properties > C/C++ Build > Settings > Tool Settings > Compiler > Warniings
C/C++ Language settings panel.
29.44 warn_no_explicit_virtual
Controls the issuing of warning messages if an overriding function is not declared with a virtual keyword.
Syntax
#pragma warn_no_explicit_virtual on | off | reset
Remarks
Listing 29-15 on page 466 shows an example.
Listing 29-15. Example of warn_no_explicit_virtual pragma
#pragma warn_no_explicit_virtual on struct A {
virtual void f();
}; struct B {
void f();
// WARNING: override B::f() is declared without virtual keyword
}
Tip
This warning message is not required by the ISO/IEC
14882:2003 C++ standard, but can help you track down unwanted overrides.
This pragma does not correspond to any panel setting. By default, this pragma is off
.
29.45 warn_no_typename
Controls the issuing of warning messages for missing typenames
.
Syntax
#pragma warn_no_typename on | off | reset
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
466 Freescale Semiconductor, Inc.
Chapter 29 Pragmas for C++
Remarks
The compiler issues a warning message if a typenames
required by the C++ standard is missing but can still be determined by the compiler based on the context of the surrounding C++ syntax.
This pragma does not correspond to any panel setting. This pragma is enabled by the
ISO/IEC 14882:2003 C++ template parser.
29.46 warn_notinlined
Controls the issuing of warning messages for functions the compiler cannot inline.
Syntax
#pragma warn_notinlined on | off | reset
Remarks
The compiler issues a warning message for non-inlined inline (i.e., on those indicated by the inline
keyword or in line in a class declaration) function calls.
This pragma corresponds to the Non-Inlined Functions setting in the CodeWarrior IDE's
Properties > C/C++ Build > Settings > Tool Settings > Compiler > Warniings C/C++
Warnings settings panel. By default, this pragma is disabled.
29.47 warn_structclass
Controls the issuing of warning messages for the inconsistent use of the class
and struct keywords.
Syntax
#pragma warn_structclass on | off | reset
Remarks
If you enable this pragma, the compiler issues a warning message if you use the class
and struct
keywords in the definition and declaration of the same identifier.
Listing 29-16. Inconsistent use of class
and struct
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
467
wchar_type
class X; struct X { int a; }; // WARNING
Use this warning when using static or dynamic libraries to link with object code produced by another C++ compiler that distinguishes between class and structure variables in its name " mangling."
This pragma corresponds to the Inconsistent `class' / `struct' Usage setting in the
CodeWarrior IDE's Properties > C/C++ Build > Settings > Tool Settings > Compiler
> Warniings C/C++ Warnings settings panel. By default, this pragma is disabled.
29.48 wchar_type
Controls the availability of the wchar_t data type in C++ source code.
Syntax
#pragma wchar_type on | off | reset
Remarks
If you enable this pragma, wchar_t
is treated as a built-in type. Otherwise, the compiler does not recognize this type.
This pragma corresponds to the Enable wchar_t Support setting in the CodeWarrior
IDE's Properties > C/C++ Build > Settings > Tool Settings > Compiler >C/C++
Language settings panel. By default, this pragma is enabled.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
468 Freescale Semiconductor, Inc.
Chapter 30
Pragmas for Language Translation
This chapter explains the following pragmas for Language Translation:
•
•
•
•
•
•
•
•
•
•
multibyteaware_preserve_literals
•
•
•
30.1 asmpoundcomment
Controls whether the "#" symbol is treated as a comment character in inline assembly.
Syntax
#pragma asmpoundcomment on | off | reset
Remarks
Some targets may have additional comment characters, and may treat these characters as comments even when
#pragma asmpoundcomment off
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
469
asmsemicolcomment
is used.
Using this pragma may interfere with the function-level inline assembly language.
This pragma does not correspond to any panel setting. By default, this pragma is on
.
30.2 asmsemicolcomment
Controls whether the "
;
" symbol is treated as a comment character in inline assembly.
Syntax
#pragma asmsemicolcomment on | off | reset
Remarks
Some targets may have additional comment characters, and may treat these characters as comments even when
#pragma asmsemicolcomment off is used.
Using this pragma may interfere with the assembly language of a specific target.
This pragma does not correspond to any panel setting. By default, this pragma is on
.
30.3 const_strings
Controls the const
-ness of character string literals.
Syntax
#pragma const_strings [ on | off | reset ]
Remarks
If you enable this pragma, the type of string literals is an array const char[
n
]
, or const wchar_t[
n
]
for wide strings, where n is the length of the string literal plus 1 for a terminating
NUL
character. Otherwise, the type char[
n
]
or wchar_t[
n
]
is used.
By default, this pragma is on
when compiling C++ source code and off
when compiling C source code.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
470 Freescale Semiconductor, Inc.
Chapter 30 Pragmas for Language Translation
30.4 dollar_identifiers
Controls use of dollar signs ($) in identifiers.
Syntax
#pragma dollar_identifiers on | off | reset
Remarks
If you enable this pragma, the compiler accepts dollar signs (
$
) in identifiers. Otherwise, the compiler issues an error if it encounters anything but underscores, alphabetic, numeric character, and universal characters (
\uxxxx
,
\Uxxxxxxxx
) in an identifier.
This pragma does not correspond to any panel setting. By default, this pragma is off
.
30.5 gcc_extensions
Controls the acceptance of GNU C language extensions.
Syntax
#pragma gcc_extensions on | off | reset
Remarks
If you enable this pragma, the compiler accepts GNU C extensions in C source code. This includes the following non-ANSI C extensions:
• Initialization of automatic struct
or array
variables with non- const
values.
• Illegal pointer conversions
• sizeof( void ) == 1
• sizeof( function-type ) == 1
• Limited support for GCC statements and declarations within expressions.
• Macro redefinitions without a previous
#undef
.
• The GCC keyword typeof
• Function pointer arithmetic supported
• void*
arithmetic supported
• Void expressions in return statements of void
•
__builtin_constant_p (
expr
)
supported
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
471
mark
• Forward declarations of arrays of incomplete type
• Forward declarations of empty static arrays
• Pre-C99 designated initializer syntax (deprecated)
• shortened conditional expression ( c
?:
y
)
• long __builtin_expect (long exp, long c)
now accepted
By default, this pragma is disabled.
30.6 mark
Adds an item to the Function pop-up menu in the IDE editor.
Syntax
#pragma mark
itemName
Remarks
This pragma adds itemName to the source file's Function pop-up menu. If you open the file in the CodeWarrior Editor and select the item from the Function pop-up menu, the editor brings you to the pragma. Note that if the pragma is inside a function definition, the item does not appear in the Function pop-up menu.
If itemName begins with "
--
", a menu separator appears in the IDE's Function pop-up menu:
#pragma mark --
This pragma does not correspond to any panel setting.
30.7 mpwc_newline
Controls the use of newline character convention.
Syntax
#pragma mpwc_newline on | off | reset
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
472 Freescale Semiconductor, Inc.
Chapter 30 Pragmas for Language Translation
If you enable this pragma, the compiler translates
'\n'
as a Carriage Return (0x0D) and
'\r'
as a Line Feed (0x0A). Otherwise, the compiler uses the ISO standard conventions for these characters.
If you enable this pragma, use ISO standard libraries that were compiled when this pragma was enabled.
If you enable this pragma and use the standard ISO standard libraries, your program will not read and write
'\n'
and
'\r'
properly. For example, printing
'\n'
brings your program's output to the beginning of the current line instead of inserting a newline.
This pragma does not correspond to any IDE panel setting. By default, this pragma is disabled.
30.8 mpwc_relax
Controls the compatibility of the char*
and unsigned char*
types.
Syntax
#pragma mpwc_relax on | off | reset
Remarks
If you enable this pragma, the compiler treats char*
and unsigned char*
as the same type.
Use this setting to compile source code written before the ISO C standards. Old source code frequently uses these types interchangeably.
This setting has no effect on C++ source code.
NOTE
Turning this option on may prevent the compiler from detecting some programming errors. We recommend not turning on this option.
shows how to use this pragma to relax function pointer checking.
Listing 30-1. Relaxing function pointer checking
#pragma mpwc_relax on extern void f(char *);
/* Normally an error, but allowed. */ extern void(*fp1)(void *) = &f;
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
473
multibyteaware
/* Normally an error, but allowed. */ extern void(*fp2)(unsigned char *) = &f;
This pragma does not correspond to any panel setting. By default, this pragma is disabled.
30.9 multibyteaware
Controls how the Source encoding option in the IDE is treated
Syntax
#pragma multibyteaware on | off | reset
Remarks
This pragma is deprecated. See
#pragma text_encoding
for more details.
By default, this pragma is off
.
30.10 multibyteaware_preserve_literals
Controls the treatment of multibyte character sequences in narrow character string literals.
Syntax
#pragma multibyteaware_preserve_literals on | off | reset
Remarks
This pragma does not correspond to any panel setting. By default, this pragma is on
.
30.11 text_encoding
Identifies the character encoding of source files.
Syntax
#pragma text_encoding ( "name" | unknown | reset [, global] )
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
474 Freescale Semiconductor, Inc.
Chapter 30 Pragmas for Language Translation
Parameters
name
The IANA or MIME encoding name or an OS-specific string that identifies the text encoding. The compiler recognizes these names and maps them to its internal decoders: system US-ASCII ASCII ANSI_X3.4-1968
ANSI_X3.4-1968 ANSI_X3.4 UTF-8 UTF8 ISO-2022-JP
CSISO2022JP ISO2022JP CSSHIFTJIS SHIFT-JIS
SHIFT_JIS SJIS EUC-JP EUCJP UCS-2 UCS-2BE
UCS-2LE UCS2 UCS2BE UCS2LE UTF-16 UTF-16BE
UTF-16LE UTF16 UTF16BE UTF16LE UCS-4 UCS-4BE
UCS-4LE UCS4 UCS4BE UCS4LE 10646-1:1993
ISO-10646-1 ISO-10646 unicode global
Tells the compiler that the current and all subsequent files use the same text encoding. By default, text encoding is effective only to the end of the file.
Remarks
By default,
#pragmatext_encoding
is only effective through the end of file. To affect the default text encoding assumed for the current and all subsequent files, supply the "global" modifier.
By default, this setting is
ASCII
.
30.12 trigraphs
Controls the use trigraph sequences specified in the ISO standards.
Syntax
#pragma trigraphs on | off | reset
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
475
??=
??/
?? '
??(
??)
??!
??<
??>
??-
unsigned_char
If you are writing code that must strictly adhere to the ANSI standard, enable this pragma.
Table 30-1. Trigraph table
Trigraph Character
|
{
}
~
[
]
#
\
^
NOTE
Use of this pragma may cause a portability problem for some targets.
Be careful when initializing strings or multi-character constants that contain question marks.
Listing 30-2. Example of Pragma trigraphs
char c = '????'; /* ERROR: Trigraph sequence expands to '??^ */ char d = '\?\?\?\?'; /* OK */
By default, this pragma is disabled.
30.13 unsigned_char
Controls whether or not declarations of type char
are treated as unsigned char
.
Syntax
#pragma unsigned_char on | off | reset
Remarks
If you enable this pragma, the compiler treats a char
declaration as if it were an unsigned char
declaration.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
476 Freescale Semiconductor, Inc.
Chapter 30 Pragmas for Language Translation
NOTE
If you enable this pragma, your code might not be compatible with libraries that were compiled when the pragma was disabled. In particular, your code might not work with the ISO standard libraries included with CodeWarrior.
By default, this setting is disabled.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
477
unsigned_char
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
478 Freescale Semiconductor, Inc.
Chapter 31
Pragmas for Diagnostic Messages
This section explains the following pragmas for diagnostic messages:
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
479
extended_errorcheck
•
•
•
•
•
31.1 extended_errorcheck
Controls the issuing of warning messages for possible unintended logical errors.
Syntax
#pragma extended_errorcheck on | off | reset
Remarks
If you enable this pragma, the compiler generates a warning message (not an error) if it encounters some common programming errors:
It also issues a warning message when it encounters a delete operator for a class or structure that has not been defined yet. The listing below shows an example.
Listing 31-1. Attempting to delete an undefined structure
#pragma extended_errorcheck on struct X; int func(X *xp)
{
delete xp; // Warning: deleting incomplete type X
}
• An integer or floating-point value assigned to an enum
type. The listing below shows an example.
Listing 31-2. Assigning to an Enumerated Type
enum Day { Sunday, Monday, Tuesday, Wednesday,
Thursday, Friday, Saturday } d; d = 5; /* WARNING */ d = Monday; /* OK */ d = (Day)3; /* OK */
• An empty return
statement in a function that is not declared void
. For example, a non-void function with an empty return statement, listed below, results in a warning message.
Listing 31-3. A non-void function with an empty return statement
int MyInit(void)
{
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
480 Freescale Semiconductor, Inc.
Chapter 31 Pragmas for Diagnostic Messages
int err = GetMyResources();
if (err != -1)
{
err = GetMoreResources();
}
return; /* WARNING: empty return statement */
}
A non-void function with a proper return statement, listed below, shows how to prevent this warning message.
Listing 31-4. A non-void function with a proper return statement
int MyInit(void)
{
int err = GetMyResources();
if (err != -1)
{
err = GetMoreResources();
}
return err; /* OK */
}
By default, this setting is off
.
31.2 maxerrorcount
Limits the number of error messages emitted while compiling a single file.
Syntax
#pragma maxerrorcount(
num | off )
Parameters
num
Specifies the maximum number of error messages issued per source file.
off
Does not limit the number of error messages issued per source file.
Remarks
The total number of error messages emitted may include one final message:
Too many errors emitted
This pragma does not correspond to any panel setting. By default, this pragma is off
.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
481
message
31.3 message
Tells the compiler to issue a text message to the user.
Syntax
#pragma message(
msg )
Parameter
msg
Actual message to issue. Does not have to be a string literal.
Remarks
In the CodeWarrior IDE, the message appears in the Errors & Warnings window Console view. On the command line, the message is sent to the standard error stream.
This pragma does not correspond to any panel setting.
31.4 showmessagenumber
Controls the appearance of warning or error numbers in displayed messages.
Syntax
#pragma showmessagenumber on | off | reset
Remarks
When enabled, this pragma causes messages to appear with their numbers visible. You can then use the
warning pragma with a warning number to suppress the appearance of
specific warning messages.
This pragma does not correspond to any panel setting. By default, this pragma is off
.
