This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [PATCH] PR/7344, O(n^2) branch prediction
- From: tm <tm at mail dot kloo dot net>
- To: Brad Lucier <lucier at math dot purdue dot edu>
- Cc: Zdenek Dvorak <rakdver at atrey dot karlin dot mff dot cuni dot cz>,Daniel Berlin <dberlin at dberlin dot org>, jh at suse dot cz,gcc-patches at gcc dot gnu dot org
- Date: Sat, 7 Sep 2002 23:53:02 -0700 (PDT)
- Subject: Re: [PATCH] PR/7344, O(n^2) branch prediction
On Sat, 7 Sep 2002, Brad Lucier wrote:
> Did you test this new patch to see if it reduced the long runtimes
> you reported for branch prediction? I ran it on the test-case in
>
> http://gcc.gnu.org/ml/gcc/2002-08/msg01559.html
>
> and discovered no difference in run times.
>
> Brad
>
> > I think you're right. Here is the revised patch.
> >
> > 2002-09-06 Toshiyasu Morita <toshiyasu.morita@hsa.hitachi.com>
> >
> > * predict.c (process_note_prediction): Ignore multiway
> > branches to avoid excessively long compile times (PR/7344)
>
This patch is to fix PR/7344, so it doesn't fix any other problems.
Yep, I know there are many other problems. The testcase I "fixed" was
reduced from infinite compile time down to about 2.x minutes.
If you look at the PR/7344 testcase there really isn't any code in it
except for a switch statement. Yet GCSE winds up taking about 30% of the
2.x minutes of compile time for some reason, and instruction scheduling
winds up another 30% or so.
That testcase really shouldn't take more than a few seconds to compile
IMHO.
Toshi