1; RUN: llc -O3 -mcpu=pwr8 -mtriple=powerpc64le-unknown-linux-gnu < %s | FileCheck %s 2; RUN: llc -O3 -mcpu=pwr8 -disable-ppc-vsx-swap-removal -mtriple=powerpc64le-unknown-linux-gnu < %s | FileCheck -check-prefix=NOOPTSWAP %s 3 4; This test was generated from the following source: 5; 6; #define N 4096 7; int ca[N] __attribute__((aligned(16))); 8; int cb[N] __attribute__((aligned(16))); 9; int cc[N] __attribute__((aligned(16))); 10; int cd[N] __attribute__((aligned(16))); 11; 12; void foo () 13; { 14; int i; 15; for (i = 0; i < N; i++) { 16; ca[i] = (cb[i] + cc[i]) * cd[i]; 17; } 18; } 19 20@cb = common global [4096 x i32] zeroinitializer, align 16 21@cc = common global [4096 x i32] zeroinitializer, align 16 22@cd = common global [4096 x i32] zeroinitializer, align 16 23@ca = common global [4096 x i32] zeroinitializer, align 16 24 25define void @foo() { 26entry: 27 br label %vector.body 28 29vector.body: 30 %index = phi i64 [ 0, %entry ], [ %index.next.3, %vector.body ] 31 %0 = getelementptr inbounds [4096 x i32], [4096 x i32]* @cb, i64 0, i64 %index 32 %1 = bitcast i32* %0 to <4 x i32>* 33 %wide.load = load <4 x i32>, <4 x i32>* %1, align 16 34 %2 = getelementptr inbounds [4096 x i32], [4096 x i32]* @cc, i64 0, i64 %index 35 %3 = bitcast i32* %2 to <4 x i32>* 36 %wide.load13 = load <4 x i32>, <4 x i32>* %3, align 16 37 %4 = add nsw <4 x i32> %wide.load13, %wide.load 38 %5 = getelementptr inbounds [4096 x i32], [4096 x i32]* @cd, i64 0, i64 %index 39 %6 = bitcast i32* %5 to <4 x i32>* 40 %wide.load14 = load <4 x i32>, <4 x i32>* %6, align 16 41 %7 = mul nsw <4 x i32> %4, %wide.load14 42 %8 = getelementptr inbounds [4096 x i32], [4096 x i32]* @ca, i64 0, i64 %index 43 %9 = bitcast i32* %8 to <4 x i32>* 44 store <4 x i32> %7, <4 x i32>* %9, align 16 45 %index.next = add nuw nsw i64 %index, 4 46 %10 = getelementptr inbounds [4096 x i32], [4096 x i32]* @cb, i64 0, i64 %index.next 47 %11 = bitcast i32* %10 to <4 x i32>* 48 %wide.load.1 = load <4 x i32>, <4 x i32>* %11, align 16 49 %12 = getelementptr inbounds [4096 x i32], [4096 x i32]* @cc, i64 0, i64 %index.next 50 %13 = bitcast i32* %12 to <4 x i32>* 51 %wide.load13.1 = load <4 x i32>, <4 x i32>* %13, align 16 52 %14 = add nsw <4 x i32> %wide.load13.1, %wide.load.1 53 %15 = getelementptr inbounds [4096 x i32], [4096 x i32]* @cd, i64 0, i64 %index.next 54 %16 = bitcast i32* %15 to <4 x i32>* 55 %wide.load14.1 = load <4 x i32>, <4 x i32>* %16, align 16 56 %17 = mul nsw <4 x i32> %14, %wide.load14.1 57 %18 = getelementptr inbounds [4096 x i32], [4096 x i32]* @ca, i64 0, i64 %index.next 58 %19 = bitcast i32* %18 to <4 x i32>* 59 store <4 x i32> %17, <4 x i32>* %19, align 16 60 %index.next.1 = add nuw nsw i64 %index.next, 4 61 %20 = getelementptr inbounds [4096 x i32], [4096 x i32]* @cb, i64 0, i64 %index.next.1 62 %21 = bitcast