|
| 1 | +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py |
| 2 | +; RUN: opt < %s -O3 -S -mtriple=x86_64-- -mcpu=x86-64-v2 | FileCheck %s |
| 3 | +; RUN: opt < %s -O3 -S -mtriple=x86_64-- -mcpu=x86-64-v3 | FileCheck %s |
| 4 | +; RUN: opt < %s -O3 -S -mtriple=x86_64-- -mcpu=x86-64-v4 | FileCheck %s |
| 5 | + |
| 6 | +define <4 x i64> @PR67803(<4 x i64> %x, <4 x i64> %y, <4 x i64> %a, <4 x i64> %b) { |
| 7 | +; CHECK-LABEL: @PR67803( |
| 8 | +; CHECK-NEXT: entry: |
| 9 | +; CHECK-NEXT: [[TMP0:%.*]] = bitcast <4 x i64> [[X:%.*]] to <8 x i32> |
| 10 | +; CHECK-NEXT: [[TMP1:%.*]] = bitcast <4 x i64> [[Y:%.*]] to <8 x i32> |
| 11 | +; CHECK-NEXT: [[TMP2:%.*]] = icmp sgt <8 x i32> [[TMP0]], [[TMP1]] |
| 12 | +; CHECK-NEXT: [[CMP_I21:%.*]] = shufflevector <8 x i1> [[TMP2]], <8 x i1> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3> |
| 13 | +; CHECK-NEXT: [[SEXT_I22:%.*]] = sext <4 x i1> [[CMP_I21]] to <4 x i32> |
| 14 | +; CHECK-NEXT: [[TMP3:%.*]] = bitcast <4 x i32> [[SEXT_I22]] to <2 x i64> |
| 15 | +; CHECK-NEXT: [[CMP_I:%.*]] = shufflevector <8 x i1> [[TMP2]], <8 x i1> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7> |
| 16 | +; CHECK-NEXT: [[SEXT_I:%.*]] = sext <4 x i1> [[CMP_I]] to <4 x i32> |
| 17 | +; CHECK-NEXT: [[TMP4:%.*]] = bitcast <4 x i32> [[SEXT_I]] to <2 x i64> |
| 18 | +; CHECK-NEXT: [[SHUFFLE_I:%.*]] = shufflevector <2 x i64> [[TMP3]], <2 x i64> [[TMP4]], <4 x i32> <i32 0, i32 1, i32 2, i32 3> |
| 19 | +; CHECK-NEXT: [[TMP5:%.*]] = bitcast <4 x i64> [[A:%.*]] to <32 x i8> |
| 20 | +; CHECK-NEXT: [[TMP6:%.*]] = shufflevector <32 x i8> [[TMP5]], <32 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15> |
| 21 | +; CHECK-NEXT: [[TMP7:%.*]] = bitcast <4 x i64> [[B:%.*]] to <32 x i8> |
| 22 | +; CHECK-NEXT: [[TMP8:%.*]] = shufflevector <32 x i8> [[TMP7]], <32 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15> |
| 23 | +; CHECK-NEXT: [[TMP9:%.*]] = bitcast <4 x i64> [[SHUFFLE_I]] to <32 x i8> |
| 24 | +; CHECK-NEXT: [[TMP10:%.*]] = shufflevector <32 x i8> [[TMP9]], <32 x i8> poison, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 8, i32 9, i32 10, i32 11, i32 12, i32 13, i32 14, i32 15> |
| 25 | +; CHECK-NEXT: [[TMP11:%.*]] = tail call <16 x i8> @llvm.x86.sse41.pblendvb(<16 x i8> [[TMP6]], <16 x i8> [[TMP8]], <16 x i8> [[TMP10]]) |
| 26 | +; CHECK-NEXT: [[TMP12:%.*]] = bitcast <16 x i8> [[TMP11]] to <2 x i64> |
| 27 | +; CHECK-NEXT: [[TMP13:%.*]] = bitcast <4 x i64> [[A]] to <32 x i8> |
| 28 | +; CHECK-NEXT: [[TMP14:%.*]] = shufflevector <32 x i8> [[TMP13]], <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> |
| 29 | +; CHECK-NEXT: [[TMP15:%.*]] = bitcast <4 x i64> [[B]] to <32 x i8> |
| 30 | +; CHECK-NEXT: [[TMP16:%.*]] = shufflevector <32 x i8> [[TMP15]], <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> |
| 31 | +; CHECK-NEXT: [[TMP17:%.*]] = bitcast <4 x i64> [[SHUFFLE_I]] to <32 x i8> |
| 32 | +; CHECK-NEXT: [[TMP18:%.*]] = shufflevector <32 x i8> [[TMP17]], <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> |
| 33 | +; CHECK-NEXT: [[TMP19:%.*]] = tail call <16 x i8> @llvm.x86.sse41.pblendvb(<16 x i8> [[TMP14]], <16 x i8> [[TMP16]], <16 x i8> [[TMP18]]) |
| 34 | +; CHECK-NEXT: [[TMP20:%.*]] = bitcast <16 x i8> [[TMP19]] to <2 x i64> |
| 35 | +; CHECK-NEXT: [[SHUFFLE_I23:%.*]] = shufflevector <2 x i64> [[TMP12]], <2 x i64> [[TMP20]], <4 x i32> <i32 0, i32 1, i32 2, i32 3> |
