https://gcc.gnu.org/bugzilla/show_bug.cgi?id=83203
Richard Biener <rguenth at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Keywords| |needs-bisection
Status|UNCONFIRMED |NEW
Last reconfirmed| |2017-11-29
CC| |jakub at gcc dot gnu.org,
| |rguenth at gcc dot gnu.org
Target Milestone|--- |8.0
Summary|Inefficient int to avx2 |[8 Regression] Inefficient
|vector conversion |int to avx2 vector
| |conversion
Ever confirmed|0 |1
--- Comment #1 from Richard Biener <rguenth at gcc dot gnu.org> ---
Confirmed. On trunk we even genrate
foo:
.LFB5186:
.cfi_startproc
pushq %rbp
.cfi_def_cfa_offset 16
.cfi_offset 6, -16
movq %rsp, %rbp
.cfi_def_cfa_register 6
andq $-32, %rsp
movq %rdi, -8(%rsp)
vmovq -8(%rsp), %xmm1
vmovq %xmm1, %xmm0
vmovdqa %xmm0, %xmm0
leave
huh...
We expand from
foo (long int x)
{
__m256i _2;
<bb 2> [100.00%]:
_2 = {x_1(D)};
return _2;
that's a regression on trunk (gcc7 code confirmed). Probably caused by
the vec_init expansion reorg?
