https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113090
Bug ID: 113090
Summary: Suboptimal vector permuation for 64-bit vector.
Product: gcc
Version: 14.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: target
Assignee: unassigned at gcc dot gnu.org
Reporter: liuhongt at gcc dot gnu.org
Target Milestone: ---
When working on PR113079, loop vectorizer try to reduc sum of v2si with
permuation, x86 backend generates
typedef int v2si __attribute__((vector_size(8)));
v2si
foo (v2si a, v2si b)
{
return __builtin_shufflevector (a, b, 1, 2);
}
foo(int __vector(2), int __vector(2)):
vpshufb xmm0, xmm0, XMMWORD PTR .LC0[rip]
vpshufb xmm1, xmm1, XMMWORD PTR .LC1[rip]
vpor xmm0, xmm0, xmm1
But it can be better with
.cfi_startproc
punpcklqdq %xmm1, %xmm0
pshufd $153, %xmm0, %xmm0
ret