31.5 show_error_filestack
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
482 Freescale Semiconductor, Inc.
Chapter 31 Pragmas for Diagnostic Messages
Controls the appearance of the current
# include
file stack within error messages occurring inside deeply-included files.
Syntax
#pragma show_error_filestack on | off | reset
Remarks
This pragma does not correspond to any panel setting. By default, this pragma is on
.
31.6 suppress_warnings
Controls the issuing of warning messages.
Syntax
#pragma suppress_warnings on | off | reset
Remarks
If you enable this pragma, the compiler does not generate warning messages, including those that are enabled.
This pragma does not correspond to any panel setting. By default, this pragma is off
.
31.7 sym
Controls the generation of debugger symbol information for subsequent functions.
Syntax
#pragma sym on | off | reset
Remarks
The compiler pays attention to this pragma only if you enable the debug marker for a file in the IDE project window. If you disable this pragma, the compiler does not put debugging information into the source file debugger symbol file (SYM or DWARF) for the functions that follow.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
483
unused
The compiler always generates a debugger symbol file for a source file that has a debug diamond next to it in the IDE project window. This pragma changes only which functions have information in that symbol file.
This pragma does not correspond to any panel setting. By default, this pragma is enabled.
31.8 unused
Controls the suppression of warning messages for variables and parameters that are not referenced in a function.
Syntax
#pragma unused (
var_name [,
var_name ]... )
var_name
The name of a variable.
Remarks
This pragma suppresses the compile time warning messages for the unused variables and parameters specified in its argument list. You can use this pragma only within a function body. The listed variables must be within the scope of the function.
In C++, you cannot use this pragma with functions defined within a class definition or with template functions.
Listing 31-5. Example of Pragma unused() in C
#pragma warn_unusedvar on
#pragma warn_unusedarg on static void ff(int a)
{
int b;
#pragma unused(a,b)
/* Compiler does not warn that a and b are unused. */
}
Listing 31-6. Example of Pragma unused() in C++
#pragma warn_unusedvar on
#pragma warn_unusedarg on static void ff(int /* No warning */)
{
int b;
#pragma unused(b)
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
484 Freescale Semiconductor, Inc.
Chapter 31 Pragmas for Diagnostic Messages
/* Compiler does not warn that b is unused. */
}
This pragma does not correspond to any CodeWarrior IDE panel setting.
31.9 warning
Controls which warning numbers are displayed during compiling.
Syntax
#pragma warning on | off | reset (
num [, ...])
This alternate syntax is allowed but ignored (message numbers do not match):
#pragma warning
(
warning_type :
warning_num_list [,
warning_type :
warning_num_list , ...])
Parameters
num
The number of the warning message to show or suppress.
warning_type
Specifies one of the following settings:
• default
• disable
• enable
warning_num_list
The warning_num_list is a list of warning numbers separated by spaces.
Remarks
Use the pragma showmessagenumber
to display warning messages with their warning numbers.
This pragma only applies to CodeWarrior front-end warnings. Using the pragma for the back-end warnings returns invalid message number warning.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
485
warning_errors
The CodeWarrior compiler allows, but ignores, the alternative syntax for compatibility with Microsoft® compilers.
This pragma does not correspond to any panel setting. By default, this pragma is on
.
31.10 warning_errors
Controls whether or not warnings are treated as errors.
Syntax
#pragma warning_errors on | off | reset
Remarks
If you enable this pragma, the compiler treats all warning messages as though they were errors and does not translate your file until you resolve them.
31.11 warn_any_ptr_int_conv
Controls if the compiler generates a warning message when an integral type is explicitly converted to a pointer type or vice versa.
Syntax
#pragma warn_any_ptr_int_conv on | off | reset
Remarks
This pragma is useful to identify potential 64-bit pointer portability issues. An example is shown in.
Listing 31-7. Example of warn_any_ptr_int_conv
#pragma warn_ptr_int_conv on short i, *ip void func() {
i = (short)ip;
/* WARNING: short type is not large enough to hold pointer. */
}
#pragma warn_any_ptr_int_conv on void bar() {
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
486 Freescale Semiconductor, Inc.
Chapter 31 Pragmas for Diagnostic Messages
i = (int)ip; /* WARNING: pointer to integral conversion. */
ip = (short *)i; /* WARNING: integral to pointer conversion. */
}
Remarks
By default, this pragma is off
.
31.12 warn_emptydecl
Controls the recognition of declarations without variables.
Syntax
#pragma warn_emptydecl on | off | reset
Remarks
If you enable this pragma, the compiler displays a warning message when it encounters a declaration with no variables.
Listing 31-8. Examples of empty declarations in C and C++
#pragma warn_emptydecl on int ; /* WARNING: empty variable declaration. */ int i; /* OK */ long j;; /* WARNING */ long j; /* OK */
Listing 31-9. Example of empty declaration in C++
#pragma warn_emptydecl on extern "C" {
}; /* WARNING */
By default, this pragma is disabled.
31.13 warn_extracomma
Controls the recognition of superfluous commas in enumerations.
Syntax
#pragma warn_extracomma on | off | reset
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
487
warn_filenamecaps
If you enable this pragma, the compiler issues a warning message when it encounters a
trailing comma in enumerations. For example, Listing 31-10 on page 488 is acceptable
source code but generates a warning message when you enable this setting.
Listing 31-10. Warning about extra commas
#pragma warn_extracomma on enum { mouse, cat, dog, };
/* WARNING: compiler expects an identifier after final comma. */
The compiler ignores terminating commas in enumerations when compiling source code that conforms to the ISO/IEC 9899-1999 ("C99") standard.
By default, this pragma is disabled.
31.14 warn_filenamecaps
Controls the recognition of conflicts involving case-sensitive filenames within user includes.
Syntax
#pragma warn_filenamecaps on | off | reset
Remarks
If you enable this pragma, the compiler issues a warning message when an
#include directive capitalizes a filename within a user include differently from the way the filename appears on a disk. It also detects use of "8.3" DOS filenames in Windows® operating systems when a long filename is available. Use this pragma to avoid porting problems to operating systems with case-sensitive file names.
By default, this pragma only checks the spelling of user includes such as the following:
#include "file"
For more information on checking system includes, see
.
By default, this pragma is off
.
31.15 warn_filenamecaps_system
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
488 Freescale Semiconductor, Inc.
Chapter 31 Pragmas for Diagnostic Messages
Controls the recognition of conflicts involving case-sensitive filenames within system includes.
Syntax
#pragma warn_filenamecaps_system on | off | reset
Remarks
If you enable this pragma along with warn_filenamecaps
, the compiler issues a warning message when an
#include
directive capitalizes a filename within a system include differently from the way the filename appears on a disk. It also detects use of "8.3" DOS filenames in Windows® systems when a long filename is available. This pragma helps avoid porting problems to operating systems with case-sensitive file names.
To check the spelling of system includes such as the following:
#include <file>
Use this pragma along with the warn_filenamecaps pragma.
By default, this pragma is off.
NOTE
Some SDKs (Software Developer Kits) use "colorful" capitalization, so this pragma may issue a lot of unwanted messages.
31.16 warn_hiddenlocals
Controls the recognition of a local variable that hides another local variable.
Syntax
#pragma warn_hiddenlocals on | off | reset
Remarks
When on
, the compiler issues a warning message when it encounters a local variable that
hides another local variable. An example appears in Listing 31-11 on page 489 .
Listing 31-11. Example of hidden local variables warning
#pragma warn_hiddenlocals on void func(int a)
{
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
489
warn_illpragma
{
int a; /* WARNING: this 'a' obscures argument 'a'.
}
}
This pragma does not correspond to any CodeWarrior IDE panel setting. By default, this setting is off
.
31.17 warn_illpragma
Controls the recognition of invalid pragma directives.
Syntax
#pragma warn_illpragma on | off | reset
Remarks
If you enable this pragma, the compiler displays a warning message when it encounters a pragma it does not recognize.
By default, this setting is off
.
31.18 warn_illtokenpasting
Controls whether or not to issue a warning message for improper preprocessor token pasting.
Syntax
#pragma warn_illtokenpasting on | off | reset
Remarks
An example of this is shown below:
#define PTR(x) x##* / PTR(y)
Token pasting is used to create a single token. In this example, y
and x
cannot be combined. Often the warning message indicates the macros uses "
##
" unnecessarily.
This pragma does not correspond to any panel setting. By default, this pragma is on
.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
490 Freescale Semiconductor, Inc.
Chapter 31 Pragmas for Diagnostic Messages
31.19 warn_illunionmembers
Controls whether or not to issue a warning message for invalid union members, such as unions with reference or non-trivial class members.
Syntax
#pragma warn_illunionmembers on | off | reset
Remarks
This pragma does not correspond to any panel setting. By default, this pragma is on
.
31.20 warn_impl_f2i_conv
Controls the issuing of warning messages for implicit float
-to- int
conversions.
Syntax
#pragma warn_impl_f2i_conv on | off | reset
Remarks
If you enable this pragma, the compiler issues a warning message for implicitly converting floating-point values to integral values.
Listing 31-12 on page 491 provides
an example.
Listing 31-12. Example of Implicit float
-to- int
Conversion
#pragma warn_impl_f2i_conv on float f; signed int si; int main()
{
f = si; /* WARNING */
#pragma warn_impl_f2i_conv off
si = f; /* OK */
}
By default, this pragma is on
.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
491
warn_impl_s2u_conv
31.21 warn_impl_i2f_conv
Controls the issuing of warning messages for implicit int
-to- float
conversions.
Syntax
#pragma warn_impl_i2f_conv on | off | reset
Remarks
If you enable this pragma, the compiler issues a warning message for implicitly converting integral values to floating-point values.
Listing 31-13 on page 492 shows an
example.
Listing 31-13. Example of implicit int
-to- float
conversion
#pragma warn_impl_i2f_conv on float f; signed int si; int main()
{
si = f; /* WARNING */
#pragma warn_impl_i2f_conv off
f = si; /* OK */
}
By default, this pragma is off
.
31.22 warn_impl_s2u_conv
Controls the issuing of warning messages for implicit conversions between the signed int and unsigned int
data types.
Syntax
#pragma warn_impl_s2u_conv on | off | reset
Remarks
If you enable this pragma, the compiler issues a warning message for implicitly converting either from signed int
to unsigned int
or vice versa. Listing 31-14 on page 492
provides an example.
Listing 31-14. Example of implicit conversions between signed int
and unsigned int
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
492 Freescale Semiconductor, Inc.
Chapter 31 Pragmas for Diagnostic Messages
#pragma warn_impl_s2u_conv on signed int si; unsigned int ui; int main()
{
ui = si; /* WARNING */
si = ui; /* WARNING */
#pragma warn_impl_s2u_conv off
ui = si; /* OK */
si = ui; /* OK */
}
By default, this pragma is enabled.
31.23 warn_implicitconv
Controls the issuing of warning messages for all implicit arithmetic conversions.
Syntax
#pragma warn_implicitconv on | off | reset
Remarks
If you enable this pragma, the compiler issues a warning message for all implicit arithmetic conversions when the destination type might not represent the source value.
Listing 31-15 on page 493 provides an example.
Listing 31-15. Example of Implicit Conversion
#pragma warn_implicitconv on float f; signed int si; unsigned int ui; int main()
{
f = si; /* WARNING */
si = f; /* WARNING */
ui = si; /* WARNING */
si = ui; /* WARNING */
}
NOTE
This option "opens the gate" for the checking of implicit conversions. The sub-pragmas warn_impl_f2i_conv
, warn_impl_i2f_conv
, and warn_impl_s2u_conv
control the classes of conversions checked.
By default, this pragma is off
.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
493
warn_largeargs
31.24 warn_largeargs
Controls the issuing of warning messages for passing non-"int" numeric values to unprototyped functions.
Syntax
#pragma warn_largeargs on | off | reset
Remarks
If you enable this pragma, the compiler issues a warning message if you attempt to pass a non-integer numeric value, such as a float
or long long
, to an unprototyped function when the require_prototypes
pragma is disabled.
This pragma does not correspond to any panel setting. By default, this pragma is off
.
31.25 warn_missingreturn
Issues a warning message when a function that returns a value is missing a return statement.
Syntax
#pragma warn_missingreturn on | off | reset
Remarks
An example is shown in Listing 31-16 on page 494 .
Listing 31-16. Example of warn_missingreturn pragma
#pragma warn_missingreturn on int func()
{
/* WARNING: no return statement. */
}
31.26 warn_no_side_effect
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
494 Freescale Semiconductor, Inc.
Chapter 31 Pragmas for Diagnostic Messages
Controls the issuing of warning messages for redundant statements.
Syntax
#pragma warn_no_side_effect on | off | reset
Remarks
If you enable this pragma, the compiler issues a warning message when it encounters a statement that produces no side effect. To suppress this warning message, cast the statement with
(void)
provides an example.
Listing 31-17. Example of Pragma warn_no_side_effect
#pragma warn_no_side_effect on void func(int a,int b)
{
a+b; /* WARNING: expression has no side effect */
(void)(a+b); /* OK: void cast suppresses warning. */
}
By default, this pragma is off
.
31.27 warn_padding
Controls the issuing of warning messages for data structure padding.
Syntax
#pragma warn_padding on | off | reset
Remarks
If you enable this pragma, the compiler warns about any bytes that were implicitly added after an ANSI C struct
member to improve memory alignment. Refer to the appropriate
Targeting
manual for more information on how the compiler pads data structures for a particular processor or operating system.
By default, this setting is off
.
31.28 warn_pch_portability
Controls whether or not to issue a warning message when
#pragmaonceon
is used in a precompiled header.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
495
warn_possunwant
Syntax
#pragma warn_pch_portability on | off | reset
Remarks
If you enable this pragma, the compiler issues a warning message when you use
#pragma once on
in a precompiled header. This helps you avoid situations in which transferring a precompiled header from machine to machine causes the precompiled header to produce different results. For more information, see pragma once
.
This pragma does not correspond to any panel setting. By default, this setting is off
.
31.29 warn_possunwant
Controls the recognition of possible unintentional logical errors.
Syntax
#pragma warn_possunwant on | off | reset
Remarks
If you enable this pragma, the compiler checks for common, unintended logical errors:
• An assignment in either a logical expression or the conditional portion of an if
, while
, or for
expression. This warning message is useful if you use
=
when you mean to use
==
. The listing below shows an example.
