|
| 1 | +; NOTE: Assertions have been autogenerated by utils/update_test_checks.py |
| 2 | +; RUN: opt -slp-vectorizer -slp-vectorize-hor -slp-vectorize-hor-store -S < %s -mtriple=x86_64-apple-macosx -mcpu=haswell | FileCheck %s |
| 3 | + |
| 4 | +;unsigned load_le32(unsigned char *data) { |
| 5 | +; unsigned le32 = (data[0]<<0) | (data[1]<<8) | (data[2]<<16) | (data[3]<<24); |
| 6 | +; return le32; |
| 7 | +;} |
| 8 | + |
| 9 | +define i32 @_Z9load_le32Ph(i8* nocapture readonly %data) { |
| 10 | +; CHECK-LABEL: @_Z9load_le32Ph( |
| 11 | +; CHECK-NEXT: entry: |
| 12 | +; CHECK-NEXT: [[TMP0:%.*]] = load i8, i8* [[DATA:%.*]], align 1 |
| 13 | +; CHECK-NEXT: [[CONV:%.*]] = zext i8 [[TMP0]] to i32 |
| 14 | +; CHECK-NEXT: [[ARRAYIDX1:%.*]] = getelementptr inbounds i8, i8* [[DATA]], i64 1 |
| 15 | +; CHECK-NEXT: [[TMP1:%.*]] = load i8, i8* [[ARRAYIDX1]], align 1 |
| 16 | +; CHECK-NEXT: [[CONV2:%.*]] = zext i8 [[TMP1]] to i32 |
| 17 | +; CHECK-NEXT: [[SHL3:%.*]] = shl nuw nsw i32 [[CONV2]], 8 |
| 18 | +; CHECK-NEXT: [[OR:%.*]] = or i32 [[SHL3]], [[CONV]] |
| 19 | +; CHECK-NEXT: [[ARRAYIDX4:%.*]] = getelementptr inbounds i8, i8* [[DATA]], i64 2 |
| 20 | +; CHECK-NEXT: [[ARRAYIDX8:%.*]] = getelementptr inbounds i8, i8* [[DATA]], i64 3 |
| 21 | +; CHECK-NEXT: [[TMP2:%.*]] = bitcast i8* [[ARRAYIDX4]] to <2 x i8>* |
| 22 | +; CHECK-NEXT: [[TMP3:%.*]] = load <2 x i8>, <2 x i8>* [[TMP2]], align 1 |
| 23 | +; CHECK-NEXT: [[TMP4:%.*]] = zext <2 x i8> [[TMP3]] to <2 x i32> |
| 24 | +; CHECK-NEXT: [[TMP5:%.*]] = shl nuw <2 x i32> [[TMP4]], <i32 16, i32 24> |
| 25 | +; CHECK-NEXT: [[TMP6:%.*]] = extractelement <2 x i32> [[TMP5]], i32 0 |
| 26 | +; CHECK-NEXT: [[OR7:%.*]] = or i32 [[OR]], [[TMP6]] |
| 27 | +; CHECK-NEXT: [[TMP7:%.*]] = extractelement <2 x i32> [[TMP5]], i32 1 |
| 28 | +; CHECK-NEXT: [[OR11:%.*]] = or i32 [[OR7]], [[TMP7]] |
| 29 | +; CHECK-NEXT: ret i32 [[OR11]] |
| 30 | +; |
| 31 | +entry: |
| 32 | + %0 = load i8, i8* %data, align 1 |
| 33 | + %conv = zext i8 %0 to i32 |
| 34 | + %arrayidx1 = getelementptr inbounds i8, i8* %data, i64 1 |
| 35 | + %1 = load i8, i8* %arrayidx1, align 1 |
| 36 | + %conv2 = zext i8 %1 to i32 |
| 37 | + %shl3 = shl nuw nsw i32 %conv2, 8 |
| 38 | + %or = or i32 %shl3, %conv |
| 39 | + %arrayidx4 = getelementptr inbounds i8, i8* %data, i64 2 |
| 40 | + %2 = load i8, i8* %arrayidx4, align 1 |
| 41 | + %conv5 = zext i8 %2 to i32 |
| 42 | + %shl6 = shl nuw nsw i32 %conv5, 16 |
| 43 | + %or7 = or i32 %or, %shl6 |
| 44 | + %arrayidx8 = getelementptr inbounds i8, i8* %data, i64 3 |
| 45 | + %3 = load i8, i8* %arrayidx8, align 1 |
| 46 | + %conv9 = zext i8 %3 to i32 |
| 47 | + %shl10 = shl nuw i32 %conv9, 24 |
| 48 | + %or11 = or i32 %or7, %shl10 |
| 49 | + ret i32 %or11 |
| 50 | +} |
0 commit comments