1 ; RUN: opt < %s -scalarrepl -instcombine | \
2 ; RUN: llc -march=x86 -mcpu=yonah | not grep sub.*esp
4 ; This checks that various insert/extract idiom work without going to the
7 define void @test(<4 x float>* %F, float %f) {
9 %tmp = load <4 x float>* %F ; <<4 x float>> [#uses=2]
10 %tmp3 = fadd <4 x float> %tmp, %tmp ; <<4 x float>> [#uses=1]
11 %tmp10 = insertelement <4 x float> %tmp3, float %f, i32 0 ; <<4 x float>> [#uses=2]
12 %tmp6 = fadd <4 x float> %tmp10, %tmp10 ; <<4 x float>> [#uses=1]
13 store <4 x float> %tmp6, <4 x float>* %F
17 define void @test2(<4 x float>* %F, float %f) {
19 %G = alloca <4 x float>, align 16 ; <<4 x float>*> [#uses=3]
20 %tmp = load <4 x float>* %F ; <<4 x float>> [#uses=2]
21 %tmp3 = fadd <4 x float> %tmp, %tmp ; <<4 x float>> [#uses=1]
22 store <4 x float> %tmp3, <4 x float>* %G
23 %tmp.upgrd.1 = getelementptr <4 x float>* %G, i32 0, i32 2 ; <float*> [#uses=1]
24 store float %f, float* %tmp.upgrd.1
25 %tmp4 = load <4 x float>* %G ; <<4 x float>> [#uses=2]
26 %tmp6 = fadd <4 x float> %tmp4, %tmp4 ; <<4 x float>> [#uses=1]
27 store <4 x float> %tmp6, <4 x float>* %F
31 define void @test3(<4 x float>* %F, float* %f) {
33 %G = alloca <4 x float>, align 16 ; <<4 x float>*> [#uses=2]
34 %tmp = load <4 x float>* %F ; <<4 x float>> [#uses=2]
35 %tmp3 = fadd <4 x float> %tmp, %tmp ; <<4 x float>> [#uses=1]
36 store <4 x float> %tmp3, <4 x float>* %G
37 %tmp.upgrd.2 = getelementptr <4 x float>* %G, i32 0, i32 2 ; <float*> [#uses=1]
38 %tmp.upgrd.3 = load float* %tmp.upgrd.2 ; <float> [#uses=1]
39 store float %tmp.upgrd.3, float* %f
43 define void @test4(<4 x float>* %F, float* %f) {
45 %tmp = load <4 x float>* %F ; <<4 x float>> [#uses=2]
46 %tmp5.lhs = extractelement <4 x float> %tmp, i32 0 ; <float> [#uses=1]
47 %tmp5.rhs = extractelement <4 x float> %tmp, i32 0 ; <float> [#uses=1]
48 %tmp5 = fadd float %tmp5.lhs, %tmp5.rhs ; <float> [#uses=1]
49 store float %tmp5, float* %f