1 ; NOTE: Assertions have been autogenerated by utils/update_test_checks.py 2 ; RUN: opt < %s -mtriple=x86_64-unknown -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=SSE 3 ; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=corei7-avx -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX1 4 ; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=core-avx2 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=AVX --check-prefix=AVX2 5 ; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=bdver2 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=XOP 6 ; RUN: opt < %s -mtriple=x86_64-unknown -mcpu=bdver4 -basicaa -slp-vectorizer -S | FileCheck %s --check-prefix=CHECK --check-prefix=XOP 7 8 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128" 9 10 @src64 = common global [4 x i64] zeroinitializer, align 32 11 @dst64 = common global [4 x i64] zeroinitializer, align 32 12 @src32 = common global [8 x i32] zeroinitializer, align 32 13 @dst32 = common global [8 x i32] zeroinitializer, align 32 14 @src16 = common global [16 x i16] zeroinitializer, align 32 15 @dst16 = common global [16 x i16] zeroinitializer, align 32 16 @src8 = common global [32 x i8] zeroinitializer, align 32 17 @dst8 = common global [32 x i8] zeroinitializer, align 32 18 19 declare i64 @llvm.bitreverse.i64(i64) 20 declare i32 @llvm.bitreverse.i32(i32) 21 declare i16 @llvm.bitreverse.i16(i16) 22 declare i8 @llvm.bitreverse.i8(i8) 23 24 define void @bitreverse_2i64() #0 { 25 ; SSE-LABEL: @bitreverse_2i64( 26 ; SSE-NEXT: [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 0), align 8 27 ; SSE-NEXT: [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 1), align 8 28 ; SSE-NEXT: [[BITREVERSE0:%.*]] = call i64 @llvm.bitreverse.i64(i64 [[LD0]]) 29 ; SSE-NEXT: [[BITREVERSE1:%.*]] = call i64 @llvm.bitreverse.i64(i64 [[LD1]]) 30 ; SSE-NEXT: store i64 [[BITREVERSE0]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 0), align 8 31 ; SSE-NEXT: store i64 [[BITREVERSE1]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 1), align 8 32 ; SSE-NEXT: ret void 33 ; 34 ; AVX-LABEL: @bitreverse_2i64( 35 ; AVX-NEXT: [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 0), align 8 36 ; AVX-NEXT: [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 1), align 8 37 ; AVX-NEXT: [[BITREVERSE0:%.*]] = call i64 @llvm.bitreverse.i64(i64 [[LD0]]) 38 ; AVX-NEXT: [[BITREVERSE1:%.*]] = call i64 @llvm.bitreverse.i64(i64 [[LD1]]) 39 ; AVX-NEXT: store i64 [[BITREVERSE0]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 0), align 8 40 ; AVX-NEXT: store i64 [[BITREVERSE1]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 1), align 8 41 ; AVX-NEXT: ret void 42 ; 43 ; XOP-LABEL: @bitreverse_2i64( 44 ; XOP-NEXT: [[TMP1:%.*]] = load <2 x i64>, <2 x i64>* bitcast ([4 x i64]* @src64 to <2 x i64>*), align 8 45 ; XOP-NEXT: [[TMP2:%.*]] = call <2 x i64> @llvm.bitreverse.v2i64(<2 x i64> [[TMP1]]) 46 ; XOP-NEXT: store <2 x i64> [[TMP2]], <2 x i64>* bitcast ([4 x i64]* @dst64 to <2 x i64>*), align 8 47 ; XOP-NEXT: ret void 48 ; 49 %ld0 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 0), align 8 50 %ld1 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i32 0, i64 1), align 8 51 %bitreverse0 = call i64 @llvm.bitreverse.i64(i64 %ld0) 52 %bitreverse1 = call i64 @llvm.bitreverse.i64(i64 %ld1) 53 store i64 %bitreverse0, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 0), align 8 54 store i64 %bitreverse1, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i32 0, i64 1), align 8 55 ret void 56 } 57 58 define void @bitreverse_4i64() #0 { 59 ; SSE-LABEL: @bitreverse_4i64( 60 ; SSE-NEXT: [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 0), align 4 61 ; SSE-NEXT: [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 1), align 4 62 ; SSE-NEXT: [[LD2:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 2), align 4 63 ; SSE-NEXT: [[LD3:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 3), align 4 64 ; SSE-NEXT: [[BITREVERSE0:%.*]] = call i64 @llvm.bitreverse.i64(i64 [[LD0]]) 65 ; SSE-NEXT: [[BITREVERSE1:%.*]] = call i64 @llvm.bitreverse.i64(i64 [[LD1]]) 66 ; SSE-NEXT: [[BITREVERSE2:%.*]] = call i64 @llvm.bitreverse.i64(i64 [[LD2]]) 67 ; SSE-NEXT: [[BITREVERSE3:%.*]] = call i64 @llvm.bitreverse.i64(i64 [[LD3]]) 68 ; SSE-NEXT: store i64 [[BITREVERSE0]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 0), align 4 69 ; SSE-NEXT: store i64 [[BITREVERSE1]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 1), align 4 70 ; SSE-NEXT: store i64 [[BITREVERSE2]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 2), align 4 71 ; SSE-NEXT: store i64 [[BITREVERSE3]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 3), align 4 72 ; SSE-NEXT: ret void 73 ; 74 ; AVX1-LABEL: @bitreverse_4i64( 75 ; AVX1-NEXT: [[LD0:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 0), align 4 76 ; AVX1-NEXT: [[LD1:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 1), align 4 77 ; AVX1-NEXT: [[LD2:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 2), align 4 78 ; AVX1-NEXT: [[LD3:%.*]] = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 3), align 4 79 ; AVX1-NEXT: [[BITREVERSE0:%.*]] = call i64 @llvm.bitreverse.i64(i64 [[LD0]]) 80 ; AVX1-NEXT: [[BITREVERSE1:%.*]] = call i64 @llvm.bitreverse.i64(i64 [[LD1]]) 81 ; AVX1-NEXT: [[BITREVERSE2:%.*]] = call i64 @llvm.bitreverse.i64(i64 [[LD2]]) 82 ; AVX1-NEXT: [[BITREVERSE3:%.*]] = call i64 @llvm.bitreverse.i64(i64 [[LD3]]) 83 ; AVX1-NEXT: store i64 [[BITREVERSE0]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 0), align 4 84 ; AVX1-NEXT: store i64 [[BITREVERSE1]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 1), align 4 85 ; AVX1-NEXT: store i64 [[BITREVERSE2]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 2), align 4 86 ; AVX1-NEXT: store i64 [[BITREVERSE3]], i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 3), align 4 87 ; AVX1-NEXT: ret void 88 ; 89 ; AVX2-LABEL: @bitreverse_4i64( 90 ; AVX2-NEXT: [[TMP1:%.*]] = load <4 x i64>, <4 x i64>* bitcast ([4 x i64]* @src64 to <4 x i64>*), align 4 91 ; AVX2-NEXT: [[TMP2:%.*]] = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> [[TMP1]]) 92 ; AVX2-NEXT: store <4 x i64> [[TMP2]], <4 x i64>* bitcast ([4 x i64]* @dst64 to <4 x i64>*), align 4 93 ; AVX2-NEXT: ret void 94 ; 95 ; XOP-LABEL: @bitreverse_4i64( 96 ; XOP-NEXT: [[TMP1:%.*]] = load <4 x i64>, <4 x i64>* bitcast ([4 x i64]* @src64 to <4 x i64>*), align 4 97 ; XOP-NEXT: [[TMP2:%.