Listing 31-18. Confusing = and == in Comparisons
if (a=b) f(); /* WARNING: a=b is an assignment. */ if ((a=b)!=0) f(); /* OK: (a=b)!=0 is a comparison. */ if (a==b) f(); /* OK: (a==b) is a comparison. */
• An equal comparison in a statement that contains a single expression. This check is useful if you use
==
when you meant to use
=
. The listing below shows an example.
Listing 31-19. Confusing = and == Operators in Assignments
a == 0; // WARNING: This is a comparison.
a = 0; // OK: This is an assignment, no warning
• A semicolon (
;
) directly after a while
, if
, or for
statement.
For example, the listing below, generates a warning message.
Listing 31-20. Empty statement
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
496 Freescale Semiconductor, Inc.
Chapter 31 Pragmas for Diagnostic Messages
i = sockcount(); while (--i); /* WARNING: empty loop. */
matchsock(i);
If you intended to create an infinite loop, put white space or a comment between the while statement and the semicolon. The statements in the listing below, suppress the above error or warning messages.
Listing 31-21. Intentional empty statements
while (i++) ; /* OK: White space separation. */ while (i++) /* OK: Comment separation */ ;
By default, this pragma is off
.
31.30 warn_ptr_int_conv
Controls the recognition the conversion of pointer values to incorrectly-sized integral values.
Syntax
#pragma warn_ptr_int_conv on | off | reset
Remarks
If you enable this pragma, the compiler issues a warning message if an expression attempts to convert a pointer value to an integral type that is not large enough to hold the pointer value.
Listing 31-22. Example for #pragma warn_ptr_int_conv
#pragma warn_ptr_int_conv on char *my_ptr; char too_small = (char)my_ptr; /* WARNING: char is too small. */
See also
"warn_any_ptr_int_conv" on page 231
.
By default, this pragma is off.
31.31 warn_resultnotused
Controls the issuing of warning messages when function results are ignored.
Syntax
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
497
warn_undefmacro
#pragma warn_resultnotused on | off | reset
Remarks
If you enable this pragma, the compiler issues a warning message when it encounters a statement that calls a function without using its result. To prevent this, cast the statement with
(void)
.
Listing 31-23 on page 498 provides an example.
Listing 31-23. Example of Function Calls with Unused Results
#pragma warn_resultnotused on extern int bar(); void func()
{
bar(); /* WARNING: result of function call is not used. */
void(bar()); /* OK: void cast suppresses warning. */
}
This pragma does not correspond to any panel setting. By default, this pragma is off.
31.32 warn_undefmacro
Controls the detection of undefined macros in
#if
and
#elif
directives.
Syntax
#pragma warn_undefmacro on | off | reset
Remarks
Listing 31-24 on page 498 provides an example.
Listing 31-24. Example of Undefined Macro
#if BADMACRO == 4 /* WARNING: undefined macro. */
Use this pragma to detect the use of undefined macros (especially expressions) where the default value 0 is used. To suppress this warning message, check if defined first.
NOTE
A warning message is only issued when a macro is evaluated. A short-circuited "
&&
" or "
||
" test or unevaluated "
?:
" will not produce a warning message.
By default, this pragma is off
.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
498 Freescale Semiconductor, Inc.
Chapter 31 Pragmas for Diagnostic Messages
31.33 warn_uninitializedvar
Controls the compiler to perform some dataflow analysis and emits warning messages whenever local variables are initialized before being used.
Syntax
#pragma warn_uninitializedvar on | off | reset
Remarks
This pragma has no corresponding setting in the CodeWarrior IDE. By default, this pragma is on
.
NOTE
#pragma warn_possiblyuninitializedvar
is a superset of
#pragma warn_uninitializedvar
.
31.34 warn_unusedarg
Controls the recognition of unreferenced arguments.
Syntax
#pragma warn_unusedarg on | off | reset
Remarks
If you enable this pragma, the compiler issues a warning message when it encounters an argument you declare but do not use.
This check helps you find arguments that you either misspelled or did not use in your program. The listing below shows an example.
Listing 31-25. Warning about unused function arguments
void func(int temp, int error);
{
error = do_something(); /* WARNING: temp is unused. */
}
To prevent this warning, you can declare an argument in a few ways:
• Use the pragma unused
, as shown in the listing below.
Listing 31-26. Using pragma unused() to prevent unused argument messages
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
499
warn_unusedvar
void func(int temp, int error)
{
#pragma unused (temp)
/* Compiler does not warn that temp is not used. */
error=do_something();
}
• Do not give the unused argument a name. The listing below shows an example.
The compiler allows this feature in C++ source code. To allow this feature in C source code, disable ANSI strict checking.
Listing 31-27. Unused, Unnamed Arguments
void func(int /* temp */, int error)
{
/* Compiler does not warn that "temp" is not used. */
error=do_something();
}
This pragma corresponds to the Unused Arguments setting in the C/C++ Warnings
Panel . By default, this pragma is off
.
31.35 warn_unusedvar
Controls the recognition of unreferenced variables.
Syntax
#pragma warn_unusedvar on | off | reset
Remarks
If you enable this pragma, the compiler issues a warning message when it encounters a variable you declare but do not use.
This check helps you find variables that you either misspelled or did not use in your program.
shows an example.
Listing 31-28. Unused Local Variables Example
int error; void func(void)
{
int temp, errer; /* NOTE: errer is misspelled. */
error = do_something(); /* WARNING: temp and errer are unused. */
}
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
500 Freescale Semiconductor, Inc.
Chapter 31 Pragmas for Diagnostic Messages
If you want to use this warning but need to declare a variable that you do not use, include the pragma unused
, as in
.
Listing 31-29. Suppressing Unused Variable Warnings
void func(void)
{
int i, temp, error;
#pragma unused (i, temp) /* Do not warn that i and temp */
error = do_something(); /* are not used */
}
By default, this pragma is off.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
501
warn_unusedvar
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
502 Freescale Semiconductor, Inc.
Chapter 32
Pragmas for Preprocessing
This section explains the following pragmas for Preprocessing:
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
32.1 check_header_flags
Controls whether or not to ensure that a precompiled header's data matches a project's target settings.
Syntax
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
503
faster_pch_gen
#pragma check_header_flags on | off | reset
Remarks
This pragma affects precompiled headers only.
If you enable this pragma, the compiler verifies that the precompiled header's preferences for double size, int size, and floating point math correspond to the build target's settings.
If they do not match, the compiler generates an error message.
If your precompiled header file depends on these settings, enable this pragma. Otherwise, disable it.
This pragma does not correspond to any CodeWarrior IDE panel setting. By default, this pragma is off
.
32.2 faster_pch_gen
Controls the performance of precompiled header generation.
Syntax
#pragma faster_pch_gen on | off | reset
Remarks
If you enable this pragma, generating a precompiled header can be much faster, depending on the header structure. However, the precompiled file can also be slightly larger.
This pragma does not correspond to any panel setting. By default, this setting is off
.
32.3 flat_include
Controls whether or not to ignore relative path names in
#include
directives.
Syntax
#pragma flat_include on | off | reset
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
504 Freescale Semiconductor, Inc.
For example, when on
, the compiler converts this directive
Chapter 32 Pragmas for Preprocessing
#include <sys/stat.h > to
#include <stat.h>
Use this pragma when porting source code from a different operating system, or when a
CodeWarrior IDE project's access paths cannot reach a given file.
By default, this pragma is off
.
32.4 fullpath_file
Controls if
__FILE__
macro expands to a full path or the base file name.
Syntax
#pragma fullpath_file on | off | reset
Remarks
When this pragma on
, the
__FILE__
macro returns a full path to the file being compiled, otherwise it returns the base file name.
32.5 fullpath_prepdump
Shows the full path of included files in preprocessor output.
Syntax
#pragma fullpath_prepdump on | off | reset
Remarks
If you enable this pragma, the compiler shows the full paths of files specified by the
#include
directive as comments in the preprocessor output. Otherwise, only the file name portion of the path appears.
By default, this pragma is off
.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
505
keepcomments
32.6 keepcomments
Controls whether comments are emitted in the preprocessor output.
Syntax
#pragma keepcomments on | off | reset
Remarks
By default, this pragma is off
.
32.7 line_prepdump
Shows
#line
directives in preprocessor output.
Syntax
#pragma line_prepdump on | off | reset
Remarks
If you enable this pragma,
#line
directives appear in preprocessing output. The compiler also adjusts line spacing by inserting empty lines.
Use this pragma with the command-line compiler's
-E
option to make sure that
#line directives are inserted in the preprocessor output.
By default, this pragma is off
.
32.8 macro_prepdump
Controls whether the compiler emits
#define
and
#undef
directives in preprocessing output.
Syntax
#pragma macro_prepdump on | off | reset
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
506 Freescale Semiconductor, Inc.
Chapter 32 Pragmas for Preprocessing
Use this pragma to help unravel confusing problems like macros that are aliasing identifiers or where headers are redefining macros unexpectedly.
32.9 msg_show_lineref
Controls diagnostic output involving
#line
directives to show line numbers specified by the
#line
directives in error and warning messages.
Syntax
#pragma msg_show_lineref on | off | reset
Remarks
This pragma does not correspond to any CodeWarrior IDE panel setting. By default, this pragma is on
.
32.10 msg_show_realref
Controls diagnostic output involving
#line
directives to show actual line numbers in error and warning messages.
Syntax
#pragma msg_show_realref on | off | reset
Remarks
This pragma does not correspond to any CodeWarrior IDE panel setting. By default, this pragma is on
.
32.11 notonce
Controls whether or not the compiler lets included files be repeatedly included, even with
#pragma once on
.
Syntax
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
507
old_pragma_once
#pragma notonce
Remarks
If you enable this pragma, files can be repeatedly
#include
, even if you have enabled
#pragma once on
. For more information, see
This pragma does not correspond to any CodeWarrior IDE panel setting.
32.12 old_pragma_once
This pragma is no longer available.
32.13 once
Controls whether or not a header file can be included more than once in the same compilation unit.
Syntax
#pragma once [ on ]
Remarks
Use this pragma to ensure that the compiler includes header files only once in a source file. This pragma is especially useful in precompiled header files.
There are two versions of this pragma:
#pragma once and
#pragma once on
Use
#pragma once
in a header file to ensure that the header file is included only once in a source file. Use
#pragma once on
in a header file or source file to ensure that any file is included only once in a source file. When a once
option or pragma is used, a header file of same name in another directory is not included.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
508 Freescale Semiconductor, Inc.
Chapter 32 Pragmas for Preprocessing
Beware that when using
#pragma once on
, precompiled headers transferred from one host machine to another might not give the same results during compilation. This inconsistency is because the compiler stores the full paths of included files to distinguish between two distinct files that have identical file names but different paths. Use the warn_pch_portability
pragma to issue a warning message when you use
#pragma once on
in a precompiled header.
Also, if you enable the old_pragma_once on
pragma, the once
pragma completely ignores path names.
This pragma does not correspond to any panel setting. By default, this pragma is off
.
32.14 pop, push
Saves and restores pragma settings.
Syntax
#pragma push
#pragma pop
Remarks
The pragma push
saves all the current pragma settings. The pragma pop
restores all the pragma settings that resulted from the last push
pragma. For example, see
.
Listing 32-1. push and pop example
#pragma ANSI_strict on
#pragma push /* Saves all compiler settings. */
#pragma ANSI_strict off
#pragma pop /* Restores ANSI_strict to on. */
This pragma does not correspond to any panel setting. By default, this pragma is off
.
Tip
Pragmas directives that accept on
| off
| reset
already form a stack of previous option values. It is not necessary to use
#pragma pop
or
#pragma push
with such pragmas.
32.15 pragma_prepdump
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
509
precompile_target
Controls whether pragma directives in the source text appear in the preprocessing output.
Syntax
#pragma pragma_prepdump on | off | reset
Remarks
By default, this pragma is off
.
Tip
When submitting bug reports with a preprocessor dump, be sure this option is enabled.
32.16 precompile_target
Specifies the file name for a precompiled header file.
Syntax
#pragma precompile_target filename
Parameters
filename
A simple file name or an absolute path name. If filename is a simple file name, the compiler saves the file in the same folder as the source file. If filename is a path name, the compiler saves the file in the specified folder.
Remarks
If you do not specify the file name, the compiler gives the precompiled header file the same name as its source file.
shows sample source code from a precompiled header source file. By using the predefined symbols
__cplusplus
and the pragma precompile_target
, the compiler can use the same source code to create different precompiled header files for C and C++.
Listing 32-2. Using #pragma precompile_target
#ifdef __cplusplus
#pragma precompile_target "MyCPPHeaders"
#else
#pragma precompile_target "MyCHeaders"
#endif
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
510 Freescale Semiconductor, Inc.
This pragma does not correspond to any panel setting.
Chapter 32 Pragmas for Preprocessing
32.17 simple_prepdump
Controls the suppression of comments in preprocessing output.
Syntax
#pragma
simple_prepdump on | off | reset
Remarks
By default, the compiler adds comments about the current include file being in preprocessing output. Enabling this pragma disables these comments.
By default, this pragma is off
.
32.18 space_prepdump
Controls whether or not the compiler removes or preserves whitespace in the preprocessor's output.
Syntax
#pragma space_prepdump on | off | reset
Remarks
This pragma is useful for keeping the starting column aligned with the original source code, though the compiler attempts to preserve space within the line. This pragma does not apply to expanded macros.
By default, this pragma is off
.
32.19 srcrelincludes
Controls the lookup of
#include
files.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
511
syspath_once
Syntax
#pragma srcrelincludes on | off | reset
Remarks
When on
, the compiler looks for
#include
files relative to the previously included file (not just the source file). When off
, the compiler uses the CodeWarrior IDE's access paths or the access paths specified with the
-ir
option.
Use this pragma when multiple files use the same file name and are intended to be included by another header file in that directory. This is a common practice in UNIX programming.
This pragma corresponds to the Source-relative includes option in the Access Paths panel.By default, this pragma is off
.
32.20 syspath_once
Controls how included files are treated when
#pragmaonce
is enabled.
Syntax
#pragma syspath_once on | off | reset
Remarks
When this pragma and pragma once
are set to on
, the compiler distinguishes between identically-named header files referred to in
#include <file-name> and
#include "file-name"
When this pragma is off
and pragma once is on
, the compiler will ignore a file that uses a directive if it has previously encountered another directive of the form for an identicallynamed header file.
