https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64286
--- Comment #2 from Jeffrey A. Law <law at redhat dot com> --- On 12/16/14 09:10, izamyatin at gmail dot com wrote: > https://gcc.gnu.org/bugzilla/show_bug.cgi?id=64286 > > --- Comment #1 from Igor Zamyatin <izamyatin at gmail dot com> --- > Perhaps something like below to restrict ree for such cases? > > diff --git a/gcc/ree.c b/gcc/ree.c > index 3376901..92370ea 100644 > --- a/gcc/ree.c > +++ b/gcc/ree.c > @@ -1004,6 +1004,11 @@ add_removable_extension (const_rtx expr, rtx_insn > *insn, > struct df_link *defs, *def; > ext_cand *cand; > > + if (!SCALAR_INT_MODE_P (GET_MODE (dest)) > + && (GET_MODE_UNIT_PRECISION (mode) != > + GET_MODE_UNIT_PRECISION (GET_MODE (XEXP (src, 0))))) > + return; > + > /* First, make sure we can get all the reaching definitions. */ > defs = get_defs (insn, XEXP (src, 0), NULL); > if (!defs) > Funny, this is exactly the situation Jakub was concerned about in a comment for PR 59754. I think you should remove the SCALAR_INT_MODE_P test in combine_reaching_defs which I think should be dead code if we filter things in add_removable_extension. If you could make that change and confirm with Kirill that avx512f still looks good (in particular avx512f-vpmovzxwd-2.c) as well as doing a bootstrap and regression test, it'd be appreciated. Thanks, jeff