*]] = call <4 x i64> @llvm.bitreverse.v4i64(<4 x i64> [[TMP1]]) 98 ; XOP-NEXT: store <4 x i64> [[TMP2]], <4 x i64>* bitcast ([4 x i64]* @dst64 to <4 x i64>*), align 4 99 ; XOP-NEXT: ret void 100 ; 101 %ld0 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 0), align 4 102 %ld1 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 1), align 4 103 %ld2 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 2), align 4 104 %ld3 = load i64, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @src64, i64 0, i64 3), align 4 105 %bitreverse0 = call i64 @llvm.bitreverse.i64(i64 %ld0) 106 %bitreverse1 = call i64 @llvm.bitreverse.i64(i64 %ld1) 107 %bitreverse2 = call i64 @llvm.bitreverse.i64(i64 %ld2) 108 %bitreverse3 = call i64 @llvm.bitreverse.i64(i64 %ld3) 109 store i64 %bitreverse0, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 0), align 4 110 store i64 %bitreverse1, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 1), align 4 111 store i64 %bitreverse2, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 2), align 4 112 store i64 %bitreverse3, i64* getelementptr inbounds ([4 x i64], [4 x i64]* @dst64, i64 0, i64 3), align 4 113 ret void 114 } 115 116 define void @bitreverse_4i32() #0 { 117 ; SSE-LABEL: @bitreverse_4i32( 118 ; SSE-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 4 119 ; SSE-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 4 120 ; SSE-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 4 121 ; SSE-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 4 122 ; SSE-NEXT: [[BITREVERSE0:%.*]] = call i32 @llvm.bitreverse.i32(i32 [[LD0]]) 123 ; SSE-NEXT: [[BITREVERSE1:%.*]] = call i32 @llvm.bitreverse.i32(i32 [[LD1]]) 124 ; SSE-NEXT: [[BITREVERSE2:%.*]] = call i32 @llvm.bitreverse.i32(i32 [[LD2]]) 125 ; SSE-NEXT: [[BITREVERSE3:%.*]] = call i32 @llvm.bitreverse.i32(i32 [[LD3]]) 126 ; SSE-NEXT: store i32 [[BITREVERSE0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 4 127 ; SSE-NEXT: store i32 [[BITREVERSE1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 4 128 ; SSE-NEXT: store i32 [[BITREVERSE2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 4 129 ; SSE-NEXT: store i32 [[BITREVERSE3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 4 130 ; SSE-NEXT: ret void 131 ; 132 ; AVX-LABEL: @bitreverse_4i32( 133 ; AVX-NEXT: [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([8 x i32]* @src32 to <4 x i32>*), align 4 134 ; AVX-NEXT: [[TMP2:%.*]] = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> [[TMP1]]) 135 ; AVX-NEXT: store <4 x i32> [[TMP2]], <4 x i32>* bitcast ([8 x i32]* @dst32 to <4 x i32>*), align 4 136 ; AVX-NEXT: ret void 137 ; 138 ; XOP-LABEL: @bitreverse_4i32( 139 ; XOP-NEXT: [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([8 x i32]* @src32 to <4 x i32>*), align 4 140 ; XOP-NEXT: [[TMP2:%.*]] = call <4 x i32> @llvm.bitreverse.v4i32(<4 x i32> [[TMP1]]) 141 ; XOP-NEXT: store <4 x i32> [[TMP2]], <4 x i32>* bitcast ([8 x i32]* @dst32 to <4 x i32>*), align 4 142 ; XOP-NEXT: ret void 143 ; 144 %ld0 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 4 145 %ld1 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 4 146 %ld2 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 4 147 %ld3 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 4 148 %bitreverse0 = call i32 @llvm.bitreverse.i32(i32 %ld0) 149 %bitreverse1 = call i32 @llvm.bitreverse.i32(i32 %ld1) 150 %bitreverse2 = call i32 @llvm.bitreverse.i32(i32 %ld2) 151 %bitreverse3 = call i32 @llvm.bitreverse.i32(i32 %ld3) 152 store i32 %bitreverse0, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 4 153 store i32 %bitreverse1, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 4 154 store i32 %bitreverse2, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 4 155 store i32 %bitreverse3, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 4 156 ret void 157 } 158 159 define void @bitreverse_8i32() #0 { 160 ; SSE-LABEL: @bitreverse_8i32( 161 ; SSE-NEXT: [[LD0:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2 162 ; SSE-NEXT: [[LD1:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2 163 ; SSE-NEXT: [[LD2:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2 164 ; SSE-NEXT: [[LD3:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2 165 ; SSE-NEXT: [[LD4:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2 166 ; SSE-NEXT: [[LD5:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2 167 ; SSE-NEXT: [[LD6:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2 168 ; SSE-NEXT: [[LD7:%.*]] = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2 169 ; SSE-NEXT: [[BITREVERSE0:%.*]] = call i32 @llvm.bitreverse.i32(i32 [[LD0]]) 170 ; SSE-NEXT: [[BITREVERSE1:%.*]] = call i32 @llvm.bitreverse.i32(i32 [[LD1]]) 171 ; SSE-NEXT: [[BITREVERSE2:%.*]] = call i32 @llvm.bitreverse.i32(i32 [[LD2]]) 172 ; SSE-NEXT: [[BITREVERSE3:%.*]] = call i32 @llvm.bitreverse.i32(i32 [[LD3]]) 173 ; SSE-NEXT: [[BITREVERSE4:%.*]] = call i32 @llvm.bitreverse.i32(i32 [[LD4]]) 174 ; SSE-NEXT: [[BITREVERSE5:%.*]] = call i32 @llvm.bitreverse.i32(i32 [[LD5]]) 175 ; SSE-NEXT: [[BITREVERSE6:%.*]] = call i32 @llvm.bitreverse.i32(i32 [[LD6]]) 176 ; SSE-NEXT: [[BITREVERSE7:%.*]] = call i32 @llvm.bitreverse.i32(i32 [[LD7]]) 177 ; SSE-NEXT: store i32 [[BITREVERSE0]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2 178 ; SSE-NEXT: store i32 [[BITREVERSE1]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2 179 ; SSE-NEXT: store i32 [[BITREVERSE2]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2 180 ; SSE-NEXT: store i32 [[BITREVERSE3]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2 181 ; SSE-NEXT: store i32 [[BITREVERSE4]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2 182 ; SSE-NEXT: store i32 [[BITREVERSE5]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2 183 ; SSE-NEXT: store i32 [[BITREVERSE6]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2 184 ; SSE-NEXT: store i32 [[BITREVERSE7]], i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2 185 ; SSE-NEXT: ret void 186 ; 187 ; AVX-LABEL: @bitreverse_8i32( 188 ; AVX-NEXT: [[TMP1:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([8 x i32]* @src32 to <8 x i32>*), align 2 189 ; AVX-NEXT: [[TMP2:%.*]] = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> [[TMP1]]) 190 ; AVX-NEXT: store <8 x i32> [[TMP2]], <8 x i32>* bitcast ([8 x i32]* @dst32 to <8 x i32>*), align 2 191 ; AVX-NEXT: ret void 192 ; 193 ; XOP-LABEL: @bitreverse_8i32( 194 ; XOP-NEXT: [[TMP1:%.*]] = load <8 x i32>, <8 x i32>* bitcast ([8 x i32]* @src32 to <8 x i32>*), align 2 195 ; XOP-NEXT: [[TMP2:%.