https://gcc.gnu.org/bugzilla/show_bug.cgi?id=123343
--- Comment #15 from rguenther at suse dot de <rguenther at suse dot de> --- On Wed, 14 Jan 2026, chenzhongyao.hit at gmail dot com wrote: > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=123343 > > --- Comment #14 from Zhongyao Chen <chenzhongyao.hit at gmail dot com> --- > Created attachment 63341 > --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=63341&action=edit > patch > > Please take a quick look. If this looks OK, I'll submit to the mailing list. So the vect_analyze_slp_reduction_group hunk looks reasonable to me, the second hunk is better placed after the group_size = scalar_stmts.length (); and should be a compile-time optimization only. The effect of the patch should be that we get the reduction-chain by re-association handling more often. So please test and post the patch with the above adjustment.
