1
; RUN: llc -mtriple=x86_64-unknown-unknown -mattr=+sse2 < %s | FileCheck %s --check-prefix=SSE
2
; RUN: llc -mtriple=x86_64-unknown-unknown -mattr=+avx < %s | FileCheck %s --check-prefix=AVX
4
; Verify that we're folding the load into the math instruction.
6
define float @rcpss(float* %a) {
9
; SSE-NEXT: rcpss (%rdi), %xmm0
14
; AVX-NEXT: vrcpss (%rdi), %xmm0, %xmm0
16
%ld = load float, float* %a
17
%ins = insertelement <4 x float> undef, float %ld, i32 0
18
%res = tail call <4 x float> @llvm.x86.sse.rcp.ss(<4 x float> %ins)
19
%ext = extractelement <4 x float> %res, i32 0
23
define float @rsqrtss(float* %a) {
26
; SSE-NEXT: rsqrtss (%rdi), %xmm0
31
; AVX-NEXT: vrsqrtss (%rdi), %xmm0, %xmm0
33
%ld = load float, float* %a
34
%ins = insertelement <4 x float> undef, float %ld, i32 0
35
%res = tail call <4 x float> @llvm.x86.sse.rsqrt.ss(<4 x float> %ins)
36
%ext = extractelement <4 x float> %res, i32 0
40
define float @sqrtss(float* %a) {
43
; SSE-NEXT: sqrtss (%rdi), %xmm0
48
; AVX-NEXT: vsqrtss (%rdi), %xmm0, %xmm0
50
%ld = load float, float* %a
51
%ins = insertelement <4 x float> undef, float %ld, i32 0
52
%res = tail call <4 x float> @llvm.x86.sse.sqrt.ss(<4 x float> %ins)
53
%ext = extractelement <4 x float> %res, i32 0
57
define double @sqrtsd(double* %a) {
60
; SSE-NEXT: sqrtsd (%rdi), %xmm0
65
; AVX-NEXT: vsqrtsd (%rdi), %xmm0, %xmm0
67
%ld = load double, double* %a
68
%ins = insertelement <2 x double> undef, double %ld, i32 0
69
%res = tail call <2 x double> @llvm.x86.sse2.sqrt.sd(<2 x double> %ins)
70
%ext = extractelement <2 x double> %res, i32 0
75
declare <4 x float> @llvm.x86.sse.rcp.ss(<4 x float>) nounwind readnone
76
declare <4 x float> @llvm.x86.sse.rsqrt.ss(<4 x float>) nounwind readnone
77
declare <4 x float> @llvm.x86.sse.sqrt.ss(<4 x float>) nounwind readnone
78
declare <2 x double> @llvm.x86.sse2.sqrt.sd(<2 x double>) nounwind readnone