Question about instrumenting gcc

Question about instrumenting gcc

[Renata Hodovan]

Hi All,

I have to instrument gcc for some purposes. The goal is to be able to 
track what GCC functions are called during a particularly compile. 
Unfortunately I'm not really familiar with the architecture of GCC so I 
need a little help. I tried the following steps:

1) Hacking gcc/Makefile.in and adding "-finstrument-functions" flag to 
T_CFLAGS.
2) I have an already implemented and tested version of __start_test__ 
and __end_test__ functions. They are called from gcc/main.c, before and 
after toplev_main() call. The containing file is linked to gcc (the 
object is added to OBJS-common and the dependency is defined later in 
gcc/Makefile.in)
3) Downloading prerequisites with contrib/download_prerequisites.
4) Executing the configuration from a clean build directory (on the same 
level with the source dir): ./../gcc-4.6.2/configure 
--prefix="/opt/gcc-4.6.2/" --enable-languages="c,c++"
5) Starting the build with "make all"

This way I ran out of memory, although I had 28G.
Next I tried to remove the T_CFLAGS settings from the Makefile and gave 
-finstrument-functions to the make command:
make CFLAGS+="-finstrument-functions" all
The build was successful this way but when I tried to compile something 
it resulted empty output files. (Theoretically __end_test__should have 
written its result to a given file.)

What do I make wrong? Thanks in advance!
Cheers,
Reni

Re: Question about instrumenting gcc

[Ian Lance Taylor]

Renata Hodovan <hodovan@inf.u-szeged.hu> writes:

> I have to instrument gcc for some purposes. The goal is to be able to
> track what GCC functions are called during a particularly
> compile. Unfortunately I'm not really familiar with the architecture
> of GCC so I need a little help. I tried the following steps:
>
> 1) Hacking gcc/Makefile.in and adding "-finstrument-functions" flag to
> T_CFLAGS.
> 2) I have an already implemented and tested version of __start_test__
> and __end_test__ functions. They are called from gcc/main.c, before
> and after toplev_main() call. The containing file is linked to gcc
> (the object is added to OBJS-common and the dependency is defined
> later in gcc/Makefile.in)
> 3) Downloading prerequisites with contrib/download_prerequisites.
> 4) Executing the configuration from a clean build directory (on the
> same level with the source dir): ./../gcc-4.6.2/configure
> --prefix="/opt/gcc-4.6.2/" --enable-languages="c,c++"
> 5) Starting the build with "make all"
>
> This way I ran out of memory, although I had 28G.

You should not have run out of memory at this step, not with 28G.
Adding -finstrument-functions should not have significantly changed the
memory usage of the build.

But I suppose you could have run out of memory if your
__cyg_profile_func_enter function uses a lot of memory.  If that is the
problem, then that is what you will have to fix.


> Next I tried to remove the T_CFLAGS settings from the Makefile and
> gave -finstrument-functions to the make command:
> make CFLAGS+="-finstrument-functions" all

This won't build the compiler itself with -finstrument-functions.  For
that you need to use
    make BOOT_CFLAGS="-g -O2 -finstrument-functions"

Ian

Re: Question about instrumenting gcc

[Renata Hodovan]

Hi Ian,

thank you very much your help. I redesigned __cyg_profile_enter and 
__cyg_profile_exit functions and I stay in the right memory interval. 
Furthermore I'm using now main_constructor and main_destructor instead 
of start_test and end_test. The settings of make also seem good, it 
creates the output of instrumentation that way what I expected. But the 
compilation failed after a point because it couldn't understand e.g. 
size_t in filenames.h and a lot of similar errors in files what I 
haven't touched. I thought, since I add file IO operations to the main.c 
or linking the instrument file to the project, I distracted somehow the 
dependency. At this point I tried to compile the gcc --disable-bootstrap 
but then the instrumentation flag hadn't any results. What is the 
problem now? :S