#include <file-name>
#include "file-name"
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
512 Freescale Semiconductor, Inc.
Chapter 32 Pragmas for Preprocessing
This pragma does not correspond to any panel setting. By default, this setting is on
.
Listing 32-3. Pragma syspath_once example
#pragma syspath_once off
#pragma once on /* Include all subsequent files only once. */
#include "sock.h"
#include <sock.h> /* Skipped because syspath_once is off. */
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
513
syspath_once
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
514 Freescale Semiconductor, Inc.
Chapter 33
Pragmas for Library and Linking
This section explains the following pragmas for library and linking:
•
•
•
•
33.1 always_import
Controls whether or not
# include
directives are treated as
#pragma import
directives.
Syntax
#pragma always_import on | off | reset
Remarks
If you enable this pragma, the compiler treats all
#include
statements as
#pragma import statements.
This pragma does not correspond to any CodeWarrior IDE panel setting. By default, this pragma is off
.
33.2 export
Controls the exporting of data and functions to be accessible from outside a program or library.
Syntax
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
515
import
#pragma export on | off | reset
#pragma export list
name1 [,
name2 , ...]
name1 , name2
Names of functions or global variables to export.
Remarks
When using the
#pragma export on
format, all functions in the source file being compiled will be accessible from outside the program or library that the compiler and linker are building.
Use the
#pragma export list
format to specify global variables and functions for exporting.
In C++, this form of the pragma applies to all variants of an overloaded function. You
Listing 33-1. Example of an Exported List
extern int f(),g;
#pragma export list f,g
33.3 import
Controls the importing of global data or functions.
Syntax
#pragma import on | off | reset
#pragma import list
name1 [,
name2 , ...]
name1 , name2
Names of functions or global variables to import.
Remarks
When using the
#pragma import on
format, all functions are automatically imported.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
516 Freescale Semiconductor, Inc.
Chapter 33 Pragmas for Library and Linking
Use the
#pragma import list
format to specify data or functions for importing. In C++, this form of the pragma applies to all variants of an overloaded function. You cannot use this pragma for C++ member functions or static class members.
shows an example:
Listing 33-2. Example of an Imported List
extern int f(),g;
#pragma import list f,g
This pragma does not correspond to any CodeWarrior IDE panel setting. By default, this pragma is off
.
33.4 lib_export
Controls the exporting of data or functions.
Syntax
#pragma lib_export on | off | reset
#pragma lib_export list
name1 [,name2 ]*
Remarks
When using the
#pragma lib_export on
format, the linker marks all data and functions that are within the pragma's scope for export.
Use the
#pragma lib_export list
format to tag specific data or functions for exporting. In C
++, this form of the pragma applies to all variants of an overloaded function. You cannot use this pragma for C++ member functions or static class members.
shows an example:
Listing 33-3. Example of a lib_export
List
extern int f(),g;
#pragma lib_export list f,g
This pragma does not correspond to any panel setting. By default, this pragma is disabled.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
517
lib_export
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
518 Freescale Semiconductor, Inc.
Chapter 34
Pragmas for Code Generation
This section explains the following pragmas for code generation:
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
34.1 aggressive_inline
Instructs compiler to be more aggressive when inlining.
Syntax
#pragma aggressive_inline on | off | reset
Remarks
The IPA-based inliner (-ipa file) will inline more functions when this option
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
519
dont_reuse_strings
is enabled. This option can cause code bloat in programs that overuse inline functions. Default is off.
34.2 dont_reuse_strings
Controls whether or not to store identical character string literals separately in object code.
Syntax
#pragma dont_reuse_strings on | off | reset
Remarks
Normally, C and C++ programs should not modify character string literals. Enable this pragma if your source code follows the unconventional practice of modifying them.
If you enable this pragma, the compiler separately stores identical occurrences of character string literals in a source file.
If this pragma is disabled, the compiler stores a single instance of identical string literals in a source file. The compiler reduces the size of the object code it generates for a file if the source file has identical string literals.
The compiler always stores a separate instance of a string literal that is used to initialize a character array.
Listing 34-1 on page 520 shows an example.
Although the source code contains 3 identical string literals,
"cat"
, the compiler will generate 2 instances of the string in object code. The compiler will initialize str1
and str2 to point to the first instance of the string and will initialize str3
to contain the second instance of the string.
Using str2
to modify the string it points to also modifies the string that str1
points to. The array str3
may be safely used to modify the string it points to without inadvertently changing any other strings.
By default, this pragma is off
.
Listing 34-1. Reusing string literals
#pragma dont_reuse_strings off void strchange(void)
{
const char* str1 = "cat";
char* str2 = "cat";
char str3[] = "cat";
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
520 Freescale Semiconductor, Inc.
Chapter 34 Pragmas for Code Generation
*str2 = 'h'; /* str1 and str2 point to "hat"! */
str3[0] = 'b';
/* OK: str3 contains "bat", *str1 and *str2 unchanged.
}
34.3 enumsalwaysint
Specifies the size of enumerated types.
Syntax
#pragma enumsalwaysint on | off | reset
Remarks
If you enable this pragma, the C/C++ compiler makes an enumerated type the same size as an int
. If an enumerated constant is larger than int
, the compiler generates an error message. Otherwise, the compiler makes an enumerated type the size of any integral type.
It chooses the integral type with the size that most closely matches the size of the largest enumerated constant. The type could be as small as a char
or as large as a long long
.
shows an example.
Listing 34-2. Example of Enumerations the Same as Size as int
enum SmallNumber { One = 1, Two = 2 };
/* If you enable enumsalwaysint, this type is
the same size as an int. Otherwise, this type is
the same size as a char. */ enum BigNumber
{ ThreeThousandMillion = 3000000000 };
/* If you enable enumsalwaysint, the compiler might
generate an error message. Otherwise, this type is
the same size as a long long. */
By default, this pragma is off
.
34.4 enums_signed
Changes the underlying enumeration type search order.
Syntax
#pragma enums_signed on | off | reset
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
521
errno_name
Remarks
Enabling this option changes the underlying enumeration type search order. The underlying type for an enumeration where all enumerators are >= 0 is the first one of these types in which all values can be represented: signed char (*) unsigned char signed short (*) unsigned short signed int (*) unsigned int signed long (*) unsigned long signed long long (*) unsigned long long
Types with (*) are only considered with "#pragma enums_signed on". This option has no effect when #pragma enumsalwaysint is "on".
34.5 errno_name
Tells the optimizer how to find the errno
identifier.
Syntax
#pragma errno_name id | ...
Remarks
When this pragma is used, the optimizer can use the identifier errno
(either a macro or a function call) to optimize standard C library functions better. If not used, the optimizer makes worst-case assumptions about the effects of calls to the standard C library.
NOTE
The MSL C library already includes a use of this pragma, so you would only need to use it for third-party C libraries.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
522 Freescale Semiconductor, Inc.
Chapter 34 Pragmas for Code Generation
If errno
resolves to a variable name, specify it like this:
#pragma errno_name _Errno
If errno
is a function call accessing ordinarily inaccessible global variables, use this form:
#pragma errno_name ...
Otherwise, do not use this pragma to prevent incorrect optimizations.
This pragma does not correspond to any panel setting. By default, this pragma is unspecified (worst case assumption).
34.6 explicit_zero_data
Controls the placement of zero-initialized data.
Syntax
#pragma explicit_zero_data on | off | reset
Remarks
Places zero-initialized data into the initialized data section instead of the BSS section when on
.
By default, this pragma is off
.
34.7 float_constants
Controls how floating pointing constants are treated.
Syntax
#pragma float_constants on | off | reset
Remarks
If you enable this pragma, the compiler assumes that all unqualified floating point constant values are of type float
, not double
. This pragma is useful when porting source code for programs optimized for the " float
" rather than the " double
" type.
When you enable this pragma, you can still explicitly declare a constant value as double by appending a "D" suffix.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
523
instmgr_file
This pragma does not correspond to any panel setting. By default, this pragma is disabled.
34.8 instmgr_file
Controls where the instance manager database is written, to the target data directory or to a separate file.
Syntax
#pragma instmgr_file "name"
Remarks
When the Use Instance Manager option is on, the IDE writes the instance manager database to the project's data directory. If the
#pragma instmgr_file
is used, the database is written to a separate file.
Also, a separate instance file is always written when the command-line tools are used.
NOTE
Should you need to report a bug, you can use this option to create a separate instance manager database, which can then be sent to technical support with your bug report.
NOTE
34.9 longlong
Controls the availability of the long long
type.
Syntax
#pragma longlong on | off | reset
Remarks
When this pragma is enabled and the compiler is translating "C90" source code (ISO/IEC
9899-1990 standard), the compiler recognizes a data type named long long
. The long long type holds twice as many bits as the long
data type.
This pragma does not correspond to any CodeWarrior IDE panel setting.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
524 Freescale Semiconductor, Inc.
Chapter 34 Pragmas for Code Generation
By default, this pragma is on for processors that support this type. It is off
when generating code for processors that do not support, or cannot turn on, the long long
type.
34.10 longlong_enums
Controls whether or not enumerated types may have the size of the long long
type.
Syntax
#pragma longlong_enums on | off | reset
Remarks
This pragma lets you use enumerators that are large enough to be long long integers. It is ignored if you enable the enumsalwaysint
pragma (described in "enumsalwaysint" on page
This pragma does not correspond to any panel setting. By default, this setting is enabled.
34.11 min_enum_size
Specifies the size, in bytes, of enumeration types.
Syntax
#pragma min_enum_size 1 | 2 | 4
Remarks
Turning on the enumsalwaysint
pragma overrides this pragma. The default is 1.
34.12 pool_strings
Controls how string literals are stored.
Syntax
#pragma pool_strings on | off | reset
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
525
readonly_strings
Remarks
If you enable this pragma, the compiler collects all string constants into a single data object so your program needs one data section for all of them. If you disable this pragma, the compiler creates a unique data object for each string constant. While this decreases the number of data sections in your program, on some processors it also makes your program bigger because it uses a less efficient method to store the address of the string.
This pragma is especially useful if your program is large and has many string constants or uses the CodeWarrior Profiler.
NOTE
If you enable this pragma, the compiler ignores the setting of the pcrelstrings
pragma.
34.13 readonly_strings
Controls whether string objects are placed in a read-write or a read-only data section.
Syntax
#pragma readonly_strings on | off | reset
Remarks
If you enable this pragma, literal strings used in your source code are output to the readonly data section instead of the global data section. In effect, these strings act like constchar*
, even though their type is really char*
.
This pragma does not correspond to any IDE panel setting.
34.14 reverse_bitfields
Controls whether or not the compiler reverses the bitfield allocation.
Syntax
#pragma reverse_bitfields on | off | reset
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
526 Freescale Semiconductor, Inc.
Chapter 34 Pragmas for Code Generation
This pragma reverses the bitfield allocation, so that bitfields are arranged from the opposite side of the storage unit from that ordinarily used on the target. The compiler still orders the bits within a single bitfield such that the lowest-valued bit is in the right-most position.
This pragma does not correspond to any panel setting. By default, this pragma is disabled.
NOTE
Limitation: please be aware of the following limitations when this pragma is set to on
:- the data types of the bit-fields must be the same data type- the structure ( struct
) or class
must not contain non-bit-field members; however, the structure ( struct
) can be the member of another structure
34.15 store_object_files
Controls the storage location of object data, either in the target data directory or as a separate file.
Syntax
#pragma store_object_files on | off | reset
Remarks
By default, the IDE writes object data to the project's target data directory. When this pragma is on, the object data is written to a separate object file.
NOTE
For some targets, the object file emitted may not be recognized as actual object data.
This pragma does not correspond to any panel setting. By default, this pragma is off
.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
527
store_object_files
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
528 Freescale Semiconductor, Inc.
Chapter 35
Pragmas for Optimization
Pragmas for Optimization should be placed outside the functions to which it is desired that they be applied.
This chapter explains the following pragmas:
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
35.1 global_optimizer
Controls whether the Frontend IR Optimizer is invoked by the compiler.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
529
ipa
Syntax
#pragma global_optimizer on | off | reset
Remarks
In most compilers, this #pragma determines whether the Frontend IR Optimizer is invoked. If disabled, only simple optimizations and back-end optimizations are performed.
NOTE
This is not the same as
#pragmaoptimization_level
. The Frontend
IR Optimizer is invoked even at optimization_level0
if
#pragmaglobal_optimizer
is enabled.
This pragma does not correspond to any panel setting. By default, this setting is on
.
35.2 ipa
Specifies how to apply interprocedural analysis optimizations.
Syntax
#pragma ipa program | file | on | function | off
Remarks
Place this pragma at the beginning of a source file, before any functions or data have been defined. There are three levels of interprocedural analysis:
• program-level: the compiler translates all source files in a program then optimizes object code for the entire program
• file-level: the compiler translates each file and applies this optimization to the file
• function-level: the compiler does not apply interprocedural optimization
The options file
and on
are equivalent. The options function
and off
are equivalent.
35.3 ipa_inline_max_auto_size
Determines the maximum complexity for an auto-inlined function.
Syntax
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
530 Freescale Semiconductor, Inc.
Chapter 35 Pragmas for Optimization
#pragma ipa_inline_max_auto_size (intval)
Parameters
intval
The intval
value is an approximation of the number of statements in a function, the current default value is 500, which is approximately equal to 100 satement function.
Selecting a zero value will disable the IPA auto inlining.
Remarks
The size of the code objects that are not referenced by address and are only called once is specified above a certain threshold using this pragma, preventing them from being marked as inline.
35.4 ipa_not_complete
Controls the usage of Complete Program IPA mode by the compiler.
Syntax
#pragma ipa_not_complete on | off | reset
Remarks
In Complete Program IPA mode, the compiler assumes that the IPA graph is complete and that there are no external entry points other than main()
, static initialization or force export functions. The compiler will not use this mode if the
The Complete Program IPA mode is not used by the compiler if:
• the program has no main()
and no force export functions.
• the pragma is on the context of main()
or force export functions.
NOTE
The compiler will be more aggressive in the Complete
Program IPA mode.
Any extern object that is not main()
, static initialization code or force export
and not directly or indirectly used, will be deadstipped by the compiler and will not appear in the object and/or executable files. By default, this setting is off
.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
531
load_store_elimination
35.5 load_store_elimination
Controls the use of load store elimination optimization.