i32* %20 to <4 x i32>* 63 %wide.load.2 = load <4 x i32>, <4 x i32>* %21, align 16 64 %22 = getelementptr inbounds [4096 x i32], [4096 x i32]* @cc, i64 0, i64 %index.next.1 65 %23 = bitcast i32* %22 to <4 x i32>* 66 %wide.load13.2 = load <4 x i32>, <4 x i32>* %23, align 16 67 %24 = add nsw <4 x i32> %wide.load13.2, %wide.load.2 68 %25 = getelementptr inbounds [4096 x i32], [4096 x i32]* @cd, i64 0, i64 %index.next.1 69 %26 = bitcast i32* %25 to <4 x i32>* 70 %wide.load14.2 = load <4 x i32>, <4 x i32>* %26, align 16 71 %27 = mul nsw <4 x i32> %24, %wide.load14.2 72 %28 = getelementptr inbounds [4096 x i32], [4096 x i32]* @ca, i64 0, i64 %index.next.1 73 %29 = bitcast i32* %28 to <4 x i32>* 74 store <4 x i32> %27, <4 x i32>* %29, align 16 75 %index.next.2 = add nuw nsw i64 %index.next.1, 4 76 %30 = getelementptr inbounds [4096 x i32], [4096 x i32]* @cb, i64 0, i64 %index.next.2 77 %31 = bitcast i32* %30 to <4 x i32>* 78 %wide.load.3 = load <4 x i32>, <4 x i32>* %31, align 16 79 %32 = getelementptr inbounds [4096 x i32], [4096 x i32]* @cc, i64 0, i64 %index.next.2 80 %33 = bitcast i32* %32 to <4 x i32>* 81 %wide.load13.3 = load <4 x i32>, <4 x i32>* %33, align 16 82 %34 = add nsw <4 x i32> %wide.load13.3, %wide.load.3 83 %35 = getelementptr inbounds [4096 x i32], [4096 x i32]* @cd, i64 0, i64 %index.next.2 84 %36 = bitcast i32* %35 to <4 x i32>* 85 %wide.load14.3 = load <4 x i32>, <4 x i32>* %36, align 16 86 %37 = mul nsw <4 x i32> %34, %wide.load14.3 87 %38 = getelementptr inbounds [4096 x i32], [4096 x i32]* @ca, i64 0, i64 %index.next.2 88 %39 = bitcast i32* %38 to <4 x i32>* 89 store <4 x i32> %37, <4 x i32>* %39, align 16 90 %index.next.3 = add nuw nsw i64 %index.next.2, 4 91 %40 = icmp eq i64 %index.next.3, 4096 92 br i1 %40, label %for.end, label %vector.body 93 94for.end: 95 ret void 96} 97 98; CHECK-LABEL: @foo 99; CHECK-NOT: xxpermdi 100; CHECK-NOT: xxswapd 101 102; CHECK: lxvd2x 103; CHECK: lxvd2x 104; CHECK-DAG: lxvd2x 105; CHECK-DAG: vadduwm 106; CHECK: vmuluwm 107; CHECK: stxvd2x 108 109; CHECK: lxvd2x 110; CHECK: lxvd2x 111; CHECK-DAG: lxvd2x 112; CHECK-DAG: vadduwm 113; CHECK: vmuluwm 114; CHECK: stxvd2x 115 116; CHECK: lxvd2x 117; CHECK: lxvd2x 118; CHECK-DAG: lxvd2x 119; CHECK-DAG: vadduwm 120; CHECK: vmuluwm 121; CHECK: stxvd2x 122 123; CHECK: lxvd2x 124; CHECK: lxvd2x 125; CHECK-DAG: lxvd2x 126; CHECK-DAG: vadduwm 127; CHECK: vmuluwm 128; CHECK: stxvd2x 129 130 131; NOOPTSWAP-LABEL: @foo 132 133; NOOPTSWAP: lxvd2x 134; NOOPTSWAP-DAG: lxvd2x 135; NOOPTSWAP-DAG: lxvd2x 136; NOOPTSWAP-DAG: xxswapd 137; NOOPTSWAP-DAG: xxswapd 138; NOOPTSWAP-DAG: xxswapd 139; NOOPTSWAP-DAG: vadduwm 140; NOOPTSWAP: vmuluwm 141; NOOPTSWAP: xxswapd 142; NOOPTSWAP-DAG: xxswapd 143; NOOPTSWAP-DAG: xxswapd 144; NOOPTSWAP-DAG: stxvd2x 145; NOOPTSWAP-DAG: stxvd2x 146; NOOPTSWAP: stxvd2x 147 148