1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py 2; RUN: opt < %s -passes=vector-combine -S -mtriple=x86_64-- -mattr=SSE2 | FileCheck %s --check-prefixes=CHECK,SSE 3; RUN: opt < %s -passes=vector-combine -S -mtriple=x86_64-- -mattr=AVX2 | FileCheck %s --check-prefixes=CHECK,AVX 4 5; x86 does not have a cheap v16i8 shuffle until SSSE3 (pshufb) 6 7define <16 x i8> @bitcast_shuf_narrow_element(<4 x i32> %v) { 8; SSE-LABEL: @bitcast_shuf_narrow_element( 9; SSE-NEXT: [[SHUF:%.*]] = shufflevector <4 x i32> [[V:%.*]], <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 10; SSE-NEXT: [[R:%.*]] = bitcast <4 x i32> [[SHUF]] to <16 x i8> 11; SSE-NEXT: ret <16 x i8> [[R]] 12; 13; AVX-LABEL: @bitcast_shuf_narrow_element( 14; AVX-NEXT: [[TMP1:%.*]] = bitcast <4 x i32> [[V:%.*]] to <16 x i8> 15; AVX-NEXT: [[R:%.*]] = shufflevector <16 x i8> [[TMP1]], <16 x i8> poison, <16 x i32> <i32 12, i32 13, i32 14, i32 15, i32 8, i32 9, i32 10, i32 11, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3> 16; AVX-NEXT: ret <16 x i8> [[R]] 17; 18 %shuf = shufflevector <4 x i32> %v, <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 19 %r = bitcast <4 x i32> %shuf to <16 x i8> 20 ret <16 x i8> %r 21} 22 23; v4f32 is the same cost as v4i32, so this always works 24 25define <4 x float> @bitcast_shuf_same_size(<4 x i32> %v) { 26; CHECK-LABEL: @bitcast_shuf_same_size( 27; CHECK-NEXT: [[TMP1:%.*]] = bitcast <4 x i32> [[V:%.*]] to <4 x float> 28; CHECK-NEXT: [[R:%.*]] = shufflevector <4 x float> [[TMP1]], <4 x float> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 29; CHECK-NEXT: ret <4 x float> [[R]] 30; 31 %shuf = shufflevector <4 x i32> %v, <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 32 %r = bitcast <4 x i32> %shuf to <4 x float> 33 ret <4 x float> %r 34} 35 36; Length-changing shuffles 37 38define <16 x i8> @bitcast_shuf_narrow_element_subvector(<2 x i32> %v) { 39; SSE-LABEL: @bitcast_shuf_narrow_element_subvector( 40; SSE-NEXT: [[SHUF:%.*]] = shufflevector <2 x i32> [[V:%.*]], <2 x i32> poison, <4 x i32> <i32 1, i32 0, i32 1, i32 0> 41; SSE-NEXT: [[R:%.*]] = bitcast <4 x i32> [[SHUF]] to <16 x i8> 42; SSE-NEXT: ret <16 x i8> [[R]] 43; 44; AVX-LABEL: @bitcast_shuf_narrow_element_subvector( 45; AVX-NEXT: [[TMP1:%.*]] = bitcast <2 x i32> [[V:%.*]] to <8 x i8> 46; AVX-NEXT: [[R:%.*]] = shufflevector <8 x i8> [[TMP1]], <8 x i8> poison, <16 x i32> <i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3> 47; AVX-NEXT: ret <16 x i8> [[R]] 48; 49 %shuf = shufflevector <2 x i32> %v, <2 x i32> poison, <4 x i32> <i32 1, i32 0, i32 1, i32 0> 50 %r = bitcast <4 x i32> %shuf to <16 x i8> 51 ret <16 x i8> %r 52} 53 54define <16 x i16> @bitcast_shuf_narrow_element_concat_subvectors(<2 x i64> %v) { 55; SSE-LABEL: @bitcast_shuf_narrow_element_concat_subvectors( 56; SSE-NEXT: [[SHUF:%.*]] = shufflevector <2 x i64> [[V:%.*]], <2 x i64> poison, <4 x i32> <i32 0, i32 1, i32 0, i32 1> 57; SSE-NEXT: [[R:%.*]] = bitcast <4 x i64> [[SHUF]] to <16 x i16> 58; SSE-NEXT: ret <16 x i16> [[R]] 59; 60; AVX-LABEL: @bitcast_shuf_narrow_element_concat_subvectors( 61; AVX-NEXT: [[TMP1:%.*]] = bitcast <2 x i64> [[V:%.