xref: /llvm-project/llvm/test/Transforms/SLPVectorizer/extended-vectorized-gathered-inst.ll (revision 38fffa630ee80163dc65e759392ad29798905679)
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
2; RUN: %if x86-registered-target %{ opt -S --passes=slp-vectorizer -mtriple=x86_64-unknown-linux < %s | FileCheck %s %}
3; RUN: %if aarch64-registered-target %{ opt -S --passes=slp-vectorizer -mtriple=aarch64-unknown-linux < %s | FileCheck %s %}
4
5define void @test(ptr %top) {
6; CHECK-LABEL: define void @test(
7; CHECK-SAME: ptr [[TOP:%.*]]) {
8; CHECK-NEXT:  entry:
9; CHECK-NEXT:    [[TMP0:%.*]] = load <4 x i8>, ptr [[TOP]], align 1
10; CHECK-NEXT:    [[TMP1:%.*]] = mul <4 x i8> [[TMP0]], zeroinitializer
11; CHECK-NEXT:    [[TMP2:%.*]] = extractelement <4 x i8> [[TMP0]], i32 2
12; CHECK-NEXT:    [[TMP3:%.*]] = zext i8 [[TMP2]] to i32
13; CHECK-NEXT:    [[TMP4:%.*]] = trunc i32 [[TMP3]] to i8
14; CHECK-NEXT:    [[TMP5:%.*]] = insertelement <4 x i8> <i8 0, i8 0, i8 0, i8 poison>, i8 [[TMP4]], i32 3
15; CHECK-NEXT:    [[TMP6:%.*]] = or <4 x i8> [[TMP1]], [[TMP5]]
16; CHECK-NEXT:    [[TMP7:%.*]] = or <4 x i8> [[TMP6]], zeroinitializer
17; CHECK-NEXT:    [[TMP8:%.*]] = lshr <4 x i8> [[TMP7]], splat (i8 2)
18; CHECK-NEXT:    br label [[FOR_COND_I:%.*]]
19; CHECK:       for.cond.i:
20; CHECK-NEXT:    store <4 x i8> [[TMP8]], ptr null, align 1
21; CHECK-NEXT:    br label [[FOR_COND_I]]
22;
23entry:
24  %0 = load i8, ptr %top, align 1
25  %conv2.i = zext i8 %0 to i32
26  %mul.i = mul i32 %conv2.i, 0
27  %add.i = or i32 %mul.i, 0
28  %arrayidx3.i = getelementptr i8, ptr %top, i64 1
29  %1 = load i8, ptr %arrayidx3.i, align 1
30  %conv4.i = zext i8 %1 to i32
31  %add5.i = or i32 %add.i, 0
32  %shr.i = lshr i32 %add5.i, 2
33  %conv7.i = trunc i32 %shr.i to i8
34  %mul12.i = mul i32 %conv4.i, 0
35  %arrayidx14.i = getelementptr i8, ptr %top, i64 2
36  %2 = load i8, ptr %arrayidx14.i, align 1
37  %conv15.i = zext i8 %2 to i32
38  %add16.i = or i32 %mul12.i, 0
39  %add17.i = or i32 %add16.i, 0
40  %shr18.i = lshr i32 %add17.i, 2
41  %conv19.i = trunc i32 %shr18.i to i8
42  %mul25.i = mul i32 %conv15.i, 0
43  %arrayidx27.i = getelementptr i8, ptr %top, i64 3
44  %3 = load i8, ptr %arrayidx27.i, align 1
45  %conv28.i = zext i8 %3 to i32
46  %add29.i = or i32 %mul25.i, 0
47  %add30.i = or i32 %add29.i, 0
48  %shr31.i = lshr i32 %add30.i, 2
49  %conv32.i = trunc i32 %shr31.i to i8
50  %mul38.i = mul i32 %conv28.i, 0
51  %add39.i = or i32 %mul38.i, %conv15.i
52  %add42.i = or i32 %add39.i, 0
53  %shr44.i = lshr i32 %add42.i, 2
54  %conv45.i = trunc i32 %shr44.i to i8
55  br label %for.cond.i
56
57for.cond.i:
58  store i8 %conv7.i, ptr null, align 1
59  %vals.sroa.5.0.add.ptr.sroa_idx.i = getelementptr i8, ptr null, i64 1
60  store i8 %conv19.i, ptr %vals.sroa.5.0.add.ptr.sroa_idx.i, align 1
61  %vals.sroa.7.0.add.ptr.sroa_idx.i = getelementptr i8, ptr null, i64 2
62  store i8 %conv32.i, ptr %vals.sroa.7.0.add.ptr.sroa_idx.i, align 1
63  %vals.sroa.9.0.add.ptr.sroa_idx.i = getelementptr i8, ptr null, i64 3
64  store i8 %conv45.i, ptr %vals.sroa.9.0.add.ptr.sroa_idx.i, align 1
65  br label %for.cond.i
66}
67