This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
[patch] Fix for PR34976: fix interchange profitability
- From: "Sebastian Pop" <sebpop at gmail dot com>
- To: "GCC Patches" <gcc-patches at gcc dot gnu dot org>, "Daniel Berlin" <dberlin at dberlin dot org>
- Date: Sat, 26 Jan 2008 22:08:10 -0600
- Subject: [patch] Fix for PR34976: fix interchange profitability
Hi,
it makes no sense to transform the testcase in this PR: the size of
the data processed by the loop is small enough to be in the L1 cache.
This patch fixes the PR by tuning the profitability function. Okay
for trunk after regstrap?
This patch does NOT fix the problem itself, because that would be to
rewrite a major part of the perfect_nestify HACK: I'm tired to tweak
this function that should be replaced by a correct loop distribution.
If you really want to know what this function produces, it's scarry,
but anyway, from:
do i=1,6
do j=1,6
gr(i,j)=0.d0
enddo
if(i.le.3) then
gr(i,i)=1.d0
else
gr(i,i)=0.5d0
endif
enddo
perfect_nestify transforms this code into:
do i=1,6
do j=1,6
gr(i,j)=0.d0
enddo
enddo
do i=1,6
gr(i,i)=1.d0
gr(i,i)=0.5d0
enddo
Yes, that's just removing the condition, and also other incredible
things... that end on the ICE.
--
Sebastian
AMD - GNU Tools