*]] = call <8 x i32> @llvm.bitreverse.v8i32(<8 x i32> [[TMP1]]) 196 ; XOP-NEXT: store <8 x i32> [[TMP2]], <8 x i32>* bitcast ([8 x i32]* @dst32 to <8 x i32>*), align 2 197 ; XOP-NEXT: ret void 198 ; 199 %ld0 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 0), align 2 200 %ld1 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 1), align 2 201 %ld2 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 2), align 2 202 %ld3 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 3), align 2 203 %ld4 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 4), align 2 204 %ld5 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 5), align 2 205 %ld6 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 6), align 2 206 %ld7 = load i32, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @src32, i32 0, i64 7), align 2 207 %bitreverse0 = call i32 @llvm.bitreverse.i32(i32 %ld0) 208 %bitreverse1 = call i32 @llvm.bitreverse.i32(i32 %ld1) 209 %bitreverse2 = call i32 @llvm.bitreverse.i32(i32 %ld2) 210 %bitreverse3 = call i32 @llvm.bitreverse.i32(i32 %ld3) 211 %bitreverse4 = call i32 @llvm.bitreverse.i32(i32 %ld4) 212 %bitreverse5 = call i32 @llvm.bitreverse.i32(i32 %ld5) 213 %bitreverse6 = call i32 @llvm.bitreverse.i32(i32 %ld6) 214 %bitreverse7 = call i32 @llvm.bitreverse.i32(i32 %ld7) 215 store i32 %bitreverse0, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 0), align 2 216 store i32 %bitreverse1, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 1), align 2 217 store i32 %bitreverse2, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 2), align 2 218 store i32 %bitreverse3, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 3), align 2 219 store i32 %bitreverse4, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 4), align 2 220 store i32 %bitreverse5, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 5), align 2 221 store i32 %bitreverse6, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 6), align 2 222 store i32 %bitreverse7, i32* getelementptr inbounds ([8 x i32], [8 x i32]* @dst32, i32 0, i64 7), align 2 223 ret void 224 } 225 226 define void @bitreverse_8i16() #0 { 227 ; SSE-LABEL: @bitreverse_8i16( 228 ; SSE-NEXT: [[LD0:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2 229 ; SSE-NEXT: [[LD1:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2 230 ; SSE-NEXT: [[LD2:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2 231 ; SSE-NEXT: [[LD3:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2 232 ; SSE-NEXT: [[LD4:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2 233 ; SSE-NEXT: [[LD5:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2 234 ; SSE-NEXT: [[LD6:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2 235 ; SSE-NEXT: [[LD7:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2 236 ; SSE-NEXT: [[BITREVERSE0:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD0]]) 237 ; SSE-NEXT: [[BITREVERSE1:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD1]]) 238 ; SSE-NEXT: [[BITREVERSE2:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD2]]) 239 ; SSE-NEXT: [[BITREVERSE3:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD3]]) 240 ; SSE-NEXT: [[BITREVERSE4:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD4]]) 241 ; SSE-NEXT: [[BITREVERSE5:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD5]]) 242 ; SSE-NEXT: [[BITREVERSE6:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD6]]) 243 ; SSE-NEXT: [[BITREVERSE7:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD7]]) 244 ; SSE-NEXT: store i16 [[BITREVERSE0]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2 245 ; SSE-NEXT: store i16 [[BITREVERSE1]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2 246 ; SSE-NEXT: store i16 [[BITREVERSE2]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2 247 ; SSE-NEXT: store i16 [[BITREVERSE3]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2 248 ; SSE-NEXT: store i16 [[BITREVERSE4]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2 249 ; SSE-NEXT: store i16 [[BITREVERSE5]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2 250 ; SSE-NEXT: store i16 [[BITREVERSE6]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2 251 ; SSE-NEXT: store i16 [[BITREVERSE7]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2 252 ; SSE-NEXT: ret void 253 ; 254 ; AVX-LABEL: @bitreverse_8i16( 255 ; AVX-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2 256 ; AVX-NEXT: [[TMP2:%.*]] = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> [[TMP1]]) 257 ; AVX-NEXT: store <8 x i16> [[TMP2]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2 258 ; AVX-NEXT: ret void 259 ; 260 ; XOP-LABEL: @bitreverse_8i16( 261 ; XOP-NEXT: [[TMP1:%.*]] = load <8 x i16>, <8 x i16>* bitcast ([16 x i16]* @src16 to <8 x i16>*), align 2 262 ; XOP-NEXT: [[TMP2:%.*]] = call <8 x i16> @llvm.bitreverse.v8i16(<8 x i16> [[TMP1]]) 263 ; XOP-NEXT: store <8 x i16> [[TMP2]], <8 x i16>* bitcast ([16 x i16]* @dst16 to <8 x i16>*), align 2 264 ; XOP-NEXT: ret void 265 ; 266 %ld0 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2 267 %ld1 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2 268 %ld2 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2 269 %ld3 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2 270 %ld4 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2 271 %ld5 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2 272 %ld6 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2 273 %ld7 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2 274 %bitreverse0 = call i16 @llvm.bitreverse.i16(i16 %ld0) 275 %bitreverse1 = call i16 @llvm.bitreverse.i16(i16 %ld1) 276 %bitreverse2 = call i16 @llvm.bitreverse.i16(i16 %ld2) 277 %bitreverse3 = call i16 @llvm.bitreverse.i16(i16 %ld3) 278 %bitreverse4 = call i16 @llvm.bitreverse.i16(i16 %ld4) 279 %bitreverse5 = call i16 @llvm.bitreverse.i16(i16 %ld5) 280 %bitreverse6 = call i16 @llvm.bitreverse.i16(i16 %ld6) 281 %bitreverse7 = call i16 @llvm.bitreverse.i16(i16 %ld7) 282 store i16 %bitreverse0, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2 283 store i16 %bitreverse1, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2 284 store i16 %bitreverse2, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2 285 store i16 %bitreverse3, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2 286 store i16 %bitreverse4, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2 287 store i16 %bitreverse5, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2 288 store i16 %bitreverse6, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2 289 store i16 %bitreverse7, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2 290 ret void 291 } 292 293 define void @bitreverse_16i16() #0 { 294 ; SSE-LABEL: @bitreverse_16i16( 295 ; SSE-NEXT: [[LD0:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2 296 ; SSE-NEXT: [[LD1:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2 297 ; SSE-NEXT: [[LD2:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2 298 ; SSE-NEXT: [[LD3:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2 299 ; SSE-NEXT: [[LD4:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2 300 ; SSE-NEXT: [[LD5:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2 301 ; SSE-NEXT: [[LD6:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2 302 ; SSE-NEXT: [[LD7:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2 303 ; SSE-NEXT: [[LD8:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8), align 2 304 ; SSE-NEXT: [[LD9:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 9), align 2 305 ; SSE-NEXT: [[LD10:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 10), align 2 306 ; SSE-NEXT: [[LD11:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 11), align 2 307 ; SSE-NEXT: [[LD12:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 12), align 2 308 ; SSE-NEXT: [[LD13:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 13), align 2 309 ; SSE-NEXT: [[LD14:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 14), align 2 310 ; SSE-NEXT: [[LD15:%.