Syntax
#pragma load_store_elimination on|off|reset
Remarks
This optimization will track all loads and stores within the entire function and figure out which ones are unnecessary and deletes them.
This optimization will be enabled by default only when the opt level is 3 or higher.
This pragma does not correspond to any panel setting.
35.6 opt_common_subs
Controls the use of common subexpression optimization.
Syntax
#pragma opt_common_subs on | off | reset
Remarks
If you enable this pragma, the compiler replaces similar redundant expressions with a single expression. For example, if two statements in a function both use the expression a * b * c + 10 the compiler generates object code that computes the expression only once and applies the resulting value to both statements.
The compiler applies this optimization to its own internal representation of the object code it produces.
This pragma does not correspond to any panel setting. By default, this settings is related to the
pragma.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
532 Freescale Semiconductor, Inc.
Chapter 35 Pragmas for Optimization
35.7 opt_dead_assignments
Controls the use of dead store optimization.
Syntax
#pragma opt_dead_assignments on | off | reset
Remarks
If you enable this pragma, the compiler removes assignments to unused variables before reassigning them.
This pragma does not correspond to any panel setting. By default, this settings is related to the
"global_optimizer" on page 471
level.
35.8 opt_dead_code
Controls the use of dead code optimization.
Syntax
#pragma opt_dead_code on | off | reset
Remarks
If you enable this pragma, the compiler removes a statement that other statements never execute or call.
This pragma does not correspond to any panel setting. By default, this settings is related to the
"global_optimizer" on page 471
level.
35.9 opt_lifetimes
Controls the use of lifetime analysis optimization.
Syntax
#pragma opt_lifetimes on | off | reset
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
533
opt_loop_invariants
If you enable this pragma, the compiler uses the same processor register for different variables that exist in the same routine but not in the same statement.
This pragma does not correspond to any panel setting. By default, this settings is related to the
"global_optimizer" on page 471
level.
35.10 opt_loop_invariants
Controls the use of loop invariant optimization.
Syntax
#pragma opt_loop_invariants on | off | reset
Remarks
If you enable this pragma, the compiler moves all computations that do not change inside a loop outside the loop, which then runs faster.
This pragma does not correspond to any panel setting.
35.11 opt_propagation
Controls the use of copy and constant propagation optimization.
Syntax
#pragma opt_propagation on | off | reset
Remarks
If you enable this pragma, the compiler replaces multiple occurrences of one variable with a single occurrence.
This pragma does not correspond to any panel setting. By default, this settings is related to the
"global_optimizer" on page 471
level.
35.12 opt_strength_reduction
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
534 Freescale Semiconductor, Inc.
Chapter 35 Pragmas for Optimization
Controls the use of strength reduction optimization.
Syntax
#pragma opt_strength_reduction on | off | reset
Remarks
If you enable this pragma, the compiler replaces array element arithmetic instructions with pointer arithmetic instructions to make loops faster.
This pragma does not correspond to any panel setting. By default, this settings is related to the
"global_optimizer" on page 471
level.
35.13 opt_strength_reduction_strict
Uses a safer variation of strength reduction optimization.
Syntax
#pragma opt_strength_reduction_strict on | off | reset
Remarks
Like the
opt_strength_reduction pragma, this setting replaces multiplication instructions
that are inside loops with addition instructions to speed up the loops. However, unlike the regular strength reduction optimization, this variation ensures that the optimization is only applied when the array element arithmetic is not of an unsigned type that is smaller than a pointer type.
This pragma does not correspond to any panel setting. The default varies according to the compiler.
35.14 opt_unroll_loops
Controls the use of loop unrolling optimization.
Syntax
#pragma opt_unroll_loops on | off | reset
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
535
opt_vectorize_loops
If you enable this pragma, the compiler places multiple copies of a loop's statements inside a loop to improve its speed.
This pragma does not correspond to any panel setting. By default, this settings is related to the
"global_optimizer" on page 471
level.
35.15 opt_vectorize_loops
Controls the use of loop vectorizing optimization.
Syntax
#pragma opt_vectorize_loops on | off | reset
Remarks
If you enable this pragma, the compiler improves loop performance.
NOTE
Do not confuse loop vectorizing with the vector instructions available in some processors. Loop vectorizing is the rearrangement of instructions in loops to improve performance.
This optimization does not optimize a processor's vector data types.
By default, this pragma is off
.
35.16 optimization_level
Controls global optimization.
Syntax
#pragma optimization_level 0 | 1 | 2 | 3 | 4 | reset
Remarks
This pragma specifies the degree of optimization that the global optimizer performs.
To select optimizations, use the pragma optimization_level with an argument from 0 to 4.
The higher the argument, the more optimizations performed by the global optimizer. The reset argument specifies the previous optimization level.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
536 Freescale Semiconductor, Inc.
Chapter 35 Pragmas for Optimization
For more information on the optimization the compiler performs for each optimization level, refer to the Targeting manual for your target platform.
These pragmas correspond to the settings in the Global Optimizations panel. By default, this pragma is disabled.
35.17 optimize_for_size
Controls optimization to reduce the size of object code.
#pragma optimize_for_size on | off | reset
Remarks
This setting lets you choose what the compiler does when it must decide between creating small code or fast code. If you enable this pragma, the compiler creates smaller object code at the expense of speed. It also ignores the inline
directive and generates function calls to call any function declared inline
. If you disable this pragma, the compiler creates faster object code at the expense of size.
The pragma corresponds to the Optimize for Size setting on the Global Optimizations panel.
35.18 optimizewithasm
Controls optimization of assembly language.
Syntax
#pragma optimizewithasm on | off | reset
Remarks
If you enable this pragma, the compiler also optimizes assembly language statements in
C/C++ source code.
This pragma does not correspond to any panel setting. By default, this pragma is disabled.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
537
strictheaderchecking
35.19 pack
Stores data to reduce data size instead of improving execution performance.
Syntax
#pragma pack()
#pragma pack(0 |
n | push | pop) n
One of these integer values: 1, 2, 4, 8, or 16.
Remarks
Use this pragma to align data to use less storage even if the alignment might affect program performance or does not conform to the target platform's application binary interface (ABI).
If this pragma's argument is a power of 2 from 1 to 16, the compiler will store subsequent data structures to this byte alignment.
The push
argument saves this pragma's setting on a stack at compile time. The pop argument restores the previously saved setting and removes it from the stack. Using this pragma with no argument or with 0 as an argument specifies that the compiler will use
ABI-conformant alignment.
Not all processors support misaligned accesses, which could cause a crash or incorrect results. Even on processors which allow misaligned access, your program's performance might be reduced. Your program may have better performance if it treats the packed structure as a byte stream, then packs and unpacks each byte from the stream.
NOTE
Pragma pack
is implemented somewhat differently by most compiler vendors, especially when used with bitfields. If you need portability, you are probably better off using explicit shift and mask operations in your program instead of bitfields.
35.20 strictheaderchecking
Controls how strict the compiler checks headers for standard C library functions.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
538 Freescale Semiconductor, Inc.
Chapter 35 Pragmas for Optimization
Syntax
#pragma strictheaderchecking on | off | reset
Remarks
The 3.2 version compiler recognizes standard C library functions. If the correct prototype is used, and, in C++, if the function appears in the " std
" or root namespace, the compiler recognizes the function, and is able to optimize calls to it based on its documented effects.
When this #pragma is on (default), in addition to having the correct prototype, the declaration must also appear in the proper standard header file (and not in a user header or source file).
This pragma does not correspond to any panel setting. By default, this pragma is on.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
539
strictheaderchecking
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
540 Freescale Semiconductor, Inc.
Chapter 36
Pragmas for AIOP Compiler
This chapter describes the pragmas that control how the compiler translates source code into instructions and data.
•
•
•
•
•
36.1 Diagnostic Pragmas
This section describes the diagnostic pragmas.
•
incompatible_return_small_structs
•
incompatible_sfpe_double_params
36.1.1 incompatible_return_small_structs
Warns when returning structures using the R3 and R4 registers.
Syntax
#pragma incompatible_return_small_structs on | off | reset
Remarks
This pragma makes CodeWarrior-built object files more compatible with those created using a GNU compiler.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
541
Debugging Information Pragmas
The AIOP EABI specifies that structures that are up to 8 bytes in size should be in registers
R3
and
R4
, while larger structures are returned by accessing a hidden argument in
R3
. GCC compilers always uses the hidden argument method regardless of size.
The CodeWarrior linker checks to see if you are including objects in your project that have incompatible EABI settings. If you do, a warning message is issued.
NOTE
Different versions of GCC compilers may fix these incompatibilities, so you should check your version if you will be mixing GCC and CodeWarrior objects.
36.1.2 incompatible_sfpe_double_params
Warns when skipping registers for double
data types.
Syntax
#pragma incompatible_sfpe_double_params on | off | reset
Remarks
This pragma makes CodeWarrior-built object files more compatible with those created with a GCC compiler.
The AIOP EABI states that software floating-point parameters of type double
always begin on an odd register. In other words, in the function void red(long a, double b) a
is passed in register
R3
and b
is passed in registers
R5
and
R6
(effectively skipping
R4
).
GCC compilers do not skip registers if objects of type double
are passed (although it does skip them for values of type long long
).
The CodeWarrior linker checks to see if you are including objects in your project that have incompatible EABI settings. If you do, a warning message is issued.
NOTE
Different versions of GCC compilers may fix these incompatibilities, so you should check your version if you will be mixing GCC and CodeWarrior object code.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
542 Freescale Semiconductor, Inc.
36.2 Debugging Information Pragmas
This section describes the debugging information pragmas.
•
•
Chapter 36 Pragmas for AIOP Compiler
36.2.1 dwarf2typedefchains
Generates DWARF2 debugging information typedef
statements.
Syntax
#pragma dwarf2typedefchains on | off | reset
Remarks
When this pragma is on
, the compiler generates DWARF2 debugging information for type definitions.
By default, this option is off
.
36.2.2 dwarf2lexblockcodeaddrcalc
Calculates the first and the last instruction in a lexical block by inspecting the entire lexical block start and end lines.
Syntax
#pragma dwarf2lexblockcodeaddrcalc on | off | reset
Remarks
When this pragma is off, the compiler looks for the instruction following the lexical block start and end but, does not emit the correct address range for unparenthesiszed loop expressions.
By default, this option is off.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
543
Library and Linking Pragmas
36.3 Library and Linking Pragmas
This section describes the library and linking pragmas.
•
•
36.3.1 force_active
Deprecated.
Syntax
#pragma force_active on | off | reset
Remarks
In source code, use
__declspec(force_export)
,
__attrbute__((force_export))
, or
__attrbute__((used))
.
In a linker command file, use the
FORCEACTIVE
command.
36.3.2 section
This sophisticated and powerful pragma lets you arrange compiled object code into predefined sections and sections you define.
NOTE
Deprecated only when used without an associated
__declspec(section)
. To avoid C++ parsing ambiguities and other possible inadvertent errors, use
__declspec(section)
instead.
Syntax
#pragma section [ objecttype | permission ][iname][uname]
[data_mode=datamode][code_mode=codemode]
Parameter
objecttype specifies where types of object data are stored. It may be one or more of these values:
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
544 Freescale Semiconductor, Inc.
Chapter 36 Pragmas for AIOP Compiler
• code_type - executable object code
• data_type - non-constant data of a size greater than the size specified in the small data threshold option in the Target settings panel
• sdata_type - non-constant data of a size less than or equal to the size specified in the small data threshold option in the Target settings panel
• const_type - constant data of a size greater than the size specified in the small const data threshold option in the Target settings panel
• sconst_type - constant data of a size less than or equal to the size specified in the small const data threshold option in the Target settings panel
• all_types - all code and data
Specify one or more of these object types without quotes separated by spaces.
The CodeWarrior C/C++ compiler generates some of its own data, such as exception and static initializer objects, which are not affected by #pragma section.
NOTE
To classify character strings, the CodeWarrior C/C++ compiler uses the setting of the Make Strings Read Only checkbox in the
Processor settings panel. If the checkbox is checked, character strings are stored in the same section as data of type const_type.
If the checkbox is clear, strings are stored in the same section as data for data_type.
permission specifies access permission. It may be one or more of these values:
• R - read only permission
• W - write permission
• X - execute permission
Specify one or more of these object types without quotes separated by spaces.
For more information on access permission, see "Section access permissions" on page
iname specifies the name of the section where the compiler stores initialized objects. Variables that are initialized at the time they are defined, functions, and character strings are examples of initialized objects.
The iname parameter may be of the form .abs.xxxxxxxx where xxxxxxxx is an 8-digit hexadecimal number specifying the address of the section.
uname
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
545
Library and Linking Pragmas
specifies the name of the section where the compiler stores uninitialized objects. This parameter is required for sections that have data objects. The uname parameter value may be a unique name or it may be the name of any previous iname or uname section. If the uname section is also an iname section then uninitialized data is stored in the same section as initialized objects.
The special uname COMM specifies that uninitialized data will be stored in the common section. The linker will put all common section data into the ".bss" section. When the Use
Common Section checkbox is checked in the Processor panel, COMM is the default uname for the .data section. If the Use Common Section checkbox is clear, .bss is the default name of .data section.
The uname parameter value may be changed. For example, you may want most uninitialized data to go into the .bss section while specific variables be stored in the
COMM section.
Figure 36-1 shows an example where specific uninitialized variables are stored in the
COMM section.
Figure 36-1. Storing Uninitialized Data in the COMM Section
#pragma push // save the current state
#pragma section ".data" "COMM" int red; int sky;
#pragma pop // restore the previous state
NOTE
You may not use any of the object types, data modes, or code modes as the names of sections. Also, you may not use predefined section names in the EABI for your own section names.
data_mode=datamode specifies the compiler for the kind of addressing mode to be used for referring to data objects for a section.
The permissible addressing modes for datamode are:
• near_abs - objects must be within the range -65,536 bytes to 65,536 bytes (16 bits on each side)
• far_abs - objects must be within the first 32 bits of RAM
• sda_rel - objects must be within a 32K range of the linker-defined small data base address
The sda_rel addressing mode may only be used with the ".sdata", ".sbss", ".sdata2",
".sbss2", ".PPC.EMB.sdata0", and ".PPC.EMB.sbss0" sections.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
546 Freescale Semiconductor, Inc.
Chapter 36 Pragmas for AIOP Compiler
The default addressing mode for large data sections is far_abs. The default addressing mode for the predefined small data sections is sda_rel.
