This is the mail archive of the
mailing list for the GCC project.
Re: [patch, testsuite, spu] Add --param max-completely-peel-times=4 for some test cases
- From: "Ulrich Weigand" <uweigand at de dot ibm dot com>
- To: richard dot guenther at gmail dot com (Richard Guenther)
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Thu, 16 Apr 2009 15:41:08 +0200 (CEST)
- Subject: Re: [patch, testsuite, spu] Add --param max-completely-peel-times=4 for some test cases
Richard Guenther wrote:
> Why not instead xfail the tests for SPU?
Well, this way the tests actually run and pass ... I thought this
was more useful ...
> [note that I think you should
> have adjusted max-completely-peeled-insns instead as you are
> obviously against code-growth]
The problem is that if estimated_unrolled_size *thinks* the
total size will shrink, unrolling is always done, no matter
what max-completely-peeled-insns says. Unfortunately, the
size estimated by estimated_unrolled_size isn't always right,
so this can lead to code growth ...
The only way to always disable automatic unrolling right now
seems to be to set max-completely-peel-times.
Dr. Ulrich Weigand
GNU Toolchain for Linux on System z and Cell BE