This is the mail archive of the
gcc-bugs@gcc.gnu.org
mailing list for the GCC project.
[Bug tree-optimization/68695] [6 Regression] Performance regression related to ssa patch / ifcvt
- From: "rguenth at gcc dot gnu.org" <gcc-bugzilla at gcc dot gnu dot org>
- To: gcc-bugs at gcc dot gnu dot org
- Date: Fri, 04 Dec 2015 13:06:00 +0000
- Subject: [Bug tree-optimization/68695] [6 Regression] Performance regression related to ssa patch / ifcvt
- Auto-submitted: auto-generated
- References: <bug-68695-4 at http dot gcc dot gnu dot org/bugzilla/>
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=68695
--- Comment #2 from Richard Biener <rguenth at gcc dot gnu.org> ---
On x86_64 the testcase produces
foo:
.LFB0:
.cfi_startproc
cmpl %esi, %edi
movl %edx, %ecx
movl %edx, %eax
cmovle %esi, %ecx
cmovle %edi, %eax
imull %ecx, %eax
ret
we expand from
# i_1 = PHI <x_3(D)(2), a_5(D)(3)>
# j_2 = PHI <y_4(D)(2), a_5(D)(3)>
_6 = i_1 * j_2;
return _6;
and on x86_64 we coalesce i_1 with x_3 and j_2 with y_4 which would
match your "before change" generated code (copy a_5 to both other regs).
So RTL ifcvt dumps are required here, too.