1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py 2; RUN: opt < %s -S -mtriple=x86_64-unknown -mattr=+avx -passes=slp-vectorizer | FileCheck %s 3 4 5 6define i32 @jumbled-load(ptr noalias nocapture %in, ptr noalias nocapture %inn, ptr noalias nocapture %out) { 7; CHECK-LABEL: @jumbled-load( 8; CHECK-NEXT: [[TMP2:%.*]] = load <4 x i32>, ptr [[IN:%.*]], align 4 9; CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, ptr [[INN:%.*]], align 4 10; CHECK-NEXT: [[SHUFFLE:%.*]] = shufflevector <4 x i32> [[TMP4]], <4 x i32> poison, <4 x i32> <i32 2, i32 0, i32 3, i32 1> 11; CHECK-NEXT: [[TMP5:%.*]] = mul <4 x i32> [[TMP2]], [[SHUFFLE]] 12; CHECK-NEXT: [[SHUFFLE1:%.*]] = shufflevector <4 x i32> [[TMP5]], <4 x i32> poison, <4 x i32> <i32 1, i32 3, i32 2, i32 0> 13; CHECK-NEXT: store <4 x i32> [[SHUFFLE1]], ptr [[OUT:%.*]], align 4 14; CHECK-NEXT: ret i32 undef 15; 16 %load.1 = load i32, ptr %in, align 4 17 %gep.1 = getelementptr inbounds i32, ptr %in, i64 3 18 %load.2 = load i32, ptr %gep.1, align 4 19 %gep.2 = getelementptr inbounds i32, ptr %in, i64 1 20 %load.3 = load i32, ptr %gep.2, align 4 21 %gep.3 = getelementptr inbounds i32, ptr %in, i64 2 22 %load.4 = load i32, ptr %gep.3, align 4 23 %load.5 = load i32, ptr %inn, align 4 24 %gep.4 = getelementptr inbounds i32, ptr %inn, i64 2 25 %load.6 = load i32, ptr %gep.4, align 4 26 %gep.5 = getelementptr inbounds i32, ptr %inn, i64 3 27 %load.7 = load i32, ptr %gep.5, align 4 28 %gep.6 = getelementptr inbounds i32, ptr %inn, i64 1 29 %load.8 = load i32, ptr %gep.6, align 4 30 %mul.1 = mul i32 %load.3, %load.5 31 %mul.2 = mul i32 %load.2, %load.8 32 %mul.3 = mul i32 %load.4, %load.7 33 %mul.4 = mul i32 %load.1, %load.6 34 store i32 %mul.1, ptr %out, align 4 35 %gep.8 = getelementptr inbounds i32, ptr %out, i64 1 36 store i32 %mul.2, ptr %gep.8, align 4 37 %gep.9 = getelementptr inbounds i32, ptr %out, i64 2 38 store i32 %mul.3, ptr %gep.9, align 4 39 %gep.10 = getelementptr inbounds i32, ptr %out, i64 3 40 store i32 %mul.4, ptr %gep.10, align 4 41 42 ret i32 undef 43} 44 45 46define i32 @jumbled-load-multiuses(ptr noalias nocapture %in, ptr noalias nocapture %out) { 47; CHECK-LABEL: @jumbled-load-multiuses( 48; CHECK-NEXT: [[TMP2:%.*]] = load <4 x i32>, ptr [[IN:%.*]], align 4 49; CHECK-NEXT: [[TMP3:%.*]] = shufflevector <4 x i32> [[TMP2]], <4 x i32> poison, <4 x i32> <i32 1, i32 2, i32 0, i32 3> 50; CHECK-NEXT: [[TMP4:%.*]] = mul <4 x i32> [[TMP2]], [[TMP3]] 51; CHECK-NEXT: [[SHUFFLE:%.*]] = shufflevector <4 x i32> [[TMP4]], <4 x i32> poison, <4 x i32> <i32 1, i32 3, i32 2, i32 0> 52; CHECK-NEXT: store <4 x i32> [[SHUFFLE]], ptr [[OUT:%.*]], align 4 53; CHECK-NEXT: ret i32 undef 54; 55 %load.1 = load i32, ptr %in, align 4 56 %gep.1 = getelementptr inbounds i32, ptr %in, i64 3 57 %load.2 = load i32, ptr %gep.1, align 4 58 %gep.2 = getelementptr inbounds i32, ptr %in, i64 1 59 %load.3 = load i32, ptr %gep.2, align 4 60 %gep.3 = getelementptr inbounds i32, ptr %in, i64 2 61 %load.4 = load i32, ptr %gep.3, align 4 62 %mul.1 = mul i32 %load.3, %load.4 63 %mul.2 = mul i32 %load.2, %load.2 64 %mul.3 = mul i32 %load.4, %load.1 65 %mul.4 = mul i32 %load.1, %load.3 66 store i32 %mul.1, ptr %out, align 4 67 %gep.8 = getelementptr inbounds i32, ptr %out, i64 1 68 store i32 %mul.2, ptr %gep.8, align 4 69 %gep.9 = getelementptr inbounds i32, ptr %out, i64 2 70 store i32 %mul.3, ptr %gep.9, align 4 71 %gep.10 = getelementptr inbounds i32, ptr %out, i64 3 72 store i32 %mul.4, ptr %gep.10, align 4 73 74 ret i32 undef 75} 76