xref: /llvm-project/llvm/test/Transforms/SLPVectorizer/X86/opaque-ptr.ll (revision c7657cf7d1ee57f9cb9133164536591a1842b43c)
1; NOTE: Assertions have been autogenerated by utils/update_test_checks.py
2; RUN: opt -S -passes=slp-vectorizer -mtriple=x86_64-apple-macosx -mcpu=haswell < %s | FileCheck %s
3
4define void @test(ptr %r, ptr %p, ptr %q) #0 {
5; CHECK-LABEL: @test(
6; CHECK-NEXT:    [[P0:%.*]] = getelementptr inbounds i64, ptr [[P:%.*]], i64 0
7; CHECK-NEXT:    [[Q0:%.*]] = getelementptr inbounds i64, ptr [[Q:%.*]], i64 0
8; CHECK-NEXT:    [[TMP1:%.*]] = load <4 x i64>, ptr [[P0]], align 2
9; CHECK-NEXT:    [[TMP2:%.*]] = load <4 x i64>, ptr [[Q0]], align 2
10; CHECK-NEXT:    [[TMP3:%.*]] = sub nsw <4 x i64> [[TMP1]], [[TMP2]]
11; CHECK-NEXT:    [[TMP4:%.*]] = extractelement <4 x i64> [[TMP3]], i32 0
12; CHECK-NEXT:    [[G0:%.*]] = getelementptr inbounds i32, ptr [[R:%.*]], i64 [[TMP4]]
13; CHECK-NEXT:    [[TMP5:%.*]] = extractelement <4 x i64> [[TMP3]], i32 1
14; CHECK-NEXT:    [[G1:%.*]] = getelementptr inbounds i32, ptr [[R]], i64 [[TMP5]]
15; CHECK-NEXT:    [[TMP6:%.*]] = extractelement <4 x i64> [[TMP3]], i32 2
16; CHECK-NEXT:    [[G2:%.*]] = getelementptr inbounds i32, ptr [[R]], i64 [[TMP6]]
17; CHECK-NEXT:    [[TMP7:%.*]] = extractelement <4 x i64> [[TMP3]], i32 3
18; CHECK-NEXT:    [[G3:%.*]] = getelementptr inbounds i32, ptr [[R]], i64 [[TMP7]]
19; CHECK-NEXT:    ret void
20;
21  %p0 = getelementptr inbounds i64, ptr %p, i64 0
22  %p1 = getelementptr inbounds i64, ptr %p, i64 1
23  %p2 = getelementptr inbounds i64, ptr %p, i64 2
24  %p3 = getelementptr inbounds i64, ptr %p, i64 3
25
26  %q0 = getelementptr inbounds i64, ptr %q, i64 0
27  %q1 = getelementptr inbounds i64, ptr %q, i64 1
28  %q2 = getelementptr inbounds i64, ptr %q, i64 2
29  %q3 = getelementptr inbounds i64, ptr %q, i64 3
30
31  %x0 = load i64, ptr %p0, align 2
32  %x1 = load i64, ptr %p1, align 2
33  %x2 = load i64, ptr %p2, align 2
34  %x3 = load i64, ptr %p3, align 2
35
36  %y0 = load i64, ptr %q0, align 2
37  %y1 = load i64, ptr %q1, align 2
38  %y2 = load i64, ptr %q2, align 2
39  %y3 = load i64, ptr %q3, align 2
40
41  %sub0 = sub nsw i64 %x0, %y0
42  %sub1 = sub nsw i64 %x1, %y1
43  %sub2 = sub nsw i64 %x2, %y2
44  %sub3 = sub nsw i64 %x3, %y3
45
46  %g0 = getelementptr inbounds i32, ptr %r, i64 %sub0
47  %g1 = getelementptr inbounds i32, ptr %r, i64 %sub1
48  %g2 = getelementptr inbounds i32, ptr %r, i64 %sub2
49  %g3 = getelementptr inbounds i32, ptr %r, i64 %sub3
50  ret void
51}
52
53define void @test2(ptr %a, ptr %b) {
54; CHECK-LABEL: @test2(
55; CHECK-NEXT:    [[TMP1:%.*]] = insertelement <2 x ptr> poison, ptr [[A:%.*]], i32 0
56; CHECK-NEXT:    [[TMP2:%.*]] = insertelement <2 x ptr> [[TMP1]], ptr [[B:%.*]], i32 1
57; CHECK-NEXT:    [[TMP3:%.*]] = getelementptr i64, <2 x ptr> [[TMP2]], <2 x i64> <i64 1, i64 3>
58; CHECK-NEXT:    [[A1:%.*]] = getelementptr inbounds i64, ptr [[A]], i64 1
59; CHECK-NEXT:    [[TMP4:%.*]] = ptrtoint <2 x ptr> [[TMP3]] to <2 x i64>
60; CHECK-NEXT:    [[TMP5:%.*]] = load <2 x i64>, ptr [[A1]], align 8
61; CHECK-NEXT:    [[TMP6:%.*]] = add <2 x i64> [[TMP4]], [[TMP5]]
62; CHECK-NEXT:    store <2 x i64> [[TMP6]], ptr [[A1]], align 8
63; CHECK-NEXT:    ret void
64;
65  %a1 = getelementptr inbounds i64, ptr %a, i64 1
66  %a2 = getelementptr inbounds i64, ptr %a, i64 2
67  %i1 = ptrtoint ptr %a1 to i64
68  %b3 = getelementptr inbounds i64, ptr %b, i64 3
69  %i2 = ptrtoint ptr %b3 to i64
70  %v1 = load i64, ptr %a1, align 8
71  %v2 = load i64, ptr %a2, align 8
72  %add1 = add i64 %i1, %v1
73  %add2 = add i64 %i2, %v2
74  store i64 %add1, ptr %a1, align 8
75  store i64 %add2, ptr %a2, align 8
76  ret void
77}
78