1 ; RUN: opt < %s -instcombine -S | FileCheck %s 2 target datalayout = "E-p:64:64:64-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:32:64-f32:32:32-f64:64:64-v64:64:64-v128:128:128-a0:0:64" 3 @C.0.1248 = internal constant [128 x float] [ float -1.000000e+00, float -1.000000e+00, float -1.000000e+00, float 0.000000e+00, float -1.000000e+00, float -1.000000e+00, float 0.000000e+00, float -1.000000e+00, float -1.000000e+00, float -1.000000e+00, float 0.000000e+00, float 1.000000e+00, float -1.000000e+00, float -1.000000e+00, float 1.000000e+00, float 0.000000e+00, float -1.000000e+00, float 0.000000e+00, float -1.000000e+00, float -1.000000e+00, float -1.000000e+00, float 0.000000e+00, float -1.000000e+00, float 1.000000e+00, float -1.000000e+00, float 0.000000e+00, float 1.000000e+00, float -1.000000e+00, float -1.000000e+00, float 0.000000e+00, float 1.000000e+00, float 1.000000e+00, float -1.000000e+00, float 1.000000e+00, float -1.000000e+00, float 0.000000e+00, float -1.000000e+00, float 1.000000e+00, float 0.000000e+00, float -1.000000e+00, float -1.000000e+00, float 1.000000e+00, float 0.000000e+00, float 1.000000e+00, float -1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 0.000000e+00, float 0.000000e+00, float -1.000000e+00, float -1.000000e+00, float -1.000000e+00, float 0.000000e+00, float -1.000000e+00, float -1.000000e+00, float 1.000000e+00, float 0.000000e+00, float -1.000000e+00, float 1.000000e+00, float -1.000000e+00, float 0.000000e+00, float -1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float -1.000000e+00, float -1.000000e+00, float 0.000000e+00, float 1.000000e+00, float -1.000000e+00, float 0.000000e+00, float -1.000000e+00, float 1.000000e+00, float -1.000000e+00, float 0.000000e+00, float 1.000000e+00, float 1.000000e+00, float -1.000000e+00, float 1.000000e+00, float 0.000000e+00, float 1.000000e+00, float 0.000000e+00, float -1.000000e+00, float -1.000000e+00, float 1.000000e+00, float 0.000000e+00, float -1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 0.000000e+00, float 1.000000e+00, float -1.000000e+00, float 1.000000e+00, float 0.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float -1.000000e+00, float 0.000000e+00, float 1.000000e+00, float 1.000000e+00, float 0.000000e+00, float -1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 0.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00, float 0.000000e+00, float 0.000000e+00, float 1.000000e+00, float -1.000000e+00, float -1.000000e+00, float 0.000000e+00, float 1.000000e+00, float -1.000000e+00, float 1.000000e+00, float 0.000000e+00, float 1.000000e+00, float 1.000000e+00, float -1.000000e+00, float 0.000000e+00, float 1.000000e+00, float 1.000000e+00, float 1.000000e+00 ], align 32 ; <[128 x float]*> [#uses=1] 4 5 define float @test1(i32 %hash, float %x, float %y, float %z, float %w) { 6 entry: 7 %lookupTable = alloca [128 x float], align 16 ; <[128 x float]*> [#uses=5] 8 %lookupTable1 = bitcast [128 x float]* %lookupTable to i8* ; <i8*> [#uses=1] 9 call void @llvm.memcpy.p0i8.p0i8.i64(i8* %lookupTable1, i8* bitcast ([128 x float]* @C.0.1248 to i8*), i64 512, i32 16, i1 false) 10 11 ; CHECK: @test1 12 ; CHECK-NOT: alloca 13 ; CHECK-NOT: call{{.*}}@llvm.memcpy 14 15 %tmp3 = shl i32 %hash, 2 ; <i32> [#uses=1] 16 %tmp5 = and i32 %tmp3, 124 ; <i32> [#uses=4] 17 %tmp753 = getelementptr [128 x float]* %lookupTable, i32 0, i32 %tmp5 ; <float*> [#uses=1] 18 %tmp9 = load float* %tmp753 ; <float> [#uses=1] 19 %tmp11 = fmul float %tmp9, %x ; <float> [#uses=1] 20 %tmp13 = fadd float %tmp11, 0.000000e+00 ; <float> [#uses=1] 21 %tmp17.sum52 = or i32 %tmp5, 1 ; <i32> [#uses=1] 22 %tmp1851 = getelementptr [128 x float]* %lookupTable, i32 0, i32 %tmp17.sum52 ; <float*> [#uses=1] 23 %tmp19 = load float* %tmp1851 ; <float> [#uses=1] 24 %tmp21 = fmul float %tmp19, %y ; <float> [#uses=1] 25 %tmp23 = fadd float %tmp21, %tmp13 ; <float> [#uses=1] 26 %tmp27.sum50 = or i32 %tmp5, 2 ; <i32> [#uses=1] 27 %tmp2849 = getelementptr [128 x float]* %lookupTable, i32 0, i32 %tmp27.sum50 ; <float*> [#uses=1] 28 %tmp29 = load float* %tmp2849 ; <float> [#uses=1] 29 %tmp31 = fmul float %tmp29, %z ; <float> [#uses=1] 30 %tmp33 = fadd float %tmp31, %tmp23 ; <float> [#uses=1] 31 %tmp37.sum48 = or i32 %tmp5, 3 ; <i32> [#uses=1] 32 %tmp3847 = getelementptr [128 x float]* %lookupTable, i32 0, i32 %tmp37.sum48 ; <float*> [#uses=1] 33 %tmp39 = load float* %tmp3847 ; <float> [#uses=1] 34 %tmp41 = fmul float %tmp39, %w ; <float> [#uses=1] 35 %tmp43 = fadd float %tmp41, %tmp33 ; <float> [#uses=1] 36 ret float %tmp43 37 } 38 39 declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture, i8* nocapture, i64, i32, i1) nounwind 40 41 %T = type { i8, [123 x i8] } 42 %U = type { i32, i32, i32, i32, i32 } 43 44 @G = constant %T {i8 1, [123 x i8] zeroinitializer } 45 @H = constant [2 x %U] zeroinitializer, align 16 46 47 define void @test2() { 48 %A = alloca %T 49 %B = alloca %T 50 %a = bitcast %T* %A to i8* 51 %b = bitcast %T* %B to i8* 52 53 ; CHECK: @test2 54 55 ; %A alloca is deleted 56 ; CHECK-NEXT: alloca [124 x i8] 57 ; CHECK-NEXT: getelementptr inbounds [124 x i8]* 58 59 ; use @G instead of %A 60 ; CHECK-NEXT: call void @llvm.memcpy.p0i8.p0i8.i64(i8* %{{.*}}, i8* getelementptr inbounds (%T* @G, i64 0, i32 0) 61 call void @llvm.memcpy.p0i8.p0i8.i64(i8* %a, i8* bitcast (%T* @G to i8*), i64 124, i32 4, i1 false) 62 call void @llvm.memcpy.p0i8.p0i8.i64(i8* %b, i8* %a, i64 124, i32 4, i1 false) 63 call void @bar(i8* %b) 64 ret void 65 } 66 67 declare void @bar(i8*) 68 69 70 ;; Should be able to eliminate the alloca. 71 define void @test3() { 72 %A = alloca %T 73 %a = bitcast %T* %A to i8* 74 call void @llvm.memcpy.p0i8.p0i8.i64(i8* %a, i8* bitcast (%T* @G to i8*), i64 124, i32 4, i1 false) 75 call void @bar(i8* %a) readonly 76 ; CHECK: @test3 77 ; CHECK-NEXT: call void @bar(i8* getelementptr inbounds (%T* @G, i64 0, i32 0)) 78 ret void 79 } 80 81 define void @test4() { 82 %A = alloca %T 83 %a = bitcast %T* %A to i8* 84 call void @llvm.memcpy.p0i8.p0i8.i64(i8* %a, i8* bitcast (%T* @G to i8*), i64 124, i32 4, i1 false) 85 call void @baz(i8* byval %a) 86 ; CHECK: @test4 87 ; CHECK-NEXT: call void @baz(i8* byval getelementptr inbounds (%T* @G, i64 0, i32 0)) 88 ret void 89 } 90 91 declare void @llvm.lifetime.start(i64, i8*) 92 define void @test5() { 93 %A = alloca %T 94 %a = bitcast %T* %A to i8* 95 call void @llvm.lifetime.start(i64 -1, i8* %a) 96 call void @llvm.memcpy.p0i8.p0i8.i64(i8* %a, i8* bitcast (%T* @G to i8*), i64 124, i32 4, i1 false) 97 call void @baz(i8* byval %a) 98 ; CHECK: @test5 99 ; CHECK-NEXT: call void @baz(i8* byval getelementptr inbounds (%T* @G, i64 0, i32 0)) 100 ret void 101 } 102 103 104 declare void @baz(i8* byval) 105 106 107 define void @test6() { 108 %A = alloca %U, align 16 109 %a = bitcast %U* %A to i8* 110 call void @llvm.memcpy.p0i8.p0i8.i64(i8* %a, i8* bitcast ([2 x %U]* @H to i8*), i64 20, i32 16, i1 false) 111 call void @bar(i8* %a) readonly 112 ; CHECK: @test6 113 ; CHECK-NEXT: call void @bar(i8* bitcast ([2 x %U]* @H to i8*)) 114 ret void 115 } 116 117 define void @test7() { 118 %A = alloca %U, align 16 119 %a = bitcast %U* %A to i8* 120 call void @llvm.memcpy.p0i8.p0i8.i64(i8* %a, i8* bitcast (%U* getelementptr ([2 x %U]* @H, i64 0, i32 0) to i8*), i64 20, i32 4, i1 false) 121 call void @bar(i8* %a) readonly 122 ; CHECK: @test7 123 ; CHECK-NEXT: call void @bar(i8* bitcast ([2 x %U]* @H to i8*)) 124 ret void 125 } 126 127 define void @test8() { 128 %A = alloca %U, align 16 129 %a = bitcast %U* %A to i8* 130 call void @llvm.memcpy.p0i8.p0i8.i64(i8* %a, i8* bitcast (%U* getelementptr ([2 x %U]* @H, i64 0, i32 1) to i8*), i64 20, i32 4, i1 false) 131 call void @bar(i8* %a) readonly 132 ; CHECK: @test8 133 ; CHECK: llvm.memcpy 134 ; CHECK: bar 135 ret void 136 } 137
