1 ; RUN: llc < %s -mtriple=x86_64-apple-darwin | FileCheck %s 2 3 define <4 x float> @foo(<4 x float> %val, <4 x float> %test) nounwind { 4 ; CHECK-LABEL: LCPI0_0: 5 ; CHECK-NEXT: .long 1065353216 ## 0x3f800000 6 ; CHECK-NEXT: .long 1065353216 ## 0x3f800000 7 ; CHECK-NEXT: .long 1065353216 ## 0x3f800000 8 ; CHECK-NEXT: .long 1065353216 ## 0x3f800000 9 ; CHECK-LABEL: foo: 10 ; CHECK: ## %bb.0: 11 ; CHECK-NEXT: cmpeqps %xmm1, %xmm0 12 ; CHECK-NEXT: andps {{.*}}(%rip), %xmm0 13 ; CHECK-NEXT: retq 14 %cmp = fcmp oeq <4 x float> %val, %test 15 %ext = zext <4 x i1> %cmp to <4 x i32> 16 %result = sitofp <4 x i32> %ext to <4 x float> 17 ret <4 x float> %result 18 } 19 20 ; Make sure the operation doesn't try to get folded when the sizes don't match, 21 ; as that ends up crashing later when trying to form a bitcast operation for 22 ; the folded nodes. 23 define void @foo1(<4 x float> %val, <4 x float> %test, <4 x double>* %p) nounwind { 24 ; CHECK-LABEL: LCPI1_0: 25 ; CHECK-NEXT: .long 1 ## 0x1 26 ; CHECK-NEXT: .long 1 ## 0x1 27 ; CHECK-NEXT: .long 1 ## 0x1 28 ; CHECK-NEXT: .long 1 ## 0x1 29 ; CHECK-LABEL: foo1: 30 ; CHECK: ## %bb.0: 31 ; CHECK-NEXT: cmpeqps %xmm1, %xmm0 32 ; CHECK-NEXT: andps {{.*}}(%rip), %xmm0 33 ; CHECK-NEXT: pshufd {{.*#+}} xmm1 = xmm0[2,3,0,1] 34 ; CHECK-NEXT: cvtdq2pd %xmm1, %xmm1 35 ; CHECK-NEXT: cvtdq2pd %xmm0, %xmm0 36 ; CHECK-NEXT: movaps %xmm0, (%rdi) 37 ; CHECK-NEXT: movaps %xmm1, 16(%rdi) 38 ; CHECK-NEXT: retq 39 %cmp = fcmp oeq <4 x float> %val, %test 40 %ext = zext <4 x i1> %cmp to <4 x i32> 41 %result = sitofp <4 x i32> %ext to <4 x double> 42 store <4 x double> %result, <4 x double>* %p 43 ret void 44 } 45 46 ; Also test the general purpose constant folding of int->fp. 47 define void @foo2(<4 x float>* noalias %result) nounwind { 48 ; CHECK-LABEL: LCPI2_0: 49 ; CHECK-NEXT: .long 1082130432 ## float 4 50 ; CHECK-NEXT: .long 1084227584 ## float 5 51 ; CHECK-NEXT: .long 1086324736 ## float 6 52 ; CHECK-NEXT: .long 1088421888 ## float 7 53 ; CHECK-LABEL: foo2: 54 ; CHECK: ## %bb.0: 55 ; CHECK-NEXT: movaps {{.*#+}} xmm0 = [4.000000e+00,5.000000e+00,6.000000e+00,7.000000e+00] 56 ; CHECK-NEXT: movaps %xmm0, (%rdi) 57 ; CHECK-NEXT: retq 58 %val = uitofp <4 x i32> <i32 4, i32 5, i32 6, i32 7> to <4 x float> 59 store <4 x float> %val, <4 x float>* %result 60 ret void 61 } 62 63 ; Fold explicit AND operations when the constant isn't a splat of a single 64 ; scalar value like what the zext creates. 65 define <4 x float> @foo3(<4 x float> %val, <4 x float> %test) nounwind { 66 ; CHECK-LABEL: LCPI3_0: 67 ; CHECK-NEXT: .long 1065353216 ## 0x3f800000 68 ; CHECK-NEXT: .long 0 ## 0x0 69 ; CHECK-NEXT: .long 1065353216 ## 0x3f800000 70 ; CHECK-NEXT: .long 0 ## 0x0 71 ; CHECK-LABEL: foo3: 72 ; CHECK: ## %bb.0: 73 ; CHECK-NEXT: cmpeqps %xmm1, %xmm0 74 ; CHECK-NEXT: andps {{.*}}(%rip), %xmm0 75 ; CHECK-NEXT: retq 76 %cmp = fcmp oeq <4 x float> %val, %test 77 %ext = zext <4 x i1> %cmp to <4 x i32> 78 %and = and <4 x i32> %ext, <i32 255, i32 256, i32 257, i32 258> 79 %result = sitofp <4 x i32> %and to <4 x float> 80 ret <4 x float> %result 81 } 82 83 ; Test the general purpose constant folding of uint->fp. 84 define void @foo4(<4 x float>* noalias %result) nounwind { 85 ; CHECK-LABEL: LCPI4_0: 86 ; CHECK-NEXT: .long 1065353216 ## float 1 87 ; CHECK-NEXT: .long 1123942400 ## float 127 88 ; CHECK-NEXT: .long 1124073472 ## float 128 89 ; CHECK-NEXT: .long 1132396544 ## float 255 90 ; CHECK-LABEL: foo4: 91 ; CHECK: ## %bb.0: 92 ; CHECK-NEXT: movaps {{.*#+}} xmm0 = [1.000000e+00,1.270000e+02,1.280000e+02,2.550000e+02] 93 ; CHECK-NEXT: movaps %xmm0, (%rdi) 94 ; CHECK-NEXT: retq 95 %val = uitofp <4 x i8> <i8 1, i8 127, i8 -128, i8 -1> to <4 x float> 96 store <4 x float> %val, <4 x float>* %result 97 ret void 98 } 99
