1 ; RUN: opt < %s -scalarrepl -instcombine | \ 2 ; RUN: llc -march=x86 -mcpu=yonah | not grep sub.*esp 3 4 ; This checks that various insert/extract idiom work without going to the 5 ; stack. 6 target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:32:64-v64:64:64-v128:128:128-a0:0:64-f80:32:32" 7 8 define void @test(<4 x float>* %F, float %f) { 9 entry: 10 %tmp = load <4 x float>* %F ; <<4 x float>> [#uses=2] 11 %tmp3 = fadd <4 x float> %tmp, %tmp ; <<4 x float>> [#uses=1] 12 %tmp10 = insertelement <4 x float> %tmp3, float %f, i32 0 ; <<4 x float>> [#uses=2] 13 %tmp6 = fadd <4 x float> %tmp10, %tmp10 ; <<4 x float>> [#uses=1] 14 store <4 x float> %tmp6, <4 x float>* %F 15 ret void 16 } 17 18 define void @test2(<4 x float>* %F, float %f) { 19 entry: 20 %G = alloca <4 x float>, align 16 ; <<4 x float>*> [#uses=3] 21 %tmp = load <4 x float>* %F ; <<4 x float>> [#uses=2] 22 %tmp3 = fadd <4 x float> %tmp, %tmp ; <<4 x float>> [#uses=1] 23 store <4 x float> %tmp3, <4 x float>* %G 24 %tmp.upgrd.1 = getelementptr <4 x float>* %G, i32 0, i32 2 ; <float*> [#uses=1] 25 store float %f, float* %tmp.upgrd.1 26 %tmp4 = load <4 x float>* %G ; <<4 x float>> [#uses=2] 27 %tmp6 = fadd <4 x float> %tmp4, %tmp4 ; <<4 x float>> [#uses=1] 28 store <4 x float> %tmp6, <4 x float>* %F 29 ret void 30 } 31 32 define void @test3(<4 x float>* %F, float* %f) { 33 entry: 34 %G = alloca <4 x float>, align 16 ; <<4 x float>*> [#uses=2] 35 %tmp = load <4 x float>* %F ; <<4 x float>> [#uses=2] 36 %tmp3 = fadd <4 x float> %tmp, %tmp ; <<4 x float>> [#uses=1] 37 store <4 x float> %tmp3, <4 x float>* %G 38 %tmp.upgrd.2 = getelementptr <4 x float>* %G, i32 0, i32 2 ; <float*> [#uses=1] 39 %tmp.upgrd.3 = load float* %tmp.upgrd.2 ; <float> [#uses=1] 40 store float %tmp.upgrd.3, float* %f 41 ret void 42 } 43 44 define void @test4(<4 x float>* %F, float* %f) { 45 entry: 46 %tmp = load <4 x float>* %F ; <<4 x float>> [#uses=2] 47 %tmp5.lhs = extractelement <4 x float> %tmp, i32 0 ; <float> [#uses=1] 48 %tmp5.rhs = extractelement <4 x float> %tmp, i32 0 ; <float> [#uses=1] 49 %tmp5 = fadd float %tmp5.lhs, %tmp5.rhs ; <float> [#uses=1] 50 store float %tmp5, float* %f 51 ret void 52 } 53