*]] = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 15), align 2 311 ; SSE-NEXT: [[BITREVERSE0:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD0]]) 312 ; SSE-NEXT: [[BITREVERSE1:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD1]]) 313 ; SSE-NEXT: [[BITREVERSE2:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD2]]) 314 ; SSE-NEXT: [[BITREVERSE3:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD3]]) 315 ; SSE-NEXT: [[BITREVERSE4:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD4]]) 316 ; SSE-NEXT: [[BITREVERSE5:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD5]]) 317 ; SSE-NEXT: [[BITREVERSE6:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD6]]) 318 ; SSE-NEXT: [[BITREVERSE7:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD7]]) 319 ; SSE-NEXT: [[BITREVERSE8:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD8]]) 320 ; SSE-NEXT: [[BITREVERSE9:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD9]]) 321 ; SSE-NEXT: [[BITREVERSE10:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD10]]) 322 ; SSE-NEXT: [[BITREVERSE11:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD11]]) 323 ; SSE-NEXT: [[BITREVERSE12:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD12]]) 324 ; SSE-NEXT: [[BITREVERSE13:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD13]]) 325 ; SSE-NEXT: [[BITREVERSE14:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD14]]) 326 ; SSE-NEXT: [[BITREVERSE15:%.*]] = call i16 @llvm.bitreverse.i16(i16 [[LD15]]) 327 ; SSE-NEXT: store i16 [[BITREVERSE0]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2 328 ; SSE-NEXT: store i16 [[BITREVERSE1]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2 329 ; SSE-NEXT: store i16 [[BITREVERSE2]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2 330 ; SSE-NEXT: store i16 [[BITREVERSE3]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2 331 ; SSE-NEXT: store i16 [[BITREVERSE4]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2 332 ; SSE-NEXT: store i16 [[BITREVERSE5]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2 333 ; SSE-NEXT: store i16 [[BITREVERSE6]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2 334 ; SSE-NEXT: store i16 [[BITREVERSE7]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2 335 ; SSE-NEXT: store i16 [[BITREVERSE8]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8), align 2 336 ; SSE-NEXT: store i16 [[BITREVERSE9]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 9), align 2 337 ; SSE-NEXT: store i16 [[BITREVERSE10]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 10), align 2 338 ; SSE-NEXT: store i16 [[BITREVERSE11]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 11), align 2 339 ; SSE-NEXT: store i16 [[BITREVERSE12]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 12), align 2 340 ; SSE-NEXT: store i16 [[BITREVERSE13]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 13), align 2 341 ; SSE-NEXT: store i16 [[BITREVERSE14]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 14), align 2 342 ; SSE-NEXT: store i16 [[BITREVERSE15]], i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 15), align 2 343 ; SSE-NEXT: ret void 344 ; 345 ; AVX-LABEL: @bitreverse_16i16( 346 ; AVX-NEXT: [[TMP1:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([16 x i16]* @src16 to <16 x i16>*), align 2 347 ; AVX-NEXT: [[TMP2:%.*]] = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> [[TMP1]]) 348 ; AVX-NEXT: store <16 x i16> [[TMP2]], <16 x i16>* bitcast ([16 x i16]* @dst16 to <16 x i16>*), align 2 349 ; AVX-NEXT: ret void 350 ; 351 ; XOP-LABEL: @bitreverse_16i16( 352 ; XOP-NEXT: [[TMP1:%.*]] = load <16 x i16>, <16 x i16>* bitcast ([16 x i16]* @src16 to <16 x i16>*), align 2 353 ; XOP-NEXT: [[TMP2:%.*]] = call <16 x i16> @llvm.bitreverse.v16i16(<16 x i16> [[TMP1]]) 354 ; XOP-NEXT: store <16 x i16> [[TMP2]], <16 x i16>* bitcast ([16 x i16]* @dst16 to <16 x i16>*), align 2 355 ; XOP-NEXT: ret void 356 ; 357 %ld0 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 0), align 2 358 %ld1 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 1), align 2 359 %ld2 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 2), align 2 360 %ld3 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 3), align 2 361 %ld4 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 4), align 2 362 %ld5 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 5), align 2 363 %ld6 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 6), align 2 364 %ld7 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 7), align 2 365 %ld8 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 8), align 2 366 %ld9 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 9), align 2 367 %ld10 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 10), align 2 368 %ld11 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 11), align 2 369 %ld12 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 12), align 2 370 %ld13 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 13), align 2 371 %ld14 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 14), align 2 372 %ld15 = load i16, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @src16, i16 0, i64 15), align 2 373 %bitreverse0 = call i16 @llvm.bitreverse.i16(i16 %ld0) 374 %bitreverse1 = call i16 @llvm.bitreverse.i16(i16 %ld1) 375 %bitreverse2 = call i16 @llvm.bitreverse.i16(i16 %ld2) 376 %bitreverse3 = call i16 @llvm.bitreverse.i16(i16 %ld3) 377 %bitreverse4 = call i16 @llvm.bitreverse.i16(i16 %ld4) 378 %bitreverse5 = call i16 @llvm.bitreverse.i16(i16 %ld5) 379 %bitreverse6 = call i16 @llvm.bitreverse.i16(i16 %ld6) 380 %bitreverse7 = call i16 @llvm.bitreverse.i16(i16 %ld7) 381 %bitreverse8 = call i16 @llvm.bitreverse.i16(i16 %ld8) 382 %bitreverse9 = call i16 @llvm.bitreverse.i16(i16 %ld9) 383 %bitreverse10 = call i16 @llvm.bitreverse.i16(i16 %ld10) 384 %bitreverse11 = call i16 @llvm.bitreverse.i16(i16 %ld11) 385 %bitreverse12 = call i16 @llvm.bitreverse.i16(i16 %ld12) 386 %bitreverse13 = call i16 @llvm.bitreverse.i16(i16 %ld13) 387 %bitreverse14 = call i16 @llvm.bitreverse.i16(i16 %ld14) 388 %bitreverse15 = call i16 @llvm.bitreverse.i16(i16 %ld15) 389 store i16 %bitreverse0 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 0), align 2 390 store i16 %bitreverse1 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 