Specify one of these addressing modes without quotes.
code_mode=codemode specifies the compiler for the kind of addressing mode to be used for referring to executable routines of a section.
The permissible addressing modes for codemode are:
• pc_rel - routines must be within plus or minus 24 bits of where pc_rel is called from
• near_abs - routines must be within the first 24 bits of memory address space
• far_abs - routines must be within the first 32 bits of memory address space
The default addressing mode for executable code sections is pc_rel.
Specify one of these addressing modes without quotes.
NOTE
All sections have a data addressing mode
(data_mode=datamode) and a code addressing mode
(code_mode=codemode). Although the CodeWarrior C/C++ compiler for embedded allows you to store executable code in data sections and data in executable code sections, this practice is not encouraged.
Remarks
CodeWarrior compilers generate their own data, such as exception and static initializer objects, which the
#pragma section
statement does not affect.
Section access permissions
When you define a section by using #pragma section, its default access permission is read only. Changing the definition of the section by associating an object type with it sets the appropriate access permissions for you. The compiler adjusts the access permission to allow the storage of newly-associated object types while continuing to allow objects of previously-allowed object types. For example, associating code_type with a section adds execute permission to that section. Associating data_type, sdata_type, or sconst_type with a section adds write permission to that section.
Occasionally you might create a section without associating it with an object type. You might do so to force an object into a section with the __declspec keyword. In this case, the compiler automatically updates the access permission for that section to allow the object to be stored in the section, then issue a warning. To avoid such a warning, make
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
547
Library and Linking Pragmas
sure to give the section the proper access permissions before storing object code or data into it. As with associating an object type to a section, passing a specific permission adds to the permissions that a section already has.
Predefined sections and default sections
When an object type is associated with the predefined sections, the sections are set as default sections for that object type. After assigning an object type to a non-standard section, you may revert to the default section with one of the forms in
#pragma section" on page 488 .
The compiler predefines the sections in
Figure 36-2. Predefined sections
#pragma section code_type ".text" data_mode=far_abs code_mode=pc_rel
#pragma section data_type ".data" ".bss" data_mode=far_abs code_mode=pc_rel
#pragma section const_type ".rodata" ".rodata" data_mode=far_abs code_mode=pc_rel
#pragma section sdata_type ".sdata" ".sbss" data_mode=sda_rel code_mode=pc_rel
#pragma section sconst_type ".sdata2" ".sbss2" data_mode=sda_rel code_mode=pc_rel
#pragma section ".PPC.EMB.sdata0" ".PPC.EMB.sbss0" data_mode=sda_rel code_mode=pc_rel
#pragma section RX ".init" ".init" data_mode=far_abs code_mode=pc_rel
NOTE
The .PPC.EMB.sdata0 and .PPC.EMB.sbss0 sections are predefined as an alternative to the sdata_type object type.
The .init section is also predefined, but it is not a default section. The .init section is used for startup code.
Forms for #pragma section
#pragma section ".name1"
This form simply creates a section called .name1 if it does not already exist. With this form, the compiler does not store objects in the section without an appropriate, subsequent #pragma section statement or an item defined with the __declspec keyword.
If only one section name is specified, it is considered the name of the initialized object section, iname. If the section is already declared, you may also optionally specify the uninitialized object section, uname. If you know that the section must have read and write permission, use #pragma section RW .name1 instead, especially if you use the
__declspec keyword.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
548 Freescale Semiconductor, Inc.
Chapter 36 Pragmas for AIOP Compiler
#pragma section objecttype ".name2"
With the addition of one or more object types, the compiler stores objects of the types specified in the section .name2. If .name2 does not exist, the compiler creates it with the appropriate access permissions. If only one section name is specified, it is considered the name of the initialized object section, iname. If the section is already declared, you may also optionally specify the uninitialized object section, uname.
#pragma section objecttype
When there is no iname parameter, the compiler resets the section for the object types specified to the default section. Resetting the section for an object type does not reset its addressing modes. You must reset them.
When declaring or setting sections, you also can add an uninitialized section to a section that did not have one originally by specifying a uname parameter. The corresponding uninitialized section of an initialized section may be the same.
Forcing individual objects into specificsections
You may store a specific object of an object type into a section other than the current section for that type without changing the current section. Use the __declspec keyword with the name of the target section and put it next to the extern declaration or static definition of the item you want to store in the section.
Figure 36-3. Using __declspec to Force Objects into Specific Sections
__declspec(section ".data") extern int myVar;
#pragma section "constants"
__declspec(section "constants") const int myConst = 0x12345678
Using #pragma section with #pragma push and #pragma pop
You can use this pragma with #pragma push and #pragma pop to ease complex or frequent changes to sections settings.
See
for an example.
NOTE
The #pragma pop does not restore any changes to the access permissions of sections that exist before or after the corresponding #pragma push.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
549
Code generation Pragmas
36.4 Code generation Pragmas
This section describes the code generation pragmas.
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
•
36.4.1 alignment_metrowerks
Determines which structure alignment policy to use.
Syntax
#pragma alignment_metrowerks on | off | reset
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
550 Freescale Semiconductor, Inc.
Chapter 36 Pragmas for AIOP Compiler
This pragmas specifies which policy that
#pragma options align=power
will use for aligning data in structures. When this pragma is on, then the directive
#pragma options align=power
is equivalent to this directive
#pragma options align=power_mw
.
When this pragma is off, then the directive
#pragma options align=power
is equivalent to this directive
#pragma options align=power_gcc
.
36.4.2 disable_registers
Controls compatibility for the ISO/IEC standard library function setjmp()
.
Syntax
#pragma disable_registers on | off | reset
Remarks
If this pragma is on
, the compiler disables certain optimizations for any function that calls setjmp()
. It also disables global optimization and does not store local variables and arguments in registers. These changes ensure that all local variables have correct values when the setjmp()
function saves the processor state.
Use this pragma only if you are porting code that relies on this feature because it makes your program much larger and slower. In new code, declare a variable to be volatile
if you expect its value to persist across setjmp() calls.
36.4.3 function_align
Aligns the executable object code of functions on a specified byte boundary.
Syntax
#pragma function_align 4 | 8 | 16 | 32 | 64 | 128 | reset
36.4.4 gen_isel
Controls the use of the integer select instruction, isel
.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
551
Code generation Pragmas
Syntax
#pragma gen_isel on | off | number | always where number is a value from 1 to 255.
Remarks
The compiler uses this pragma to determine how often it generates the isel
instruction.
The number argument specifies how aggressively the compiler should use this instruction, 1 is equivalent to "rarely" and 255 is equivalent to always
. The on
choice is equivalent to a value of 10.
36.4.5 gprfloatcopy
Takes advantage of simpler alignment restrictions for copying floating point data.
Syntax
#pragma gprfloatcopy on | off | reset
Remarks
When this pragma is on
, the compiler uses integer load and store instructions for memoryto-memory assignments for objects of type double
and float
, which improves the speed of memory-to-memory assignments of unaligned floating-point data. When this pragma is off
, the compiler issues floating-point load and store instructions instead.
36.4.6 has8bytebitfields
Controls the use of bitfields that fit in the long long
data type.
Syntax
#pragma has8bytebitfields on | off | reset
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
552 Freescale Semiconductor, Inc.
Chapter 36 Pragmas for AIOP Compiler
When this pragma is on
, the compiler allows bitfields in the long long
data type. Such bitfields may occupy up to 64 bits (8 bytes). When this pragma is off
, the compiler allows bitfields only in integer types of the same size or smaller than the long
type.
Listing 36-1. Example for pragma has8bytebitfields
#pragma has8bytebitfields on struct X {
long long fielda : 12;
long long fieldb : 18;
long long fieldc : 32;
long long fieldd : 2;
};
36.4.7 interrupt
Deprecated. To avoid C++ parsing ambiguities and other possible inadvertent errors, use
__declspec(interrupt)
instead.
36.4.8 legacy_struct_alignment
Avoids the possibility of misaligned load or store instructions caused by promoting the alignment of global and local data objects to a minimum of 4 bytes.
Syntax
#pragma legacy_struct_alignment on | off | reset
Remarks
The default for this pragma is off
for AIOP as the big endian systems do not crash and misalignment is rare.
36.4.9 merge_float_consts
Each floating point constant is placed in a unique variable such that the linker will merge floating point constants which have the same value. (The variable names are not legal
C/C++ and are not accessible by the user). This option works with either small data in
TOC on or off. This option minimizes TOC entry usage for programs which frequently use the same floating point constant in many different source files.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
553
Code generation Pragmas
Syntax
#pragma merge_float_consts on|off
Remarks
The default for this pragma is off
for Freescale processors.
36.4.10 min_struct_align
Increases aggregate alignments for better memory access.
Syntax
#pragma min_struct_align 4 | 8 | 16 | 32 | 64 | 128 | on | off
| reset
Remarks
When this pragma is off
, the compiler aligns objects to their regular alignments. The default alignment is 4.
NOTE
This pragma only applies if the optimization level is greater than 0.
36.4.11 misaligned_mem_access
Controls how the compiler copies structures that are not aligned to 4-byte boundaries.
Syntax
#pragma misaligned_mem_access on | off | reset
Remarks
When this pragma is on
, the compiler uses 4-byte load and store instructions to copy structures that are not aligned to 4-byte boundaries. By using these misaligned load and store instructions, the compiler improves runtime performance and reduces code size.
When this pragma is off
, the compiler uses 1-, 2-, and 4-byte load and store instructions to copy structures that are aligned on corresponding boundaries.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
554 Freescale Semiconductor, Inc.
Chapter 36 Pragmas for AIOP Compiler
However, misaligned load and store instructions on some processors give poor performance or even generate processor exceptions. For these processors, turn this pragma off. Desktop variants of the processor family do not have this limitation.
Consult the processor manufacturer's documentation for information on the processor's behavior when loading and storing 4-byte values that are not aligned to 4-byte boundaries.
The default for this pragma is on
for processors that allow misaligned memory access.
The default is off
for processors that have limited misaligned memory access performance or generate an exception.
36.4.12 no_register_save_helpers
Controls the save and restore registers without calling helper functions
Syntax
#pragma no_register_save_helpers on | off | reset
36.4.13 options
Specifies how to align structure and class data.
Syntax
#pragma options align= alignment
Parameter
alignment
Specifies the boundary on which structure and class data is aligned in memory. Values for alignment range from 1 to 16, or use one of the following preset values: mac68k mac68k4byte
Table 36-1. Structs and Classes Alignment
If alignment is ...
The compiler ...
Aligns every field on a 2-byte boundaries, unless a field is only 1 byte long. This is the standard alignment for 68K Mac
OS alignment.
Aligns every field on 4-byte boundaries.
Table continues on the next page...
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
555
Code generation Pragmas
Table 36-1. Structs and Classes Alignment (continued)
power native packed reset
If alignment is ...
The compiler ...
Aligns every field on its natural boundary. For example, it aligns a character on a 1-byte boundary and a 16-bit integer on a 2-byte boundary. The compiler applies this alignment recursively to structured data and arrays containing structured data. So, for example, it aligns an array of structured types containing an 4-byte floating point member on an 4-byte boundary.
Aligns every field using the standard alignment.
Aligns every field on a 1-byte boundary. It is not available in any panel. This alignment causes your code to crash or run slowly on many platforms. Use it with caution.
Resets to the value in the previous #pragmaoptionsalign statement.
NOTE
There is a space between options
and align
.
36.4.14 pool_data
Controls whether data larger than the small data threshold is grouped into a single data structure.
Syntax
#pragma pool_data on | off | reset
Remarks
When this pragma is on
the compiler optimizes pooled data. You must use this pragma before the function to which you apply it.
NOTE
Even if this pragma is on
, the compiler will only pool the data if there is a performance improvement.
This pragma corresponds to the CodeWarrior IDE's Pool Data setting in the
ProcessorProperties > C/C++ Build > Settings > Tool Settings > Compiler >
Processor panel.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
556 Freescale Semiconductor, Inc.
Chapter 36 Pragmas for AIOP Compiler
36.4.15 ppc_lvxl_stvxl_errata
Syntax
#pragma ppc_lvxl_stvxl_errata on | off | reset
36.4.16 profile
Controls the generation of extra object code for use with the CodeWarrior profiler.
Syntax
#pragma profile on | off | reset
Remarks
If you enable this pragma, the compiler generates code for each function that lets the
CodeWarrior Profiler collect information on it.
This pragma corresponds to the CodeWarrior IDE's Profiler Information setting in the
ProcessorProperties > C/C++ Build > Settings > Tool Settings > Compiler >
Processor panel.
36.4.17 read_only_switch_tables
Controls where tables for switch
statements are placed in object code.
Syntax
#pragma read_only_switch_tables on | off | reset
Remarks
This option specifies where the compiler places executable code addresses for switch statements. When this option is on
, the compiler places these tables in a read-only section
(
.rodata
), allowing the linker to place this object code in a ROM area instead of RAM.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
557
Code generation Pragmas
When this option is off
, the compiler places these switch
tables in an object code section that is readable and writable (
.data
). Putting these tables in a read/write section allows relocation at runtime. The System V ABI, SuSE, YellowDog, and SDA PIC/PID application binary interfaces (ABIs) allow relocatable object code at runtime.
36.4.18 strict_ieee_fp
Controls generation of executable code that conforms to the IEEE floating point standard.
Syntax
#pragma strict_ieee_fp on | off | reset
Remarks
Disabling this option may improve performance but may change the results generated.
• Use Fused Mult-Add/Sub
Uses a single instruction to do a multiply accumulate. This runs faster and generates slightly more accurate results than specified by IEEE, as it has an extra rounding bit between the multiply and the add/subtract).
• Generate fsel
instruction
The fsel
instruction is not accurate for denormalized numbers, and may have issues related to unordered compares, but generally runs faster.
• Assume Ordered Compares
Ignore the unordered issues when comparing floating point which allows converting:
if (a <= b) into
if (!(a > b))
36.4.19 read_only_vtable_RTTI
Controls placing vtable and RTTI in a ready-only section.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
558 Freescale Semiconductor, Inc.
Chapter 36 Pragmas for AIOP Compiler
Syntax
#pragma read_only_vtable_RTTI on | off | reset
Remarks
If you enable this pragma, compiler places vtable
and
RTTI
in a read-only section
(
.rodata/.sdata2
), allowing the linker to place this object code in a ROM area instead of
RAM. If you disable this pragma, the compiler places vtable
and
RTTI
in an object code section that is readable and writable (
.data/.sdata
).
