1; RUN: opt -codegenprepare -load-store-vectorizer %s -S -o - | FileCheck %s
2; RUN: opt                 -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] }, { %union, [2000 x i8] }* @global_pointer, i64 0, i32 0, i32 0, i32 1, i64 0, i64 0, i64 0, i64 %idx351.4
22  %tmp296.4 = bitcast float* %arrayidx352.4 to i32*
23  %add350.5 = add i32 5, %mul331
24  %idx351.5 = zext i32 %add350.5 to i64
25  %arrayidx352.5 = getelementptr inbounds { %union, [2000 x i8] }, { %union, [2000 x i8] }* @global_pointer, i64 0, i32 0, i32 0, i32 1, i64 0, i64 0, i64 0, i64 %idx351.5
26  %tmp296.5 = bitcast float* %arrayidx352.5 to i32*
27  %cnd = icmp ult i32 %base, 1000
28  br i1 %cnd, label %loads, label %exit
29
30loads:
31  ; If and only if the loads are in a different BB from the GEPs codegenprepare
32  ; would try to turn the GEPs into math, which makes LoadStoreVectorizer's job
33  ; harder
34  %tmp297.4 = load i32, i32* %tmp296.4, align 4, !tbaa !0
35  %tmp297.5 = load i32, i32* %tmp296.5, align 4, !tbaa !0
36  br label %exit
37
38exit:
39  ret void
40}
41
42; Function Attrs: convergent nounwind
43define void @test.codegenprepared(i32 %base) #0 {
44; CHECK-LABEL: @test.codegenprepared(
45; CHECK-NOT: load i32
46; CHECK: load <2 x i32>
47; CHECK-NOT: load i32
48entry:
49  %mul331 = and i32 %base, -4
50  %add350.4 = add i32 4, %mul331
51  %idx351.4 = zext i32 %add350.4 to i64
52  %add350.5 = add i32 5, %mul331
53  %idx351.5 = zext i32 %add350.5 to i64
54  %cnd = icmp ult i32 %base, 1000
55  br i1 %cnd, label %loads, label %exit
56
57loads:                                            ; preds = %entry
58  %sunkaddr = mul i64 %idx351.4, 4
59  %sunkaddr1 = getelementptr inbounds i8, i8* bitcast ({ %union, [2000 x i8] }* @global_pointer to i8*), i64 %sunkaddr
60  %sunkaddr2 = getelementptr inbounds i8, i8* %sunkaddr1, i64 4096
61  %0 = bitcast i8* %sunkaddr2 to i32*
62  %tmp297.4 = load i32, i32* %0, align 4, !tbaa !0
63  %sunkaddr3 = mul i64 %idx351.5, 4
64  %sunkaddr4 = getelementptr inbounds i8, i8* bitcast ({ %union, [2000 x i8] }* @global_pointer to i8*), i64 %sunkaddr3
65  %sunkaddr5 = getelementptr inbounds i8, i8* %sunkaddr4, i64 4096
66  %1 = bitcast i8* %sunkaddr5 to i32*
67  %tmp297.5 = load i32, i32* %1, align 4, !tbaa !0
68  br label %exit
69
70exit:                                             ; preds = %loads, %entry
71  ret void
72}
73
74attributes #0 = { convergent nounwind }
75
76!0 = !{!1, !1, i64 0}
77!1 = !{!"float", !2, i64 0}
78!2 = !{!"omnipotent char", !3, i64 0}
79!3 = !{!"Simple C++ TBAA"}
80