This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH, committed] Function specific option changes (IA-64, hot/cold, scheduling)
On Sun, Jul 27, 2008 at 4:28 PM, Michael Meissner <gnu@the-meissners.org> wrote:
> On the other hand, when I presented the function specific proposal at the 2007
> GCC summit, I had many people ask for this exact feature. Most of these people
> work on embedded systems where memory is often times tight. I can imagine with
> profile guided feedback that you may want cold functions to be optimized for
> space and hot functions with extra optimization (that way you don't decorate
> the functions at all, which is the ultimate goal of function specific options).
I want this feature too, I have been getting many requests for a way
to mark a function as being cold and automatically switch over to use
-Os for those cold functions. I think the issue with SPEC is that we
are marking functions which are not really cold as cold. I think we
need to better decide with the heuristics which we have than just
disabling this.
The reason why we want to switch cold functions to -Os is so they take
less memory.
Thanks,
Andrew Pinski
- References:
- [PATCH, committed] Function specific option changes (IA-64, hot/cold, scheduling)
- Re: [PATCH, committed] Function specific option changes (IA-64, hot/cold, scheduling)
- Re: [PATCH, committed] Function specific option changes (IA-64, hot/cold, scheduling)
- Re: [PATCH, committed] Function specific option changes (IA-64, hot/cold, scheduling)
- Re: [PATCH, committed] Function specific option changes (IA-64, hot/cold, scheduling)