*]] to <8 x i16> 62; AVX-NEXT: [[R:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7> 63; AVX-NEXT: ret <16 x i16> [[R]] 64; 65 %shuf = shufflevector <2 x i64> %v, <2 x i64> poison, <4 x i32> <i32 0, i32 1, i32 0, i32 1> 66 %r = bitcast <4 x i64> %shuf to <16 x i16> 67 ret <16 x i16> %r 68} 69 70define <16 x i8> @bitcast_shuf_extract_subvector(<8 x i32> %v) { 71; CHECK-LABEL: @bitcast_shuf_extract_subvector( 72; CHECK-NEXT: [[TMP1:%.*]] = bitcast <8 x i32> [[V:%.*]] to <32 x i8> 73; CHECK-NEXT: [[R:%.*]] = shufflevector <32 x i8> [[TMP1]], <32 x i8> poison, <16 x i32> <i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31> 74; CHECK-NEXT: ret <16 x i8> [[R]] 75; 76 %shuf = shufflevector <8 x i32> %v, <8 x i32> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7> 77 %r = bitcast <4 x i32> %shuf to <16 x i8> 78 ret <16 x i8> %r 79} 80 81; Negative test - must cast to vector type 82 83define i128 @bitcast_shuf_narrow_element_wrong_type(<4 x i32> %v) { 84; CHECK-LABEL: @bitcast_shuf_narrow_element_wrong_type( 85; CHECK-NEXT: [[SHUF:%.*]] = shufflevector <4 x i32> [[V:%.*]], <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 86; CHECK-NEXT: [[R:%.*]] = bitcast <4 x i32> [[SHUF]] to i128 87; CHECK-NEXT: ret i128 [[R]] 88; 89 %shuf = shufflevector <4 x i32> %v, <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 90 %r = bitcast <4 x i32> %shuf to i128 91 ret i128 %r 92} 93 94; Widen shuffle elements 95 96define <4 x i32> @bitcast_shuf_wide_element(<8 x i16> %v) { 97; CHECK-LABEL: @bitcast_shuf_wide_element( 98; CHECK-NEXT: [[TMP1:%.*]] = bitcast <8 x i16> [[V:%.*]] to <4 x i32> 99; CHECK-NEXT: [[R:%.*]] = shufflevector <4 x i32> [[TMP1]], <4 x i32> poison, <4 x i32> <i32 0, i32 0, i32 1, i32 1> 100; CHECK-NEXT: ret <4 x i32> [[R]] 101; 102 %shuf = shufflevector <8 x i16> %v, <8 x i16> poison, <8 x i32> <i32 0, i32 1, i32 0, i32 1, i32 2, i32 3, i32 2, i32 3> 103 %r = bitcast <8 x i16> %shuf to <4 x i32> 104 ret <4 x i32> %r 105} 106 107declare void @use(<4 x i32>) 108 109; Negative test - don't create an extra shuffle 110 111define <16 x i8> @bitcast_shuf_uses(<4 x i32> %v) { 112; CHECK-LABEL: @bitcast_shuf_uses( 113; CHECK-NEXT: [[SHUF:%.*]] = shufflevector <4 x i32> [[V:%.*]], <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 114; CHECK-NEXT: call void @use(<4 x i32> [[SHUF]]) 115; CHECK-NEXT: [[R:%.*]] = bitcast <4 x i32> [[SHUF]] to <16 x i8> 116; CHECK-NEXT: ret <16 x i8> [[R]] 117; 118 %shuf = shufflevector <4 x i32> %v, <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 119 call void @use(<4 x i32> %shuf) 120 %r = bitcast <4 x i32> %shuf to <16 x i8> 121 ret <16 x i8> %r 122} 123 124define <2 x i64> @PR35454_1(<2 x i64> %v) { 125; SSE-LABEL: @PR35454_1( 126; SSE-NEXT: [[BC:%.*]] = bitcast <2 x i64> [[V:%.*]] to <4 x i32> 127; SSE-NEXT: [[PERMIL:%.*]] = shufflevector <4 x i32> [[BC]], <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 128; SSE-NEXT: [[BC1:%.*]] = bitcast <4 x i32> [[PERMIL]] to <16 x i8> 129; SSE-NEXT: [[ADD:%.*]] = shl <16 x i8> [[BC1]], splat (i8 1) 130; SSE-NEXT: [[BC2:%.*]] = bitcast <16 x i8> [[ADD]] to <4 x i32> 131; SSE-NEXT: [[PERMIL1:%.*]] = shufflevector <4 x i32> [[BC2]], <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 132; SSE-NEXT: [[BC3:%.