Patchwork Reduce complette unrolling & peeling limits

login
register
mail settings
Submitter Jan Hubicka
Date Nov. 21, 2012, 4:27 p.m.
Message ID <20121121162702.GB32567@kam.mff.cuni.cz>
Download mbox | patch
Permalink /patch/200803/
State New
Headers show

Comments

Jan Hubicka - Nov. 21, 2012, 4:27 p.m.
> > Did you notice that gcc.c-torture/compile/pr43186.c regressed?  It now again
> > takes a while to compile, so times out on slow machines:
> > ...
> 
> On a 2.5Ghz Core2Duo, compiling the test with revision 192891 (2012-10-28)
> takes a small fraction of a second, while with revision 193270 (2012-11-06)
> it takes ~25s.
> 
> However this patch makes gfortran.dg/reassoc_4.f to fail
> 
> FAIL: gfortran.dg/reassoc_4.f  -O   scan-tree-dump-times reassoc1 "[0-9] \\\\* " 22
> 
> After it 22 should be replaced with 16 (thresshold max-completely-peeled-insns=138
> gives 16, =139 gives 22).

I would propose the following patch instead.  The patch anyway changes the limits on some
targets, so lets change them on all.

Honza


> 
> Dominique
Dominique Dhumieres - Nov. 25, 2012, 12:22 p.m.
> ... I believe I posted a patch?

Yes: http://gcc.gnu.org/ml/gcc-patches/2012-11/msg01799.html

I have found another fall out: I have some avatars of the polyhedron tests
where the REAL(8) have been replaced with REAL(10). Some of them are now
Should I open a new PR for that?

Cheers,

Dominique
Dominique Dhumieres - Nov. 25, 2012, 12:25 p.m.
My mailer has eaten a line in my previous mail. One should read:

I have found another fall out: I have some avatars of the polyhedron tests
where the REAL(8) have been replaced with REAL(10). Some of them are now ~50%
slower with the new value of max-completely-peeled-insns.
Should I open a new PR for that?

Sorry for the noise.

Dominique

Patch

Index: reassoc_4.f
===================================================================
--- reassoc_4.f (revision 193698)
+++ reassoc_4.f (working copy)
@@ -1,7 +1,5 @@ 
 ! { dg-do compile }
-! { dg-options "-O3 -ffast-math -fdump-tree-reassoc1" }
-! { dg-additional-options "--param max-completely-peel-times=16" { target spu-*-* } }
-! { dg-additional-options "--param max-completely-peeled-insns=400" { target s390*-*-* } }
+! { dg-options "-O3 -ffast-math -fdump-tree-reassoc1 --param max-completely-peel-times=16 --param max-completely-peeled-insns=400" }
       subroutine anisonl(w,vo,anisox,s,ii1,jj1,weight)
       integer ii1,jj1,i1,iii1,j1,jjj1,k1,l1,m1,n1
       real*8 w(3,3),vo(3,3),anisox(3,3,3,3),s(60,60),weight