1), align 2 391 store i16 %bitreverse2 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 2), align 2 392 store i16 %bitreverse3 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 3), align 2 393 store i16 %bitreverse4 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 4), align 2 394 store i16 %bitreverse5 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 5), align 2 395 store i16 %bitreverse6 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 6), align 2 396 store i16 %bitreverse7 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 7), align 2 397 store i16 %bitreverse8 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 8), align 2 398 store i16 %bitreverse9 , i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 9), align 2 399 store i16 %bitreverse10, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 10), align 2 400 store i16 %bitreverse11, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 11), align 2 401 store i16 %bitreverse12, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 12), align 2 402 store i16 %bitreverse13, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 13), align 2 403 store i16 %bitreverse14, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 14), align 2 404 store i16 %bitreverse15, i16* getelementptr inbounds ([16 x i16], [16 x i16]* @dst16, i16 0, i64 15), align 2 405 ret void 406 } 407 408 define void @bitreverse_16i8() #0 { 409 ; SSE-LABEL: @bitreverse_16i8( 410 ; SSE-NEXT: [[LD0:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1 411 ; SSE-NEXT: [[LD1:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1 412 ; SSE-NEXT: [[LD2:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1 413 ; SSE-NEXT: [[LD3:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1 414 ; SSE-NEXT: [[LD4:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1 415 ; SSE-NEXT: [[LD5:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1 416 ; SSE-NEXT: [[LD6:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1 417 ; SSE-NEXT: [[LD7:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1 418 ; SSE-NEXT: [[LD8:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1 419 ; SSE-NEXT: [[LD9:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1 420 ; SSE-NEXT: [[LD10:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1 421 ; SSE-NEXT: [[LD11:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1 422 ; SSE-NEXT: [[LD12:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1 423 ; SSE-NEXT: [[LD13:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1 424 ; SSE-NEXT: [[LD14:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1 425 ; SSE-NEXT: [[LD15:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1 426 ; SSE-NEXT: [[BITREVERSE0:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD0]]) 427 ; SSE-NEXT: [[BITREVERSE1:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD1]]) 428 ; SSE-NEXT: [[BITREVERSE2:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD2]]) 429 ; SSE-NEXT: [[BITREVERSE3:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD3]]) 430 ; SSE-NEXT: [[BITREVERSE4:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD4]]) 431 ; SSE-NEXT: [[BITREVERSE5:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD5]]) 432 ; SSE-NEXT: [[BITREVERSE6:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD6]]) 433 ; SSE-NEXT: [[BITREVERSE7:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD7]]) 434 ; SSE-NEXT: [[BITREVERSE8:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD8]]) 435 ; SSE-NEXT: [[BITREVERSE9:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD9]]) 436 ; SSE-NEXT: [[BITREVERSE10:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD10]]) 437 ; SSE-NEXT: [[BITREVERSE11:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD11]]) 438 ; SSE-NEXT: [[BITREVERSE12:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD12]]) 439 ; SSE-NEXT: [[BITREVERSE13:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD13]]) 440 ; SSE-NEXT: [[BITREVERSE14:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD14]]) 441 ; SSE-NEXT: [[BITREVERSE15:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD15]]) 442 ; SSE-NEXT: store i8 [[BITREVERSE0]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1 443 ; SSE-NEXT: store i8 [[BITREVERSE1]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1 444 ; SSE-NEXT: store i8 [[BITREVERSE2]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1 445 ; SSE-NEXT: store i8 [[BITREVERSE3]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1 446 ; SSE-NEXT: store i8 [[BITREVERSE4]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1 447 ; SSE-NEXT: store i8 [[BITREVERSE5]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1 448 ; SSE-NEXT: store i8 [[BITREVERSE6]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1 449 ; SSE-NEXT: store i8 [[BITREVERSE7]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1 450 ; SSE-NEXT: store i8 [[BITREVERSE8]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1 451 ; SSE-NEXT: store i8 [[BITREVERSE9]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1 452 ; SSE-NEXT: store i8 [[BITREVERSE10]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1 453 ; SSE-NEXT: store i8 [[BITREVERSE11]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1 454 ; SSE-NEXT: store i8 [[BITREVERSE12]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1 455 ; SSE-NEXT: store i8 [[BITREVERSE13]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1 456 ; SSE-NEXT: store i8 [[BITREVERSE14]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1 457 ; SSE-NEXT: store i8 [[BITREVERSE15]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1 458 ; SSE-NEXT: ret void 459 ; 460 ; AVX-LABEL: @bitreverse_16i8( 461 ; AVX-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1 462 ; AVX-NEXT: [[TMP2:%.*]] = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> [[TMP1]]) 463 ; AVX-NEXT: store <16 x i8> [[TMP2]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1 464 ; AVX-NEXT: ret void 465 ; 466 ; XOP-LABEL: @bitreverse_16i8( 467 ; XOP-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1 468 ; XOP-NEXT: [[TMP2:%.