thanks,
Reni


On 06/05/2012 09:38 PM, Ian Lance Taylor wrote:
> Renata Hodovan<hodovan@inf.u-szeged.hu>  writes:
>
>> I have to instrument gcc for some purposes. The goal is to be able to
>> track what GCC functions are called during a particularly
>> compile. Unfortunately I'm not really familiar with the architecture
>> of GCC so I need a little help. I tried the following steps:
>>
>> 1) Hacking gcc/Makefile.in and adding "-finstrument-functions" flag to
>> T_CFLAGS.
>> 2) I have an already implemented and tested version of __start_test__
>> and __end_test__ functions. They are called from gcc/main.c, before
>> and after toplev_main() call. The containing file is linked to gcc
>> (the object is added to OBJS-common and the dependency is defined
>> later in gcc/Makefile.in)
>> 3) Downloading prerequisites with contrib/download_prerequisites.
>> 4) Executing the configuration from a clean build directory (on the
>> same level with the source dir): ./../gcc-4.6.2/configure
>> --prefix="/opt/gcc-4.6.2/" --enable-languages="c,c++"
>> 5) Starting the build with "make all"
>>
>> This way I ran out of memory, although I had 28G.
> You should not have run out of memory at this step, not with 28G.
> Adding -finstrument-functions should not have significantly changed the
> memory usage of the build.
>
> But I suppose you could have run out of memory if your
> __cyg_profile_func_enter function uses a lot of memory.  If that is the
> problem, then that is what you will have to fix.
>
>
>> Next I tried to remove the T_CFLAGS settings from the Makefile and
>> gave -finstrument-functions to the make command:
>> make CFLAGS+="-finstrument-functions" all
> This won't build the compiler itself with -finstrument-functions.  For
> that you need to use
>      make BOOT_CFLAGS="-g -O2 -finstrument-functions"
>
> Ian

Re: Question about instrumenting gcc

[Ian Lance Taylor]

Renata Hodovan <hodovan@inf.u-szeged.hu> writes:

> thank you very much your help. I redesigned __cyg_profile_enter and
> __cyg_profile_exit functions and I stay in the right memory
> interval. Furthermore I'm using now main_constructor and
> main_destructor instead of start_test and end_test. The settings of
> make also seem good, it creates the output of instrumentation that way
> what I expected. But the compilation failed after a point because it
> couldn't understand e.g. size_t in filenames.h and a lot of similar
> errors in files what I haven't touched. I thought, since I add file IO
> operations to the main.c or linking the instrument file to the
> project, I distracted somehow the dependency. At this point I tried to
> compile the gcc --disable-bootstrap but then the instrumentation flag
> hadn't any results. What is the problem now? :S

When you configure with --disable-bootstrap, use CFLAGS rather than
BOOT_CFLAGS.  BOOT_CFLAGS only applies to stages 2 and 3.

I have no idea what the problem is when bootstrapping.  You would need
to provide more information.

Ian

Re: Question about instrumenting gcc

[Renata Hodovan]

Thank you, Ian. This way instrumenting works fine!

Cheers,
Reni


On 06/12/2012 06:41 PM, Ian Lance Taylor wrote:
> Renata Hodovan<hodovan@inf.u-szeged.hu>  writes:
>
>> thank you very much your help. I redesigned __cyg_profile_enter and
>> __cyg_profile_exit functions and I stay in the right memory
>> interval. Furthermore I'm using now main_constructor and
>> main_destructor instead of start_test and end_test. The settings of
>> make also seem good, it creates the output of instrumentation that way
>> what I expected. But the compilation failed after a point because it
>> couldn't understand e.g. size_t in filenames.h and a lot of similar
>> errors in files what I haven't touched. I thought, since I add file IO
>> operations to the main.c or linking the instrument file to the
>> project, I distracted somehow the dependency. At this point I tried to
>> compile the gcc --disable-bootstrap but then the instrumentation flag
>> hadn't any results. What is the problem now? :S
> When you configure with --disable-bootstrap, use CFLAGS rather than
> BOOT_CFLAGS.  BOOT_CFLAGS only applies to stages 2 and 3.
>
> I have no idea what the problem is when bootstrapping.  You would need
> to provide more information.
>
> Ian