This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Speculative prefetching: report
- From: Dale Johannesen <dalej at apple dot com>
- To: "gcc at gcc dot gnu dot org List" <gcc at gcc dot gnu dot org>
- Cc: Zdenek Dvorak <rakdver at atrey dot karlin dot mff dot cuni dot cz>, Dale Johannesen <dalej at apple dot com>
- Date: Tue, 12 Oct 2004 12:17:02 -0700
- Subject: Speculative prefetching: report
Has anybody been trying the recently added prefetching based on feedback
(-fspeculative-prefetching)? I found the following problems on
darwin-ppc:
1. It has a bug which causes ICEs. The fix is here.
http://gcc.gnu.org/ml/gcc-patches/2004-09/msg01277.html
Could somebody please approve?
2. In the profile-generation phase, there is a huge code bloat
resutling from
instrumenting all the loads and stores. This badly stresses the rest
of the BE,
causing out-of-memory ICEs or unacceptably slow compile times (hours) on
several SPECmarks. (With IMA to be sure.)
3. The SPECmarks which did build got consistently worse results.
I did no deep analysis, but it appears to be much too aggressive about
prefetching; it will prefetch each store of an unrolled memset loop
individually,
for example I think it needs some concept of cache line size to be
useful.
This is a Cool Idea, but in its current form I don't think it belongs in
-fprofile-generate/use. Comments?