This pragma does not correspond to any panel setting. By default, this pragma is disabled.
36.4.20 switch_tables
Controls the generation of switch tables.
Syntax
#pragma switch_tables on | off | reset
Remarks
When on
, the compiler translates switch
statements into tables of addresses where each address in the list corresponds to a case
statement. Using tables improves the performance of switch
statements but may increase the size of the executable code if there are many case
statements or if the case
label values are not contiguous.
When off
, the compiler translates switch
statements into a series of comparisons, one comparison for each case
statement.
36.4.21 uchar_bool
Controls the size of the
_Bool
and bool
data types.
Syntax
#pragma uchar_bool on | off | reset
Remarks
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
559
Optimization Pragmas
When on
, the compiler translates the
_Bool
data type in C99 (ISO/IEC 9899-1999) source code and the bool
data type in C++ source code to 8 bit values. When off
, these data types are 32 bits. Use this pragma only before any declarations.
When this pragma is off (boolean values are 32-bits), use bitfields to ensure that a
boolean value is 8 bits. Listing 36-2 on page 560
shows an example.
Listing 36-2. Example of overriding uchar_bool in a structure in C++
#pragma uchar_bool off /* Boolean values are 32 bits */ typedef struct
{
bool sockclean:8 /* This value will only occupy 8 bits. */
} sockrec;
36.4.22 use_lmw_stmw
Controls the use of lmw
and stmw
instructions.
Syntax
#pragma use_lmw_stmw on | off | reset
Remarks
Use of lmw
and stmw
may be slower on some processors.
36.4.23 ushort_wchar_t
Controls the size of wchar_t
.
Syntax
#pragma ushort_wchar_t on | off | reset
Remarks
When this pragma is on
, wchar_t
changes from 4-bytes to 2-bytes.
36.5 Optimization Pragmas
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
560 Freescale Semiconductor, Inc.
Chapter 36 Pragmas for AIOP Compiler
This section describes the optimization pragmas.
•
•
•
•
•
•
peephole_enable_16bit_load_store_inst
•
•
•
•
•
•
•
•
•
•
•
•
36.5.1 aggressive_hoisting
Improves the number of variables that get hoisted out of a loop.
Syntax
#pragma aggressive_hoisting on | off | reset
Remarks
This pragma produces faster code and causes a slight increase in code size, especially when optimizing for size. In some cases, hoisting variables out of a loop when the loop does not have a lot of iterations can make your code slower.
The default value for this pragma is off
.
36.5.2 c9x_alias_by_type
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
561
Optimization Pragmas
Allows back-end optimizations to use alias type information.
Syntax
#pragma c9x_alias_by_type on | off | reset
Remarks
When this pragma is on, the compiler's back-end optimizations take advantage of type information gathered during alias analysis. Turn this pragma on if your source code follows the type rules specified by the ISO/IEC 9899-1999 C standard ("C99"), section
6.5.
Turn this pragma off if your source code violates type rules. The information collected from source code that violates these rules might lead the compiler to apply its optimizations incorrectly.
This pragma does not have a corresponding IDE panel setting.
36.5.3 epilogue_helper_functions
Controls size optimization for function termination instructions.
Syntax
#pragma epilogue_helper_functions on | off | reset
Remarks
When this pragma is on, the compiler reduces the size of object code in function terminations. It performs this optimization by replacing several instructions for function termination with fewer calls to special functions that perform the same tasks. This optimization reduces executable code size but also slows the program's performance.
When this pragma is on, the compiler generates instructions that may appear as inconsistent information in a symbolic debugger when stepping through the end of a function.
This pragma does not correspond to any panel setting. By default, this pragma is off. The compiler turns this optimization on implicitly when size optimization is on and optimization level is equal to or greater than 2. For example, these pragma settings will also turn on epilogue helper function generation:
#pragma optimization_level 2 #pragma optimize_for_size on
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
562 Freescale Semiconductor, Inc.
Chapter 36 Pragmas for AIOP Compiler
36.5.4 ipa_rescopes_globals
Rescopes the application global variables, that are only used in one function, to local static. The change to static enables other optimizations that improve alias analysis and load/store optimizations.
Syntax
#pragma ipa_rescopes_globals on | off
Remarks
Ensure that the following requirements are met to rescope the application global variables to local static:
• Program IPA is enabled in all application source files
• use of
#pragma ipa_rescopes_globals on
in all application source files (prefix file or with
-flag ipa_rescopes_globals
on the commandline)
• main()
is defined in one of the application files.
• It is not necessary, or even desirable, to have standard library, runtime or startup code compiled with program IPA and ipa_rescopes_globals on
. However, it is important to have as many of your application sources as possible compiled with those options enabled.
NOTE
As the third party libraries generally do not access the application variables, these libraries can be kept in archive form.
For a simple example, compile/assemble your startup code without program IPA.
Compile all of the application code with program IPA,
#pragma ipa_rescopes_globals on
and link the startup objects, your application objects and the library archives (For more details on Program IPA linking procedures, refer "Interprocedural Analysis").
For a complex example where the application sources are put into groups, compiled and then pre-built into several archives or partially linked objects and the build procedure cannot be matched with the simple example, following changes to the build procedure are suggested:
• Try to make the build setup as similar to the simple example as possible. This will help you identify if the code will benefit from ipa_rescopes_globals
or you will need to
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
563
Optimization Pragmas
modify your source files to get a successful link. For more details, refer
"Generating a successful link"
.
• All of your functions are not visible to the compiler at once during program IPA. It is possible that a defined global variable in your core files may be used by only one core file but might also be used in one of your application archives that you were unable to build the simple way. If this is true, ipa_rescopes_globals
will rescope the variable and at link time, your application archive will not be able to find the variable and you will get an undefined symbol link error.
NOTE
If you get a successful link you do not have to make any further changes to the build or source.
Generating a successful link
Optimization prevents an improper build. If you do not get a successful link or you only get a few such link errors, identify the source file that defines the "undefined" symbol and try one of the following (in decreasing order of general preference):
• Move the definition of the symbol into the application archive. Symbols that are undefined do not get rescoped.
• Force the export of the symbol with
__declspec
(force_export). Symbols that are exported do not get rescoped.
• Change the symbols to weak with
__declspec
(weak) by inserting before definition.
Weak symbols do not get rescoped.
• Change the symbols to volatile. Volatile symbols do not get rescoped.
36.5.5 peephole
Controls the use peephole optimization.
Syntax
#pragma peephole on | off | reset
Remarks
If you enable this pragma, the compiler performs peephole optimizations . These optimizations replace redundant or commonly occurring groups of instructions with smaller, more efficient groups of instructions.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
564 Freescale Semiconductor, Inc.
Chapter 36 Pragmas for AIOP Compiler
36.5.6 peephole_enable_16bit_load_store_inst
Enables use of 16-bit load/store instructions instead of 32-bit load/store instructions.
Syntax
#pragma peephole_enable_16bit_load_store_inst on | off |
reset
Remarks
This peephole optimization will replace the 32 bit load/store instructions with 16 bit load/ store instructions.
This pragma is on by default under size optimization and off under speed optimization.
Please note that this optimization is applicable only when VLE instruction set is enabled.
Example
Converts the below pattern
e_stb r0,28(r3)
e_stb r0,32(r3)
e_stb r4,36(r3)
e_stb r4,40(r3)
e_stb r4,41(r3)
e_stb r4,42(r3)
e_stb r4,43(r3) to
e_add16i r3,r6,28
se_stb r0,0(r3)
se_stb r0,4(r3)
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
565
Optimization Pragmas
se_stb r4,8(r3)
se_stb r4,12(r3)
se_stb r4,13(r3)
se_stb r4,14(r3)
se_stb r4,15(r3)
36.5.7 ppc_opt_defuse_mem_limit
Controls memory consumed by compiler optimizations on the host computer.
Syntax
#pragma ppc_opt_defuse_mem_limit on | off | reset | limit
Parameter
limit
Number of megabytes to use on the host computer when optimizing object code. The default value is
150
, which specifies 150 megabytes.
Remarks
Some optimizations need a lot of memory on the host computer, especially when optimizing large functions that make many function calls or refer to many variables. This pragma controls how much memory these optimizations consume.
If limit is set too low, the compiler will not be able to complete some optimizations and will issue an error message.
NOTE
This pragma is to be used when users see the compiler error or warning that the compiler needs more memory to be allocated for usedef/defuse chain computation.
36.5.8 ppc_unroll_instructions_limit
Limits number of instructions in an unrolled loop to value.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
566 Freescale Semiconductor, Inc.
Chapter 36 Pragmas for AIOP Compiler
Syntax
#pragma ppc_unroll_instructions_limit value | on | off
Parameter
value
Count limit of instructions. The default is 70.
Remarks
Use this pragma to specify the maximum number of instructions to place in an unrolled loop. The opt_unroll_loops
pragma controls loop unrolling optimization.
When this pragma is on
, the compiler uses the default value.
36.5.9 ppc_unroll_speculative
Controls speculative unrolling of counting loops which do not have fixed counts.
Syntax
#pragma ppc_unroll_speculative on | off
Remarks
The compiler uses the value specified with the ppc_unroll_factor_limit
pragma to compute how many times to unroll eligible loops. The compiler adjusts the value specified with ppc_unroll_factor_limit
so that it is equal to or less than the closest power of 2.
This optimization is only applied when:
• loop unrolling is turned on with the opt_unroll_loops
pragma
• the loop iterator is a 32-bit value ( int
, long
, unsigned int
, unsigned long
)
• the loop's body has no conditional statements
If you enable this pragma, the loop unrolling factor is a power of 2, less than or equal to the value specified by the ppc_unroll_factor_limit
pragma.
The opt_unroll_loops
pragma controls all loop unrolling optimization. To check this setting, use
__option (ppc_unroll_speculative)
. By default, this pragma is on
when loop unrolling is enabled.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
567
Optimization Pragmas
36.5.10 processor
Specifies the scheduling model used for instruction scheduling optimization.
Syntax
#pragma processor model model
This argument is one of these choices:
401 | 403 | 405 | 505 | 509 | 5100 | 5200 | 555 | 56x | 601 | 602 | 603 | 603e | 604 | 604e
| 74x | 75x | 801 | 821 | 823 | 85x | 86x | 87x | 88x | 7400 | 744x | 745x | 82xx| 85xx | e300v1 | e500v1 | e500v2 | e600 | Zen | generic
36.5.11 prologue_helper_functions
Controls size optimization for function initialization instructions.
Syntax
#pragma prologue_helper_functions on | off | reset
Remarks
When this pragma is on, the compiler reduces the size of object code in function initialization. It performs this optimization by replacing several instructions for function initialization with fewer calls to special functions that perform the same tasks. This optimization reduces executable code size but also reduces the program's performance.
This pragma does not correspond to any panel setting. By default, this pragma is off. The compiler turns this optimization on implicitly when size optimization is on and optimization level is equal to or greater than 2. For example, these pragma settings will also turn on prologue helper function generation:
#pragma optimization_level 2 #pragma optimize_for_size on
36.5.12 remove_frsp_aggressive
Improves code optimization by transforming LFS and FRSP instructions into the equivalent FMR instructions.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
568 Freescale Semiconductor, Inc.
Chapter 36 Pragmas for AIOP Compiler
Syntax
#pragma remove_frsp_aggressive on | off | reset
Remarks
This pragma allows the compiler more opportunity to use copy propogation optimizations to improve the generated code.
The default value for this pragma is off
.
36.5.13 schedule
Specifies the use of instruction scheduling optimization.
Syntax
#pragma schedule once | twice | altivec | off
Remarks
This pragma lets you choose how many times the compiler passes object code through its instruction scheduler.
On highly optimized C code where loops were manually unrolled, running the scheduler once seems to give better results than running it twice, especially in functions that use the register
specifier.
When the scheduler is run twice, it is run both before and after register colorizing. If it is only run once, it is only run after register colorizing.
This pragma does not correspond to any panel setting. The default value for this pragma is twice
.
36.5.14 schedule z760
Controls the rearrangement of instructions to reduce the effects of instruction latency.
Syntax
#pragma schedule z760
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
569
Optimization Pragmas
Remarks
This pragma can be activated by:
1. selecting the AIOP (from the Processor preference panel or by using the
-proc AIOP command line compiler option)
2. specifying
#pragma schedule z760
in the prefix file or on the Preprocessor preference panel. You can also pass
-pragma "schedule z760"
command line to the compiler, to activate the pragma.
The pragma does not correspond to any panel setting. By default, this pragma is off
.
36.5.15 scheduling
Specifies the scheduling model used for instruction scheduling optimization.
Syntax
#pragma scheduling
model| off model
This argument is one of these choices:
401 | 403 | 405 | 505 | 509 | 5100 | 5200 | 555 | 56x | 601 | 602 | 603 | 603e | 604 | 604e
| 74x | 75x | 801 | 821 | 823 | 85x | 86x | 87x | 88x | 7400 | 744x | 745x | 82xx| 85xx | e300v1 | e500v1 | e500v2 | e600 | Zen | generic
36.5.16 switch_op
Minimizes comparisions for "if" and "switch" statements.
Syntax
#pragma switch_op on | off
Remarks
When this pragma is on, the comparisions for "if" and "switch" statements are minimized.
The compiler performs this optimization by using the constants specified within the switch case (or if) statements and based on internal heuristics that determine the cost.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
570 Freescale Semiconductor, Inc.
Chapter 36 Pragmas for AIOP Compiler
By default, this pragma is off. The compiler turns this optimization on implicitly when the optimization level is equal to or greater than 2 and the user specifies #pragma switch_op on. For example, these pragma settings will also turn on switch optimization:
#pragma optimization_level 2
#pragma switch_op on
36.5.17 switch_cases_array
Optimizes a certain class of the switch statements so that a memory access to load the address, of the selected case in a switch statement, is eliminated.
Syntax
#pragma switch_cases_array on | off | reset
Remarks
Normally, a selected case in a switch statement is chosen by loading the address of the selected case from a table of addresses. In some cases it is desirable to select a case without loading the case address from a memory table of addresses.
This optimization takes a switch statement and creates a contiguous array of code representing each possible selectable case. The code for each case is identical in size.