*]] = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> [[TMP1]]) 469 ; XOP-NEXT: store <16 x i8> [[TMP2]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1 470 ; XOP-NEXT: ret void 471 ; 472 %ld0 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1 473 %ld1 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1 474 %ld2 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1 475 %ld3 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1 476 %ld4 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1 477 %ld5 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1 478 %ld6 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1 479 %ld7 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1 480 %ld8 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1 481 %ld9 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1 482 %ld10 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1 483 %ld11 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1 484 %ld12 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1 485 %ld13 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1 486 %ld14 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1 487 %ld15 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1 488 %bitreverse0 = call i8 @llvm.bitreverse.i8(i8 %ld0) 489 %bitreverse1 = call i8 @llvm.bitreverse.i8(i8 %ld1) 490 %bitreverse2 = call i8 @llvm.bitreverse.i8(i8 %ld2) 491 %bitreverse3 = call i8 @llvm.bitreverse.i8(i8 %ld3) 492 %bitreverse4 = call i8 @llvm.bitreverse.i8(i8 %ld4) 493 %bitreverse5 = call i8 @llvm.bitreverse.i8(i8 %ld5) 494 %bitreverse6 = call i8 @llvm.bitreverse.i8(i8 %ld6) 495 %bitreverse7 = call i8 @llvm.bitreverse.i8(i8 %ld7) 496 %bitreverse8 = call i8 @llvm.bitreverse.i8(i8 %ld8) 497 %bitreverse9 = call i8 @llvm.bitreverse.i8(i8 %ld9) 498 %bitreverse10 = call i8 @llvm.bitreverse.i8(i8 %ld10) 499 %bitreverse11 = call i8 @llvm.bitreverse.i8(i8 %ld11) 500 %bitreverse12 = call i8 @llvm.bitreverse.i8(i8 %ld12) 501 %bitreverse13 = call i8 @llvm.bitreverse.i8(i8 %ld13) 502 %bitreverse14 = call i8 @llvm.bitreverse.i8(i8 %ld14) 503 %bitreverse15 = call i8 @llvm.bitreverse.i8(i8 %ld15) 504 store i8 %bitreverse0 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1 505 store i8 %bitreverse1 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1 506 store i8 %bitreverse2 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1 507 store i8 %bitreverse3 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1 508 store i8 %bitreverse4 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1 509 store i8 %bitreverse5 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1 510 store i8 %bitreverse6 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1 511 store i8 %bitreverse7 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1 512 store i8 %bitreverse8 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1 513 store i8 %bitreverse9 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1 514 store i8 %bitreverse10, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1 515 store i8 %bitreverse11, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1 516 store i8 %bitreverse12, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1 517 store i8 %bitreverse13, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1 518 store i8 %bitreverse14, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1 519 store i8 %bitreverse15, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1 520 ret void 521 } 522 523 define void @bitreverse_32i8() #0 { 524 ; SSE-LABEL: @bitreverse_32i8( 525 ; SSE-NEXT: [[LD0:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1 526 ; SSE-NEXT: [[LD1:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1 527 ; SSE-NEXT: [[LD2:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1 528 ; SSE-NEXT: [[LD3:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1 529 ; SSE-NEXT: [[LD4:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1 530 ; SSE-NEXT: [[LD5:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1 531 ; SSE-NEXT: [[LD6:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1 532 ; SSE-NEXT: [[LD7:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1 533 ; SSE-NEXT: [[LD8:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1 534 ; SSE-NEXT: [[LD9:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1 535 ; SSE-NEXT: [[LD10:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1 536 ; SSE-NEXT: [[LD11:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1 537 ; SSE-NEXT: [[LD12:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1 538 ; SSE-NEXT: [[LD13:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1 539 ; SSE-NEXT: [[LD14:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1 540 ; SSE-NEXT: [[LD15:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1 541 ; SSE-NEXT: [[LD16:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16), align 1 542 ; SSE-NEXT: [[LD17:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 17), align 1 543 ; SSE-NEXT: [[LD18:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 18), align 1 544 ; SSE-NEXT: [[LD19:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 19), align 1 545 ; SSE-NEXT: [[LD20:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 20), align 1 546 ; SSE-NEXT: [[LD21:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 21), align 1 547 ; SSE-NEXT: [[LD22:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 22), align 1 548 ; SSE-NEXT: [[LD23:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 23), align 1 549 ; SSE-NEXT: [[LD24:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 24), align 1 550 ; SSE-NEXT: [[LD25:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 25), align 1 551 ; SSE-NEXT: [[LD26:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 26), align 1 552 ; SSE-NEXT: [[LD27:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 27), align 1 553 ; SSE-NEXT: [[LD28:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 28), align 1 554 ; SSE-NEXT: [[LD29:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 29), align 1 555 ; SSE-NEXT: [[LD30:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 30), align 1 556 ; SSE-NEXT: [[LD31:%.*]] = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 31), align 1 557 ; SSE-NEXT: [[BITREVERSE0:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD0]]) 558 ; SSE-NEXT: [[BITREVERSE1:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD1]]) 559 ; SSE-NEXT: [[BITREVERSE2:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD2]]) 560 ; SSE-NEXT: [[BITREVERSE3:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD3]]) 561 ; SSE-NEXT: [[BITREVERSE4:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD4]]) 562 ; SSE-NEXT: [[BITREVERSE5:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD5]]) 563 ; SSE-NEXT: [[BITREVERSE6:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD6]]) 564 ; SSE-NEXT: [[BITREVERSE7:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD7]]) 565 ; SSE-NEXT: [[BITREVERSE8:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD8]]) 566 ; SSE-NEXT: [[BITREVERSE9:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD9]]) 567 ; SSE-NEXT: [[BITREVERSE10:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD10]]) 568 ; SSE-NEXT: [[BITREVERSE11:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD11]]) 569 ; SSE-NEXT: [[BITREVERSE12:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD12]]) 570 ; SSE-NEXT: [[BITREVERSE13:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD13]]) 571 ; SSE-NEXT: [[BITREVERSE14:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD14]]) 572 ; SSE-NEXT: [[BITREVERSE15:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD15]]) 573 ; SSE-NEXT: [[BITREVERSE16:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD16]]) 574 ; SSE-NEXT: [[BITREVERSE17:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD17]]) 575 ; SSE-NEXT: [[BITREVERSE18:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD18]]) 576 ; SSE-NEXT: [[BITREVERSE19:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD19]]) 577 ; SSE-NEXT: [[BITREVERSE20:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD20]]) 578 ; SSE-NEXT: [[BITREVERSE21:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD21]]) 579 ; SSE-NEXT: [[BITREVERSE22:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD22]]) 580 ; SSE-NEXT: [[BITREVERSE23:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD23]]) 581 ; SSE-NEXT: [[BITREVERSE24:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD24]]) 582 ; SSE-NEXT: [[BITREVERSE25:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD25]]) 583 ; SSE-NEXT: [[BITREVERSE26:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD26]]) 584 ; SSE-NEXT: [[BITREVERSE27:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD27]]) 585 ; SSE-NEXT: [[BITREVERSE28:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD28]]) 586 ; SSE-NEXT: [[BITREVERSE29:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD29]]) 587 ; SSE-NEXT: [[BITREVERSE30:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD30]]) 588 ; SSE-NEXT: [[BITREVERSE31:%.*]] = call i8 @llvm.bitreverse.i8(i8 [[LD31]]) 589 ; SSE-NEXT: store i8 [[BITREVERSE0]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1 590 ; SSE-NEXT: store i8 [[BITREVERSE1]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1 591 ; SSE-NEXT: store i8 [[BITREVERSE2]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1 592 ; SSE-NEXT: store i8 [[BITREVERSE3]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1 593 ; SSE-NEXT: store i8 [[BITREVERSE4]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1 594 ; SSE-NEXT: store i8 [[BITREVERSE5]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1 595 ; SSE-NEXT: store i8 [[BITREVERSE6]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1 596 ; SSE-NEXT: store i8 [[BITREVERSE7]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1 597 ; SSE-NEXT: store i8 [[BITREVERSE8]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1 598 ; SSE-NEXT: store i8 [[BITREVERSE9]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1 599 ; SSE-NEXT: store i8 [[BITREVERSE10]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1 600 ; SSE-NEXT: store i8 [[BITREVERSE11]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1 601 ; SSE-NEXT: store i8 [[BITREVERSE12]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1 602 ; SSE-NEXT: store i8 [[BITREVERSE13]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1 603 ; SSE-NEXT: store i8 [[BITREVERSE14]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1 604 ; SSE-NEXT: store i8 [[BITREVERSE15]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1 605 ; SSE-NEXT: store i8 [[BITREVERSE16]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16), align 1 606 ; SSE-NEXT: store i8 [[BITREVERSE17]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 17), align 1 607 ; SSE-NEXT: store i8 [[BITREVERSE18]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 18), align 1 608 ; SSE-NEXT: store i8 [[BITREVERSE19]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 19), align 1 609 ; SSE-NEXT: store i8 [[BITREVERSE20]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 20), align 1 610 ; SSE-NEXT: store i8 [[BITREVERSE21]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 21), align 1 611 ; SSE-NEXT: store i8 [[BITREVERSE22]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 22), align 1 612 ; SSE-NEXT: store i8 [[BITREVERSE23]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 23), align 1 613 ; SSE-NEXT: store i8 [[BITREVERSE24]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 24), align 1 614 ; SSE-NEXT: store i8 [[BITREVERSE25]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 25), align 1 615 ; SSE-NEXT: store i8 [[BITREVERSE26]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 26), align 1 616 ; SSE-NEXT: store i8 [[BITREVERSE27]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 27), align 1 617 ; SSE-NEXT: store i8 [[BITREVERSE28]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 28), align 1 618 ; SSE-NEXT: store i8 [[BITREVERSE29]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 29), align 1 619 ; SSE-NEXT: store i8 [[BITREVERSE30]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 30), align 1 620 ; SSE-NEXT: store i8 [[BITREVERSE31]], i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 31), align 1 621 ; SSE-NEXT: ret void 622 ; 623 ; AVX-LABEL: @bitreverse_32i8( 624 ; AVX-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1 625 ; AVX-NEXT: [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16) to <16 x i8>*), align 1 626 ; AVX-NEXT: [[TMP3:%.*]] = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> [[TMP1]]) 627 ; AVX-NEXT: [[TMP4:%.*]] = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> [[TMP2]]) 628 ; AVX-NEXT: store <16 x i8> [[TMP3]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1 629 ; AVX-NEXT: store <16 x i8> [[TMP4]], <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16) to <16 x i8>*), align 1 630 ; AVX-NEXT: ret void 631 ; 632 ; XOP-LABEL: @bitreverse_32i8( 633 ; XOP-NEXT: [[TMP1:%.*]] = load <16 x i8>, <16 x i8>* bitcast ([32 x i8]* @src8 to <16 x i8>*), align 1 634 ; XOP-NEXT: [[TMP2:%.*]] = load <16 x i8>, <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16) to <16 x i8>*), align 1 635 ; XOP-NEXT: [[TMP3:%.*]] = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> [[TMP1]]) 636 ; XOP-NEXT: [[TMP4:%.