| 36 | +; CHECK-NEXT: ret <4 x i64> [[SHUFFLE_I23]] |
| 37 | +; |
| 38 | +entry: |
| 39 | + %0 = bitcast <4 x i64> %x to <8 x i32> |
| 40 | + %extract = shufflevector <8 x i32> %0, <8 x i32> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3> |
| 41 | + %1 = bitcast <4 x i32> %extract to <2 x i64> |
| 42 | + %2 = bitcast <4 x i64> %y to <8 x i32> |
| 43 | + %extract1 = shufflevector <8 x i32> %2, <8 x i32> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3> |
| 44 | + %3 = bitcast <4 x i32> %extract1 to <2 x i64> |
| 45 | + %cmp.i21 = icmp sgt <4 x i32> %extract, %extract1 |
| 46 | + %sext.i22 = sext <4 x i1> %cmp.i21 to <4 x i32> |
| 47 | + %4 = bitcast <4 x i32> %sext.i22 to <2 x i64> |
| 48 | + %extract2 = shufflevector <8 x i32> %0, <8 x i32> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7> |
| 49 | + %5 = bitcast <4 x i32> %extract2 to <2 x i64> |
| 50 | + %extract3 = shufflevector <8 x i32> %2, <8 x i32> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7> |
| 51 | + %6 = bitcast <4 x i32> %extract3 to <2 x i64> |
| 52 | + %cmp.i = icmp sgt <4 x i32> %extract2, %extract3 |
| 53 | + %sext.i = sext <4 x i1> %cmp.i to <4 x i32> |
| 54 | + %7 = bitcast <4 x i32> %sext.i to <2 x i64> |
| 55 | + %shuffle.i = shufflevector <2 x i64> %4, <2 x i64> %7, <4 x i32> <i32 0, i32 1, i32 2, i32 3> |
| 56 | + %8 = bitcast <4 x i64> %a to <8 x i32> |
| 57 | + %extract6 = shufflevector <8 x i32> %8, <8 x i32> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3> |
| 58 | + %9 = bitcast <4 x i32> %extract6 to <2 x i64> |
| 59 | + %10 = bitcast <4 x i64> %b to <8 x i32> |
| 60 | + %extract7 = shufflevector <8 x i32> %10, <8 x i32> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3> |
| 61 | + %11 = bitcast <4 x i32> %extract7 to <2 x i64> |
| 62 | + %12 = bitcast <4 x i64> %shuffle.i to <8 x i32> |
| 63 | + %extract8 = shufflevector <8 x i32> %12, <8 x i32> poison, <4 x i32> <i32 0, i32 1, i32 2, i32 3> |
| 64 | + %13 = bitcast <4 x i32> %extract8 to <2 x i64> |
| 65 | + %14 = bitcast <2 x i64> %9 to <16 x i8> |
| 66 | + %15 = bitcast <2 x i64> %11 to <16 x i8> |
| 67 | + %16 = bitcast <2 x i64> %13 to <16 x i8> |
| 68 | + %17 = call <16 x i8> @llvm.x86.sse41.pblendvb(<16 x i8> %14, <16 x i8> %15, <16 x i8> %16) |
| 69 | + %18 = bitcast <16 x i8> %17 to <2 x i64> |
| 70 | + %extract10 = shufflevector <8 x i32> %8, <8 x i32> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7> |
| 71 | + %19 = bitcast <4 x i32> %extract10 to <2 x i64> |
| 72 | + %extract11 = shufflevector <8 x i32> %10, <8 x i32> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7> |
| 73 | + %20 = bitcast <4 x i32> %extract11 to <2 x i64> |
| 74 | + %extract12 = shufflevector <8 x i32> %12, <8 x i32> poison, <4 x i32> <i32 4, i32 5, i32 6, i32 7> |
| 75 | + %21 = bitcast <4 x i32> %extract12 to <2 x i64> |
| 76 | + %22 = bitcast <2 x i64> %19 to <16 x i8> |
| 77 | + %23 = bitcast <2 x i64> %20 to <16 x i8> |
| 78 | + %24 = bitcast <2 x i64> %21 to <16 x i8> |
| 79 | + %25 = call <16 x i8> @llvm.x86.sse41.pblendvb(<16 x i8> %22, <16 x i8> %23, <16 x i8> %24) |
| 80 | + %26 = bitcast <16 x i8> %25 to <2 x i64> |
| 81 | + %shuffle.i23 = shufflevector <2 x i64> %18, <2 x i64> %26, <4 x i32> <i32 0, i32 1, i32 2, i32 3> |
| 82 | + ret <4 x i64> %shuffle.i23 |
| 83 | +} |
0 commit comments