*]] = bitcast <4 x i32> [[PERMIL1]] to <2 x i64> 133; SSE-NEXT: ret <2 x i64> [[BC3]] 134; 135; AVX-LABEL: @PR35454_1( 136; AVX-NEXT: [[TMP1:%.*]] = bitcast <2 x i64> [[V:%.*]] to <16 x i8> 137; AVX-NEXT: [[BC1:%.*]] = shufflevector <16 x i8> [[TMP1]], <16 x i8> poison, <16 x i32> <i32 12, i32 13, i32 14, i32 15, i32 8, i32 9, i32 10, i32 11, i32 4, i32 5, i32 6, i32 7, i32 0, i32 1, i32 2, i32 3> 138; AVX-NEXT: [[ADD:%.*]] = shl <16 x i8> [[BC1]], splat (i8 1) 139; AVX-NEXT: [[BC2:%.*]] = bitcast <16 x i8> [[ADD]] to <4 x i32> 140; AVX-NEXT: [[PERMIL1:%.*]] = shufflevector <4 x i32> [[BC2]], <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 141; AVX-NEXT: [[BC3:%.*]] = bitcast <4 x i32> [[PERMIL1]] to <2 x i64> 142; AVX-NEXT: ret <2 x i64> [[BC3]] 143; 144 %bc = bitcast <2 x i64> %v to <4 x i32> 145 %permil = shufflevector <4 x i32> %bc, <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 146 %bc1 = bitcast <4 x i32> %permil to <16 x i8> 147 %add = shl <16 x i8> %bc1, <i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1, i8 1> 148 %bc2 = bitcast <16 x i8> %add to <4 x i32> 149 %permil1 = shufflevector <4 x i32> %bc2, <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 150 %bc3 = bitcast <4 x i32> %permil1 to <2 x i64> 151 ret <2 x i64> %bc3 152} 153 154define <2 x i64> @PR35454_2(<2 x i64> %v) { 155; SSE-LABEL: @PR35454_2( 156; SSE-NEXT: [[BC:%.*]] = bitcast <2 x i64> [[V:%.*]] to <4 x i32> 157; SSE-NEXT: [[PERMIL:%.*]] = shufflevector <4 x i32> [[BC]], <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 158; SSE-NEXT: [[BC1:%.*]] = bitcast <4 x i32> [[PERMIL]] to <8 x i16> 159; SSE-NEXT: [[ADD:%.*]] = shl <8 x i16> [[BC1]], splat (i16 1) 160; SSE-NEXT: [[BC2:%.*]] = bitcast <8 x i16> [[ADD]] to <4 x i32> 161; SSE-NEXT: [[PERMIL1:%.*]] = shufflevector <4 x i32> [[BC2]], <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 162; SSE-NEXT: [[BC3:%.*]] = bitcast <4 x i32> [[PERMIL1]] to <2 x i64> 163; SSE-NEXT: ret <2 x i64> [[BC3]] 164; 165; AVX-LABEL: @PR35454_2( 166; AVX-NEXT: [[TMP1:%.*]] = bitcast <2 x i64> [[V:%.*]] to <8 x i16> 167; AVX-NEXT: [[BC1:%.*]] = shufflevector <8 x i16> [[TMP1]], <8 x i16> poison, <8 x i32> <i32 6, i32 7, i32 4, i32 5, i32 2, i32 3, i32 0, i32 1> 168; AVX-NEXT: [[ADD:%.*]] = shl <8 x i16> [[BC1]], splat (i16 1) 169; AVX-NEXT: [[BC2:%.*]] = bitcast <8 x i16> [[ADD]] to <4 x i32> 170; AVX-NEXT: [[PERMIL1:%.*]] = shufflevector <4 x i32> [[BC2]], <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 171; AVX-NEXT: [[BC3:%.*]] = bitcast <4 x i32> [[PERMIL1]] to <2 x i64> 172; AVX-NEXT: ret <2 x i64> [[BC3]] 173; 174 %bc = bitcast <2 x i64> %v to <4 x i32> 175 %permil = shufflevector <4 x i32> %bc, <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 176 %bc1 = bitcast <4 x i32> %permil to <8 x i16> 177 %add = shl <8 x i16> %bc1, <i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1, i16 1> 178 %bc2 = bitcast <8 x i16> %add to <4 x i32> 179 %permil1 = shufflevector <4 x i32> %bc2, <4 x i32> poison, <4 x i32> <i32 3, i32 2, i32 1, i32 0> 180 %bc3 = bitcast <4 x i32> %permil1 to <2 x i64> 181 ret <2 x i64> %bc3 182} 183