*]] = call <16 x i8> @llvm.bitreverse.v16i8(<16 x i8> [[TMP2]]) 637 ; XOP-NEXT: store <16 x i8> [[TMP3]], <16 x i8>* bitcast ([32 x i8]* @dst8 to <16 x i8>*), align 1 638 ; XOP-NEXT: store <16 x i8> [[TMP4]], <16 x i8>* bitcast (i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16) to <16 x i8>*), align 1 639 ; XOP-NEXT: ret void 640 ; 641 %ld0 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 0), align 1 642 %ld1 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 1), align 1 643 %ld2 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 2), align 1 644 %ld3 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 3), align 1 645 %ld4 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 4), align 1 646 %ld5 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 5), align 1 647 %ld6 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 6), align 1 648 %ld7 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 7), align 1 649 %ld8 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 8), align 1 650 %ld9 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 9), align 1 651 %ld10 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 10), align 1 652 %ld11 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 11), align 1 653 %ld12 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 12), align 1 654 %ld13 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 13), align 1 655 %ld14 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 14), align 1 656 %ld15 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 15), align 1 657 %ld16 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 16), align 1 658 %ld17 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 17), align 1 659 %ld18 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 18), align 1 660 %ld19 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 19), align 1 661 %ld20 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 20), align 1 662 %ld21 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 21), align 1 663 %ld22 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 22), align 1 664 %ld23 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 23), align 1 665 %ld24 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 24), align 1 666 %ld25 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 25), align 1 667 %ld26 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 26), align 1 668 %ld27 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 27), align 1 669 %ld28 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 28), align 1 670 %ld29 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 29), align 1 671 %ld30 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 30), align 1 672 %ld31 = load i8, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @src8, i8 0, i64 31), align 1 673 %bitreverse0 = call i8 @llvm.bitreverse.i8(i8 %ld0) 674 %bitreverse1 = call i8 @llvm.bitreverse.i8(i8 %ld1) 675 %bitreverse2 = call i8 @llvm.bitreverse.i8(i8 %ld2) 676 %bitreverse3 = call i8 @llvm.bitreverse.i8(i8 %ld3) 677 %bitreverse4 = call i8 @llvm.bitreverse.i8(i8 %ld4) 678 %bitreverse5 = call i8 @llvm.bitreverse.i8(i8 %ld5) 679 %bitreverse6 = call i8 @llvm.bitreverse.i8(i8 %ld6) 680 %bitreverse7 = call i8 @llvm.bitreverse.i8(i8 %ld7) 681 %bitreverse8 = call i8 @llvm.bitreverse.i8(i8 %ld8) 682 %bitreverse9 = call i8 @llvm.bitreverse.i8(i8 %ld9) 683 %bitreverse10 = call i8 @llvm.bitreverse.i8(i8 %ld10) 684 %bitreverse11 = call i8 @llvm.bitreverse.i8(i8 %ld11) 685 %bitreverse12 = call i8 @llvm.bitreverse.i8(i8 %ld12) 686 %bitreverse13 = call i8 @llvm.bitreverse.i8(i8 %ld13) 687 %bitreverse14 = call i8 @llvm.bitreverse.i8(i8 %ld14) 688 %bitreverse15 = call i8 @llvm.bitreverse.i8(i8 %ld15) 689 %bitreverse16 = call i8 @llvm.bitreverse.i8(i8 %ld16) 690 %bitreverse17 = call i8 @llvm.bitreverse.i8(i8 %ld17) 691 %bitreverse18 = call i8 @llvm.bitreverse.i8(i8 %ld18) 692 %bitreverse19 = call i8 @llvm.bitreverse.i8(i8 %ld19) 693 %bitreverse20 = call i8 @llvm.bitreverse.i8(i8 %ld20) 694 %bitreverse21 = call i8 @llvm.bitreverse.i8(i8 %ld21) 695 %bitreverse22 = call i8 @llvm.bitreverse.i8(i8 %ld22) 696 %bitreverse23 = call i8 @llvm.bitreverse.i8(i8 %ld23) 697 %bitreverse24 = call i8 @llvm.bitreverse.i8(i8 %ld24) 698 %bitreverse25 = call i8 @llvm.bitreverse.i8(i8 %ld25) 699 %bitreverse26 = call i8 @llvm.bitreverse.i8(i8 %ld26) 700 %bitreverse27 = call i8 @llvm.bitreverse.i8(i8 %ld27) 701 %bitreverse28 = call i8 @llvm.bitreverse.i8(i8 %ld28) 702 %bitreverse29 = call i8 @llvm.bitreverse.i8(i8 %ld29) 703 %bitreverse30 = call i8 @llvm.bitreverse.i8(i8 %ld30) 704 %bitreverse31 = call i8 @llvm.bitreverse.i8(i8 %ld31) 705 store i8 %bitreverse0 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 0), align 1 706 store i8 %bitreverse1 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 1), align 1 707 store i8 %bitreverse2 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 2), align 1 708 store i8 %bitreverse3 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 3), align 1 709 store i8 %bitreverse4 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 4), align 1 710 store i8 %bitreverse5 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 5), align 1 711 store i8 %bitreverse6 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 6), align 1 712 store i8 %bitreverse7 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 7), align 1 713 store i8 %bitreverse8 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 8), align 1 714 store i8 %bitreverse9 , i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 9), align 1 715 store i8 %bitreverse10, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 10), align 1 716 store i8 %bitreverse11, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 11), align 1 717 store i8 %bitreverse12, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 12), align 1 718 store i8 %bitreverse13, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 13), align 1 719 store i8 %bitreverse14, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 14), align 1 720 store i8 %bitreverse15, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 15), align 1 721 store i8 %bitreverse16, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 16), align 1 722 store i8 %bitreverse17, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 17), align 1 723 store i8 %bitreverse18, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 18), align 1 724 store i8 %bitreverse19, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 19), align 1 725 store i8 %bitreverse20, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 20), align 1 726 store i8 %bitreverse21, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 21), align 1 727 store i8 %bitreverse22, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 22), align 1 728 store i8 %bitreverse23, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 23), align 1 729 store i8 %bitreverse24, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 24), align 1 730 store i8 %bitreverse25, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 25), align 1 731 store i8 %bitreverse26, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 26), align 1 732 store i8 %bitreverse27, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 27), align 1 733 store i8 %bitreverse28, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 28), align 1 734 store i8 %bitreverse29, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 29), align 1 735 store i8 %bitreverse30, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 30), align 1 736 store i8 %bitreverse31, i8* getelementptr inbounds ([32 x i8], [32 x i8]* @dst8, i8 0, i64 31), align 1 737 ret void 738 } 739 740 attributes #0 = { nounwind } 741 742
