This is the mail archive of the
gcc-help@gcc.gnu.org
mailing list for the GCC project.
Re: Parallel execution
- From: Ian Lance Taylor <iant at google dot com>
- To: Hasan MUTLU <bigpatboy at yahoo dot com>
- Cc: gcc-help at gcc dot gnu dot org
- Date: Wed, 13 Apr 2011 13:08:57 -0700
- Subject: Re: Parallel execution
- References: <105673.74895.qm@web130109.mail.mud.yahoo.com>
Hasan MUTLU <bigpatboy@yahoo.com> writes:
> I have implemented a multiprocessor system and want to run a code on my system.
> My idea is to have a compiler support that tells me which instructions could be
> executed in parallel without any problems. In other words, I have a given code
> that is written without considering any kind of parallelization (a standard C
> code written for single processor) and I want my compiler to tell me which
> instructions don't have any kind of dependencies, so could be run in parallel.
> This might remind you VLIW, which deals more or less with the same problem. You
> can ignore how my hardware deals with this information. My question is how close
> I can get to this information with gcc compiler. In other words, how can I
> determine instruction dependencies, the instructions that could be run in
> parallel? Any command line options (if there is) or a pointer in the source code
> of gcc would be helpful.
gcc doesn't support this directly. You can get the information by using
-fdump-rtl-sched2. Also fiddle with -fsched-verbose=N.
Ian