GCC for Embedded Engineers

Read along to understand how GCC works, find out what all those other programs in the toolchain directory do, and learn some tips and tricks to become more comfortable with most indispensable tool in your project.
The spec File

Recall that GCC is a driver program that knows what program to invoke to build a certain output, which begs the question, “How does it know that?” This information that was built in to GCC when it was built is kept in the “specs”. To see the specs, run GCC with the -dumpspecs parameters:

armv5l-linux-gcc -dumpspecs

The console will fill with a few hundred lines of output. The spec file format evolved over years of development, and it's easier for the computer to read than for a person. Each line contains instructions for what parameters to use for a given tool. From the prior example, consider the command line for the assembler (with the path names removed for readability):

"<path>/as.exe" "-mcpu=xscale" "-mfloat-abi=soft" 
 ↪"-o" "<temppath>/ccm4aB3B.o" "<temppath>/ccC39DVR.s"

The compiler has the following in the specs for the assembler:

*asm:
%{mbig-endian:-EB} %{mlittle-endian:-EL} %{mcpu=*:-mcpu=%*} 
 ↪%{march=*:-march=%*} %{mapcs-*:-mapcs-%*} 
 ↪%(subtarget_asm_float_spec) 
 ↪%{mthumb-interwork:-mthumb-interwork} 
 ↪%{msoft-float:-mfloat-abi=soft} 
 ↪%{mhard-float:-mfloat-abi=hard} %{mfloat-abi=*} 
 ↪%{mfpu=*} %(subtarget_extra_asm_spec)

This line uses some familiar constructs explained below. Adequately discussing the minutiae of the spec file would require an article series in itself.

  • *asm: this line tells GCC the following line will override the internal specification for the asm tool.

  • %{mbig-endian:-EB}: the pattern %{symbol:parameter} means if a symbol was passed to GCC, replace it with parameter; otherwise, this expands to a null string. In our example, the parameter -mfloat-abi=soft was added this way.

  • %(subtarget_extra_asm_spec): evaluate the spec string %(specname). This may result in an empty string, as it did in our case.

Most users don't need to modify the spec file for their compiler; however, frequently engineers who inherit a project need to have GCC recognize nonstandard extensions for files. For example, assembler source files may have the extension, .arm; in this case, GCC won't know what to execute, as it doesn't have a rule for that file extension. In this case, you can create a spec file containing the following:

.arm:
@asm

and use the -specs=<file> to pass that to GCC, so that it will know how to handle files with the .arm extension. The spec file on the command line will be added to the internal spec file after it has been processed.

Tips and Tricks of the Trade

The following tips and tricks should be, if they haven't already, stashed on the crib sheet of engineers who work with GCC.

Force GCC to use an alternate C library:

armv5l-linux-gcc -nostdlib -nostdinc -isystem 
 ↪<path to header files> -L<path to c library> 
 ↪-l <c library file>

This tells GCC to ignore everything it knows about where to find header files and libraries and instead uses what you tell it. Most alternate C libraries provide a script that performs this function; however, some projects can't use the wrapper scripts, and other times, when experimenting with several versions of a library, the flexibility and control of specifying this information directly is necessary.

Mixed assembler/source output:

armv5l-linux-gcc -g program.c -o binary-program
armv5l-linux-objdump -S binary-program

This is the best way to see exactly what GCC generated in relation to the input code. Doing the compilation with several different optimization settings shows what the compiler did for the given optimization. Because embedded development pushes the processor-support envelope, being able to see the generated assembler code can be instrumental in proving a defect in GCC's support for that processor. In addition, engineers can use this to validate that the proper instructions are generated when specifying processor-specific optimizations.

List predefined macros:


armv5l-linux-gcc -E -dM - < /dev/null

An invaluable tool for doing a port, this makes clear what GCC macros will be set automatically and the value. This will show not only the standard macros, but also all the ones set for the target architecture. Keeping this output and comparing it to a newer version of GCC can save hours of work when code fails to compile or run due to changes.

List dependencies:

armv5l-linux-gcc -M program.c

Formally, this command creates a separate make rule for each file on the command line showing all dependencies. The output is indispensable when trying to track down problems related to what header files a source file is using and tracking down problems related to forcing GCC to use an alternate C library. Deeply nested header files are both unavoidable and incredibly useful in any nontrivial C project and can consume hours when trying to debug. Using -MM instead of -M will show only nonsystem dependencies—useful noise reduction when the problem resides in the project files alone.

Show internal steps:

armv5l-linux-gcc -### program.c

This article already uses this command to make GCC show what steps occur internally to build a program. When a program isn't compiling or linking properly, using -### is the fastest route to see what GCC is doing. Each command is on its own line and can be run individually, so:


armv5l-linux-gcc -### program.c &> compile-commands

______________________

Comments

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.

Tabela

Tabela's picture

Geçtiğimiz yerlerde her üç adımda bir gördüğümüz Tabela lar caddeleri renklendiren unsurlardır. Tabela lar caddeleri renklendirirken aynı zamandada insanların aradıkları yerleri bulmalarında son derece yardımcıdırlar. Geçen bir tatil yöresinde gezerken çok renkli, ışıklı çok güzel bir Tabela gördüm ve bayıldım.
Resmini çekeyim şu Tabela nın derken fotoğraf makinemin pili bitmiş ve çekemedim. Önümüzdeki sene gidebilirsem yine aynı şekilde bu Tabela nın resmini çekmeyi planlıyorum kimbilir belki önümüzdeki sene kısmet olur...

http://www.dikkatdizi.com

Diziizle's picture

That's what i was looking for

Medrano's picture

That's what i was looking for =) Thanks

Medrano

You have given much

Sinemalar's picture

You have given much descriptive information,Thanks...
Sinema,Vizyondakiler,Sinemalar

really

Sinemalar's picture

You have given much descriptive information,Thanks...
Sinema,Vizyondakiler,Sinemalar

thanks for document

RHM's picture

thanks for document

Thank you for the article.

Tarihte Bugun Bir Sey Oldu's picture

Thank you for the article.

thank you

emmy's picture

thanks for document
my page: www.yuhoha.com

türkçe rap müzik

rap's picture

Linux

angellord's picture

thanks for document

Bravo

Dave Comer's picture

Thanks for an excellent article on a subject affect many embedded engineers.

Webcast
How to Build an Optimal Hadoop Cluster to Store and Maintain Unlimited Amounts of Data Using Microservers

Realizing the promise of Apache® Hadoop® requires the effective deployment of compute, memory, storage and networking to achieve optimal results. With its flexibility and multitude of options, it is easy to over or under provision the server infrastructure, resulting in poor performance and high TCO. Join us for an in depth, technical discussion with industry experts from leading Hadoop and server companies who will provide insights into the key considerations for designing and deploying an optimal Hadoop cluster.

Learn More

Sponsored by AMD

White Paper
Red Hat White Paper: Using an Open Source Framework to Catch the Bad Guy

Built-in forensics, incident response, and security with Red Hat Enterprise Linux 6

Every security policy provides guidance and requirements for ensuring adequate protection of information and data, as well as high-level technical and administrative security requirements for a system in a given environment. Traditionally, providing security for a system focuses on the confidentiality of the information on it. However, protecting the data integrity and system and data availability is just as important. For example, when processing United States intelligence information, there are three attributes that require protection: confidentiality, integrity, and availability.

Learn more about catching the bad guy in this free white paper.

Learn More

Sponsored by DLT Solutions