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 -codegenprepare -aa-pipeline=basic-aa -passes='function(load-store-vectorizer)' %s -S -o - | FileCheck %s
4 ; RUN: opt -aa-pipeline=basic-aa -passes='function(load-store-vectorizer)' %s -S -o - | FileCheck %s
6 target triple = "x86_64--"
8 %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]]] } }
10 @global_pointer = external unnamed_addr global { %union, [2000 x i8] }, align 4
12 ; Function Attrs: convergent nounwind
13 define void @test(i32 %base) #0 {
16 ; CHECK: load <2 x i32>
19 %mul331 = and i32 %base, -4
20 %add350.4 = add i32 4, %mul331
21 %idx351.4 = zext i32 %add350.4 to i64
22 %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
23 %tmp296.4 = bitcast float* %arrayidx352.4 to i32*
24 %add350.5 = add i32 5, %mul331
25 %idx351.5 = zext i32 %add350.5 to i64
26 %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
27 %tmp296.5 = bitcast float* %arrayidx352.5 to i32*
28 %cnd = icmp ult i32 %base, 1000
29 br i1 %cnd, label %loads, label %exit
32 ; If and only if the loads are in a different BB from the GEPs codegenprepare
33 ; would try to turn the GEPs into math, which makes LoadStoreVectorizer's job
35 %tmp297.4 = load i32, i32* %tmp296.4, align 4, !tbaa !0
36 %tmp297.5 = load i32, i32* %tmp296.5, align 4, !tbaa !0
43 ; Function Attrs: convergent nounwind
44 define void @test.codegenprepared(i32 %base) #0 {
45 ; CHECK-LABEL: @test.codegenprepared(
47 ; CHECK: load <2 x i32>
50 %mul331 = and i32 %base, -4
51 %add350.4 = add i32 4, %mul331
52 %idx351.4 = zext i32 %add350.4 to i64
53 %add350.5 = add i32 5, %mul331
54 %idx351.5 = zext i32 %add350.5 to i64
55 %cnd = icmp ult i32 %base, 1000
56 br i1 %cnd, label %loads, label %exit
58 loads: ; preds = %entry
59 %sunkaddr = mul i64 %idx351.4, 4
60 %sunkaddr1 = getelementptr inbounds i8, i8* bitcast ({ %union, [2000 x i8] }* @global_pointer to i8*), i64 %sunkaddr
61 %sunkaddr2 = getelementptr inbounds i8, i8* %sunkaddr1, i64 4096
62 %0 = bitcast i8* %sunkaddr2 to i32*
63 %tmp297.4 = load i32, i32* %0, align 4, !tbaa !0
64 %sunkaddr3 = mul i64 %idx351.5, 4
65 %sunkaddr4 = getelementptr inbounds i8, i8* bitcast ({ %union, [2000 x i8] }* @global_pointer to i8*), i64 %sunkaddr3
66 %sunkaddr5 = getelementptr inbounds i8, i8* %sunkaddr4, i64 4096
67 %1 = bitcast i8* %sunkaddr5 to i32*
68 %tmp297.5 = load i32, i32* %1, align 4, !tbaa !0
71 exit: ; preds = %loads, %entry
75 attributes #0 = { convergent nounwind }
78 !1 = !{!"float", !2, i64 0}
79 !2 = !{!"omnipotent char", !3, i64 0}
80 !3 = !{!"Simple C++ TBAA"}