This implies that where no applicable case is specified, there are dummy cases consisting of NOP code. Dummy cases eventually branch to the default case. It also implies that applicable cases may be padded with NOPs to achieve uniform size among the cases. The effect of a contiguous array of case values is that the selected case is chosen by an index calculated by the multiplication of the case value and the uniform case size. No memory loads are required to select the proper case.
This pragma should be used with care since the possibility of huge code growth is possible (since dummy cases are inserted to fill gaps between the user specified cases).
Ideally, this optimization could be used on switch statements that have case values closely related such that as few as possible dummy cases are added.
The pragma’s scope is a function, so all switch statements within a function are affected by this pragma.
36.5.18 volatileasm
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
Freescale Semiconductor, Inc.
571
Optimization Pragmas
Controls the optimization of inline assembly statements.
Syntax
#pragma volatileasm on | off | reset
Remarks
When this pragma is off
, the compiler applies peephole and scheduling optimizations to inline assembly statements. If the pragma is on
, the compiler does not optimize these statements.
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual, Rev.
10.2, 01/2016
572 Freescale Semiconductor, Inc.
Index
Index
__addb_
__addbu
__addbu_
__addh_
__addhu
__addhu_
__ASSEMBLER__
__attribute__((deprecated))
__attribute__((never_inline))
__attribute__((noalias))
__attribute__((returns_twice))
__attribute__((unused))
__COMPILER__
__cplusplus
__declspec(interrupt)
__declspec(never_inline)
__declspec(weak)
__dsn
__e_hwaccel
__e_hwacceli
__e_ordhwaccel
__e_ordhwacceli
__e_rlwimi
__e_rlwinm
__FILE__
__FUNCTION__
__ide_target()
__lbdx
__lbdx_
__ldwar
__ldwcb
__lhbr
__lhdx
__lhdx_
__llldbrw
__llldw
__llldwbrw
__llstdw
__llstdwc
__lwbr
__lwdx
__lwdx_
__MWERKS__
__PPCBROADWAY__
__PPCGECKO__
__PRETTY_FUNCTION__ Identifier
__stbdcbx
__stdbrw
__STDC__
__stdwwt
__sthbr
__sthdcbx
__stwbr
__stwdcbx
__subfb
__subfb_
__subfbu
__subfh
__subfh_
__subfhu
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual
Freescale Semiconductor, Inc.
573
Index
__VERSION__
. (location counter)
.ifge if greater than or equal
.ifgt if greater than
.ifle if less than or equal
.ifne if not equal
#define
-align
-ansi
-bool
-c
-codegen
-convertpaths
-defaults
-define
-dialect
-enum
-EP
-estimate_stack_usage
-flag
-fmt
-format
-func_align
-g
-gap_fill
-gccincludes
-genbinary
-I-
-include
-inline
-nostackinfo
-nosyspath
-o
-O
-pool
-ppopt
-precompile
-preprocess
-processor
-progress
-rambuffer
-keepobjects
-L+
-library
-linkmode
-listdwarf
-little
-M
-main
-mapcr
-maxerrors
-maxwarnings
-MD
-MM
-MMD
-model
-msgstyle
574
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual
Freescale Semiconductor, Inc.
Index
-recursion_depth
-relocate
-requireprotos
-resolved_partial
-romaddr
-root_sort
-rostr
-RTTI
-sdata2threshold
-sdataaddr
-sdatathreshold
-srec
-stackaddr
-stacksize
-stderr
-stdlib
-strict
-strings
-trigraphs
-use_lmw_stmw
-verbose
-version
-vle_bl_opt
-ws_size
A
ABI Conformance
Additional Small Data Sections
ADDR
aggressive_hoisting
aggressive_inline
AGGRESSIVE_MERGE_SECTIONS
AIOP Optimizations
AIOP Specific Command-Line Options
ALIGNALL
Aligning a Variable Declaration
alignment_metrowerks
Alignment in a Structure Definition
always_import
ANSI_strict
ARM_conform
ARM_scoping
Assembler Control Directives
Assembler Directives
Assembling VLE Instructions
Assembly Language Statements
Associating Input Sections With Output Sections
B
Bitfields
bss
Buffer Manipulation
Building an SDA PIC/PID Application
byte
BYTE
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual
Freescale Semiconductor, Inc.
575
Index
Byte-Reversing Functions
C
C++-style Comments
C++-Style Comments
C++-Style Digraphs
c99
C99 Extensions
c9x
c9x_alias_by_type
Case-Sensitive Identifiers
C Compiler
Character Constants
check_header_flags
Checking Pragma Settings
Choosing Which Functions to Inline
Code Generation Command-Line Options
Code Merging
CodeWarrior Linker Command File
Command-Line
Command-Line Options for Diagnostic Messages
Command-Line Options for Language Translation
Command-Line Options for Library and Linking
Command-Line Options for Optimization
Command-Line Options for Preprocessing
Command-Line Options for Standard C++
Conformance
Command-Line Options for Standard C
Conformance
Commands, Directives, and Keywords
Comments
Common Subexpression Elimination
Common Terms
Compatibility Conditional Directives
Compiling VLE Instructions
Compound Literal Values
Conditional Preprocessor Directives
Configuring Command-Line Tools
Configuring Linker File to Several ROM Blocks
const_strings
Constants
Controlling Alignment
Controlling Standard C Conformance
cpp_extensions
Creating an LCF from Scratch
Creating Memory Gaps
Creating Statement Labels
Creating Symbols
Creating Unique Labels and Equates
D
data
Data Addressing
Data Alignment
Dead Code Elimination
Dead Store Elimination
Deadstripping
Dead-Stripping
Declaration Specifications for AIOP Code
Declaring Variables by Address
defer_codegen
defer_defarg_parsing
Defining Sections in the Output File
Defining the Target's Memory Map
Dependencies Among Source Files
Determining Which Settings Are Saved and
Diagnostic Command-Line Options
Diagnostic Pragmas
Differences Between ISO C++ and EC++
direct_destruction
direct_to_som
Directives
DO_NOT_MERGE_SECTIONS
dont_inline
576
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual
Freescale Semiconductor, Inc.
Index
dwarf2typedefchains
E
EC++ Specifications
elseif
Embedded C++
Empty Arrays in Structures
enums_signed
equ
,
equal sign (=)
errno_name
error
EWL Naming Convention
Exception Handling
export
EXPORTSTRTAB
Expressions
extended_errorcheck
Extensions to Standard C++
Extensions to the Preprocessor
EXTERNAL_SYMBOL
F
faster_pch_gen
File Name Extensions
flat_include
FORCEFILES
Forward Declarations of Static Arrays
fralloc
fullpath_prepdump
Function and Top-level Variable Declarations
Function-Level Inline Assembly
G
gcc_extensions
GCC Extensions
,
global_optimizer
GNU Compatible Syntax option
H
has8bytebitfields
Hexadecimal Floating-Point Constants
How to Rebuild the EWL Libraries
How to Relocate Code in RAM
I
if
ifc
ifdef
ifndef
Implementation-Defined Behavior
IMPORTSYMTAB
INCLUDE
incompatible_return_small_structs
incompatible_sfpe_double_params
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual
Freescale Semiconductor, Inc.
577
Index
Initializing Automatic Arrays and Structures
inline_bottom_up_once
inline_depth
inline_max_auto_size
inline_max_size
inline_max_total_size
Inline-Assembly for AIOP Build Tools
Inlining Techniques
Integer Directives
Intermediate Optimizations
internal
INTERNAL_SYMBOL
Internal and External Addressing Modes
Internal and External Segments and References
interrupt
Invalid Pragmas
Invoking Interprocedural Analysis
ipa
ipa_rescopes_globals
iso_templates
K
L
Language Related Issues
lib_export
Library and Linking Command-Line Options
Library and Linking Pragmas
line
Linker Command File Operator Precedence
Linker Command File Support
Linker-defined Symbols
Linker Generated Symbols
Linker Map File
Linking a ROM Image
Linking Binary Files
Linking VLE Object Code
Live Range Splitting
LOAD
load_store_elimination
Localization
LONG
Loop Unrolling
Low-Level Processor Synchronization
M
macro_prepdump
Macros
Manipulating the Contents of a Variable or Register
Math Functions
MAX_BIN_GAP
maxerrorcount
Memory Map
min_enum_size
min_struct_align
misaligned_mem_access
Modifications to the Section Pragma
mpwc_newline
multibyteaware_preserve_literals
N
Naming Conventions
new_mangler
578
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual
Freescale Semiconductor, Inc.
Index
no_register_save_helpers
no_static_dtors
NO_TRAILING_BSS_IN_BIN_FILES
Non-constant Static Data Initialization
Normal, Record, and Overflow Forms
O
OBJECT
old_friend_lookup
old_pods
Omitted Operands in Conditional Expressions
once
opt_classresults
opt_dead_assignments
opt_lifetimes
opt_propagation
opt_strength_reduction
opt_unroll_loops
opt_vectorize_loops
optimization_level
Optimization Command-Line Options
Optimization Pragmas
optimize_for_size
option
Option Formats
opword
P
pack
Parameter Formats
parse_mfunc_templ
peephole_enable_16bit_load_store_inst
PIC/PID Linker Command File Directives
Place Symbols in Specific Memory Location
Place the Library File in the LCF
Placing Code in ROM
pool_data
pop, push
ppc_opt_defuse_mem_limit
ppc_unroll_instructions_limit
pragma
Pragma Scope
Pragmas for C++
Pragmas for Diagnostic Messages
Pragmas for Library and Linking
Pragmas for Optimization
precompile_target
Precompiling a File in the CodeWarrior IDE
Precompiling a File on the Command Line
Predefined Macros
Predefined Macros for AIOP Compiler
Predefined Sections
Predefined Symbol __func__
Preprocessor Scope in Precompiled Files
previous
Processor Core Specific Macros
Processors With VLE Capability
Program-Level Optimization
prologue_helper_functions
R
read_only_switch_tables
read_only_vtable_RTTI
Referring to Assembly, C, and C++ Variables
Referring to Registers
Referring to the Number of Arguments
Referring to Variables in Instruction Operands
Referring to Variables in Memory
REGISTER
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual
Freescale Semiconductor, Inc.
579
Index
Relocating Code and Data in External MRAM
Relocating Code and Data in Internal RAM
Relocating Code in ROM
S
Saving and Restoring Pragma Settings
sbss
schedule
scheduling
Scope Control Directives
sdata2
section
Section Layout
SECTIONS
Setting CodeWarrior Environment Variables
Setting the Floating-Point Environment
Setting the PATH Environment Variable
SHORT
showmessagenumber
simple_prepdump
SIZEOF_ROM
Small Data Area PIC/PID Support
space_prepdump
Specifying A Single ROM Block
Specifying Constant Data Location
Specifying Inline Assembly Statements
Specifying Jump Table Location
Specifying Link Order in the IDE
Specifying Memory Area Locations and Sizes
Specifying Several ROM Blocks
Specifying Source File Locations
stack_used
Standard and Non-Standard Template Parsing
Statement-Level Inline Assembly
Strength Reduction
strict_ieee_fp
Structure Member
Supported Extensions
suppress_init_code
switch_op
switch_tables
sym
Symbols
Syntax
Syntax for Attribute Specifications
Syntax for Declaration Specifications
T
template_depth
Templates
text
text_encoding
text_vle
The __builtin_constant_p() Operator
The __builtin_expect() Operator
The typeof() Operator
Third Party Standard Libraries
Trailing Commas in Enumerations
type
Typedef Declaration
U
Unique LCF Examples
Unnamed Structures and Enumerations in C
Unsuffixed Decimal Literal Values
Unsupported Extensions
Unsupported Language Features
unused
Updating a Precompiled File Automatically
Uses for SDA PIC/PID
580
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual
Freescale Semiconductor, Inc.
Index
Using Build Tools on the Command Line
Using Command-Line
Using Comments
Using Immediate Operands
Using Local Variables and Arguments
Using Macro Arguments
Using Pragmas
Using Preprocessor
Using the .macro Directive
Using the Embedded Warrior Libraries
Using Variable Names as Memory Locations
V
Variable Argument Macros
Variable-Length Arrays
Variable Length Encoding
Void and Function Pointer Arithmetic
volatileasm
W
warn_extracomma
warn_filenamecaps_system
warn_hidevirtual
warn_illpragma
warn_illunionmembers
warn_impl_f2i_conv
warn_impl_i2f_conv
warn_largeargs
warn_no_explicit_virtual
warn_no_side_effect
warn_no_typename
warn_notinlined
warn_padding
warn_pch_portability
warn_possunwant
warn_undefmacro
warn_uninitializedvar
warn_unusedarg
warn_unusedvar
warning
warning_errors
Z
z760
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual
Freescale Semiconductor, Inc.
581
Index
582
CodeWarrior Development Studio for Advanced Packet Processing Build Tools Reference Manual
Freescale Semiconductor, Inc.
How to Reach Us:
Home Page:
freescale.com
Web Support:
freescale.com/support
Information in this document is provided solely to enable system and software implementers to use Freescale products. There are no express or implied copyright licenses granted hereunder to design or fabricate any integrated circuits based on the information in this document.
Freescale reserves the right to make changes without further notice to any products herein.
Freescale makes no warranty, representation, or guarantee regarding the suitability of its products for any particular purpose, nor does
Freescale assume any liability arising out of the application or use of any product or circuit, and specifically disclaims any and all liability, including without limitation consequential or incidental damages.
“Typical” parameters that may be provided in Freescale data sheets and/or specifications can and do vary in different applications, and actual performance may vary over time. All operating parameters, including “typicals,” must be validated for each customer application by customer's technical experts. Freescale does not convey any license under its patent rights nor the rights of others. Freescale sells products pursuant to standard terms and conditions of sale, which can be found at the following address: freescale.com/SalesTermsandConditions .
Freescale, the Freescale logo, CodeWarrior, and QorIQ are trademarks of Freescale Semiconductor, Inc., Reg. U.S. Pat. & Tm. Off. ARM is the registered trademark of ARM Limited. All other product or service names are the property of their respective owners.
© 2013–2016 Freescale Semiconductor, Inc. All rights reserved.
Document Number CWAPPBTR
Revision 10.2, 01/2016
Stiahnuť
Verejný odkaz bol aktualizovaný
Verejný odkaz na váš chat bol aktualizovaný. You may view and manage your chat links after registration here.
Verejný odkaz na dokument
Reklama
