https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68775
--- Comment #5 from rguenther at suse dot de <rguenther at suse dot de> --- On Wed, 9 Dec 2015, seurer at linux dot vnet.ibm.com wrote: > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68775 > > --- Comment #4 from William Seurer <seurer at linux dot vnet.ibm.com> --- > (In reply to Richard Biener from comment #1) > > Can you try to bisect to an affected source file (just use > > -fno-tree-vectorize on all but one) and there to whether it is basic-block > > or loop vectorization > > (-fno-tree-loop-vectorize / -fno-tree-slp-vectorize) and specific > > loop/basic-block with -fdbg-cnt=vect_slp:N or -fdbg-cnt=vect_loop:N? > > > > Note that it seems that tonto also fails (since forever) with AVX2 on > > x86_64. > > It still fails with just -fno-tree-vectorize but succeeds with both > -fno-tree-loop-vectorize and -fno-tree-slp-vectorize. That's a very odd observation as -fno-tree-vectorize should be equivalent to -fno-tree-loop-vectorize -fno-tree-slp-vectorize. I understand your sentence correct in that -fno-tree-loop-vectorize makes 465.tonto no longer miscompare. The same is true for -fno-tree-slp-vectorize. But with -fno-tree-vectorize 465.tonto still miscompares? > I am about to go through it flipping those for individual modules. Thanks.
