------- Comment #8 from dberlin at gcc dot gnu dot org 2006-07-22 13:30 ------- Subject: Re: [4.1 Regression] wrong optimization with -ftree-vectorize
rguenth at gcc dot gnu dot org wrote: > ------- Comment #5 from rguenth at gcc dot gnu dot org 2006-07-21 10:41 > ------- > On the mainline we produce > > Variable: D.1848, UID 1848, float *, symbol memory tag: SMT.4 > Variable: D.1851, UID 1851, const float *, symbol memory tag: SMT.4 > > while 4.1 branch does > > Variable: D.1604, UID 1604, float *, type memory tag: TMT.5 > Variable: D.1607, UID 1607, const float *, type memory tag: TMT.6 > > anyones bell ringing? Well, the symbol difference was likely caused by the TYPE_READONLY comparison that ensured that readonly types got their own SMT. However, doing *that* was simply wrong, even though it worked the majority of the time :). Oh, i see you figured that out :) > > -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=28029
