1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py 2; RUN: opt < %s -slp-vectorizer -S | FileCheck %s 3; RUN: opt < %s -slp-vectorizer -enable-new-pm -S | FileCheck %s 4 5target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128" 6target triple = "x86_64-unknown-linux-gnu" 7 8@b = common global [4 x i32] zeroinitializer, align 16 9@c = common global [4 x i32] zeroinitializer, align 16 10@d = common global [4 x i32] zeroinitializer, align 16 11@e = common global [4 x i32] zeroinitializer, align 16 12@a = common global [4 x i32] zeroinitializer, align 16 13@fb = common global [4 x float] zeroinitializer, align 16 14@fc = common global [4 x float] zeroinitializer, align 16 15@fa = common global [4 x float] zeroinitializer, align 16 16@fd = common global [4 x float] zeroinitializer, align 16 17 18define void @addsub() { 19; CHECK-LABEL: @addsub( 20; CHECK-NEXT: entry: 21; CHECK-NEXT: br label [[BB1:%.*]] 22; CHECK: bb1: 23; CHECK-NEXT: [[TMP0:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([4 x i32]* @b to <4 x i32>*), align 16 24; CHECK-NEXT: [[TMP1:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([4 x i32]* @c to <4 x i32>*), align 16 25; CHECK-NEXT: [[TMP2:%.*]] = add nsw <4 x i32> [[TMP0]], [[TMP1]] 26; CHECK-NEXT: [[TMP3:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([4 x i32]* @d to <4 x i32>*), align 16 27; CHECK-NEXT: [[TMP4:%.*]] = load <4 x i32>, <4 x i32>* bitcast ([4 x i32]* @e to <4 x i32>*), align 16 28; CHECK-NEXT: [[TMP5:%.*]] = add nsw <4 x i32> [[TMP3]], [[TMP4]] 29; CHECK-NEXT: [[TMP6:%.*]] = add nsw <4 x i32> [[TMP2]], [[TMP5]] 30; CHECK-NEXT: [[TMP7:%.*]] = sub nsw <4 x i32> [[TMP2]], [[TMP5]] 31; CHECK-NEXT: [[TMP8:%.*]] = shufflevector <4 x i32> [[TMP6]], <4 x i32> [[TMP7]], <4 x i32> <i32 0, i32 5, i32 2, i32 7> 32; CHECK-NEXT: store <4 x i32> [[TMP8]], <4 x i32>* bitcast ([4 x i32]* @a to <4 x i32>*), align 16 33; CHECK-NEXT: ret void 34; 35entry: 36 br label %bb1 37 38bb1: 39 %0 = load i32, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @b, i64 0, i64 0), align 16 40 %1 = load i32, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @c, i64 0, i64 0), align 16 41 %add = add nsw i32 %0, %1 42 %2 = load i32, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @d, i64 0, i64 0), align 16 43 %3 = load i32, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @e, i64 0, i64 0), align 16 44 %add1 = add nsw i32 %2, %3 45 %add2 = add nsw i32 %add, %add1 46 store i32 %add2, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @a, i64 0, i64 0), align 16 47 %4 = load i32, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @b, i64 0, i64 1), align 4 48 %5 = load i32, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @c, i64 0, i64 1), align 4 49 %add3 = add nsw i32 %4, %5 50 %6 = load i32, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @d, i64 0, i64 1), align 4 51 %7 = load i32, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @e, i64 0, i64 1), align 4 52 %add4 = add nsw i32 %6, %7 53 %sub = sub nsw i32 %add3, %add4 54 store i32 %sub, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @a, i64 0, i64 1), align 4 55 %8 = load i32, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @b, i64 0, i64 2), align 8 56 %9 = load i32, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @c, i64 0, i64 2), align 8 57 %add5 = add nsw i32 %8, %9 58 %10 = load i32, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @d, i64 0, i64 2), align 8 59 %11 = load i32, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @e, i64 0, i64 2), align 8 60 %add6 = add nsw i32 %10, %11 61 %add7 = add nsw i32 %add5, %add6 62 store i32 %add7, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @a, i64 0, i64 2), align 8 63 %12 = load i32, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @b, i64 0, i64 3), align 4 64 %13 = load i32, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @c, i64 0, i64 3), align 4 65 %add8 = add nsw i32 %12, %13 66 %14 = load i32, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @d, i64 0, i64 3), align 4 67 %15 = load i32, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @e, i64 0, i64 3), align 4 68 %add9 = add nsw i32 %14, %15 69 %sub10 = sub nsw i32 %add8, %add9 70 store i32 %sub10, i32* getelementptr inbounds ([4 x i32], [4 x i32]* @a, i64 0, i64 3), align 4 71 ret void 72} 73