https://bugs.llvm.org/show_bug.cgi?id=46966
Bug ID: 46966
Summary: Suboptimal codegen for v2sf operations - should use
MOVLPS
Product: libraries
Version: trunk
Hardware: PC
OS: Linux
Status: NEW
Severity: enhancement
Priority: P
Component: Backend: X86
Assignee: unassignedb...@nondot.org
Reporter: david.bolvan...@gmail.com
CC: craig.top...@gmail.com, llvm-bugs@lists.llvm.org,
llvm-...@redking.me.uk, spatel+l...@rotateright.com
float r[2], a[2], b[2], c[2];
void test_plus(void) {
for (int i = 0; i < 2; i++) r[i] = a[i] + b[i];
}
Clang:
test_plus: # @test_plus
movss xmm0, dword ptr [rip + a] # xmm0 = mem[0],zero,zero,zero
addss xmm0, dword ptr [rip + b]
movss dword ptr [rip + r], xmm0
movss xmm0, dword ptr [rip + a+4] # xmm0 = mem[0],zero,zero,zero
addss xmm0, dword ptr [rip + b+4]
movss dword ptr [rip + r+4], xmm0
ret
GCC:
test_plus:
movq xmm0, QWORD PTR a[rip]
movq xmm1, QWORD PTR b[rip]
addps xmm0, xmm1
movlps QWORD PTR r[rip], xmm0
ret
Godbolt: https://godbolt.org/z/fnxGj1
--
You are receiving this mail because:
You are on the CC list for the bug._______________________________________________
llvm-bugs mailing list
llvm-bugs@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/llvm-bugs
