1; RUN: opt -passes='require<profile-summary>,function(codegenprepare)' -passes=load-store-vectorizer %s -S -o - | FileCheck %s
2; RUN: opt                 -passes=load-store-vectorizer %s -S -o - | FileCheck %s
3; RUN: opt -aa-pipeline=basic-aa -passes='function(load-store-vectorizer)' %s -S -o - | FileCheck %s
4
5target triple = "x86_64--"
6
7%union = type { { [4 x [4 x [4 x [16 x float]]]], [4 x [4 x [4 x [16 x float]]]], [10 x [10 x [4 x float]]] } }
8
9@global_pointer = external unnamed_addr global { %union, [2000 x i8] }, align 4
10
11; Function Attrs: convergent nounwind
12define void @test(i32 %base) #0 {
13; CHECK-LABEL: @test(
14; CHECK-NOT: load i32
15; CHECK: load <2 x i32>
16; CHECK-NOT: load i32
17entry:
18  %mul331 = and i32 %base, -4
19  %add350.4 = add i32 4, %mul331
20  %idx351.4 = zext i32 %add350.4 to i64
21  %arrayidx352.4 = getelementptr inbounds { %union, [2000 x i8] }, ptr @global_pointer, i64 0, i32 0, i32 0, i32 1, i64 0, i64 0, i64 0, i64 %idx351.4
22  %add350.5 = add i32 5, %mul331
23  %idx351.5 = zext i32 %add350.5 to i64
24  %arrayidx352.5 = getelementptr inbounds { %union, [2000 x i8] }, ptr @global_pointer, i64 0, i32 0, i32 0, i32 1, i64 0, i64 0, i64 0, i64 %idx351.5
25  %cnd = icmp ult i32 %base, 1000
26  br i1 %cnd, label %loads, label %exit
27
28loads:
29  ; If and only if the loads are in a different BB from the GEPs codegenprepare
30  ; would try to turn the GEPs into math, which makes LoadStoreVectorizer's job
31  ; harder
32  %tmp297.4 = load i32, ptr %arrayidx352.4, align 4, !tbaa !0
33  %tmp297.5 = load i32, ptr %arrayidx352.5, align 4, !tbaa !0
34  br label %exit
35
36exit:
37  ret void
38}
39
40; Function Attrs: convergent nounwind
41define void @test.codegenprepared(i32 %base) #0 {
42; CHECK-LABEL: @test.codegenprepared(
43; CHECK-NOT: load i32
44; CHECK: load <2 x i32>
45; CHECK-NOT: load i32
46entry:
47  %mul331 = and i32 %base, -4
48  %add350.4 = add i32 4, %mul331
49  %idx351.4 = zext i32 %add350.4 to i64
50  %add350.5 = add i32 5, %mul331
51  %idx351.5 = zext i32 %add350.5 to i64
52  %cnd = icmp ult i32 %base, 1000
53  br i1 %cnd, label %loads, label %exit
54
55loads:                                            ; preds = %entry
56  %sunkaddr = mul i64 %idx351.4, 4
57  %sunkaddr1 = getelementptr inbounds i8, ptr @global_pointer, i64 %sunkaddr
58  %sunkaddr2 = getelementptr inbounds i8, ptr %sunkaddr1, i64 4096
59  %tmp297.4 = load i32, ptr %sunkaddr2, align 4, !tbaa !0
60  %sunkaddr3 = mul i64 %idx351.5, 4
61  %sunkaddr4 = getelementptr inbounds i8, ptr @global_pointer, i64 %sunkaddr3
62  %sunkaddr5 = getelementptr inbounds i8, ptr %sunkaddr4, i64 4096
63  %tmp297.5 = load i32, ptr %sunkaddr5, align 4, !tbaa !0
64  br label %exit
65
66exit:                                             ; preds = %loads, %entry
67  ret void
68}
69
70attributes #0 = { convergent nounwind }
71
72!0 = !{!1, !1, i64 0}
73!1 = !{!"float", !2, i64 0}
74!2 = !{!"omnipotent char", !3, i64 0}
75!3 = !{!"Simple C++ TBAA"}
76