2 ; RUN: opt -loop-vectorize -S -mattr=avx512f --debug-only=loop-vectorize < %s 2>&1 | FileCheck %s
4 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
5 target triple = "x86_64-unknown-linux-gnu"
7 @A = global [10240 x i32] zeroinitializer, align 16
8 @B = global [10240 x i32] zeroinitializer, align 16
10 ; Function Attrs: nounwind uwtable
11 define void @load_i32_interleave4() {
12 ;CHECK-LABEL: load_i32_interleave4
13 ;CHECK: Found an estimated cost of 1 for VF 1 For instruction: %0 = load
14 ;CHECK: Found an estimated cost of 5 for VF 2 For instruction: %0 = load
15 ;CHECK: Found an estimated cost of 5 for VF 4 For instruction: %0 = load
16 ;CHECK: Found an estimated cost of 8 for VF 8 For instruction: %0 = load
17 ;CHECK: Found an estimated cost of 22 for VF 16 For instruction: %0 = load
21 for.cond.cleanup: ; preds = %for.body
24 for.body: ; preds = %entry, %for.body
25 %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
26 %arrayidx = getelementptr inbounds [10240 x i32], [10240 x i32]* @A, i64 0, i64 %indvars.iv
27 %0 = load i32, i32* %arrayidx, align 16
28 %1 = or i64 %indvars.iv, 1
29 %arrayidx2 = getelementptr inbounds [10240 x i32], [10240 x i32]* @A, i64 0, i64 %1
30 %2 = load i32, i32* %arrayidx2, align 4
31 %add3 = add nsw i32 %2, %0
32 %3 = or i64 %indvars.iv, 2
33 %arrayidx6 = getelementptr inbounds [10240 x i32], [10240 x i32]* @A, i64 0, i64 %3
34 %4 = load i32, i32* %arrayidx6, align 8
35 %add7 = add nsw i32 %add3, %4
36 %5 = or i64 %indvars.iv, 3
37 %arrayidx10 = getelementptr inbounds [10240 x i32], [10240 x i32]* @A, i64 0, i64 %5
38 %6 = load i32, i32* %arrayidx10, align 4
39 %add11 = add nsw i32 %add7, %6
40 %arrayidx13 = getelementptr inbounds [10240 x i32], [10240 x i32]* @B, i64 0, i64 %indvars.iv
41 store i32 %add11, i32* %arrayidx13, align 16
42 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 4
43 %cmp = icmp slt i64 %indvars.iv.next, 1024
44 br i1 %cmp, label %for.body, label %for.cond.cleanup
47 define void @load_i32_interleave5() {
48 ;CHECK-LABEL: load_i32_interleave5
49 ;CHECK: Found an estimated cost of 1 for VF 1 For instruction: %0 = load
50 ;CHECK: Found an estimated cost of 6 for VF 2 For instruction: %0 = load
51 ;CHECK: Found an estimated cost of 9 for VF 4 For instruction: %0 = load
52 ;CHECK: Found an estimated cost of 18 for VF 8 For instruction: %0 = load
53 ;CHECK: Found an estimated cost of 35 for VF 16 For instruction: %0 = load
57 for.cond.cleanup: ; preds = %for.body
60 for.body: ; preds = %entry, %for.body
61 %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
62 %arrayidx = getelementptr inbounds [10240 x i32], [10240 x i32]* @A, i64 0, i64 %indvars.iv
63 %0 = load i32, i32* %arrayidx, align 4
64 %1 = add nuw nsw i64 %indvars.iv, 1
65 %arrayidx2 = getelementptr inbounds [10240 x i32], [10240 x i32]* @A, i64 0, i64 %1
66 %2 = load i32, i32* %arrayidx2, align 4
67 %add3 = add nsw i32 %2, %0
68 %3 = add nuw nsw i64 %indvars.iv, 2
69 %arrayidx6 = getelementptr inbounds [10240 x i32], [10240 x i32]* @A, i64 0, i64 %3
70 %4 = load i32, i32* %arrayidx6, align 4
71 %add7 = add nsw i32 %add3, %4
72 %5 = add nuw nsw i64 %indvars.iv, 3
73 %arrayidx10 = getelementptr inbounds [10240 x i32], [10240 x i32]* @A, i64 0, i64 %5
74 %6 = load i32, i32* %arrayidx10, align 4
75 %add11 = add nsw i32 %add7, %6
76 %7 = add nuw nsw i64 %indvars.iv, 4
77 %arrayidx14 = getelementptr inbounds [10240 x i32], [10240 x i32]* @A, i64 0, i64 %7
78 %8 = load i32, i32* %arrayidx14, align 4
79 %add15 = add nsw i32 %add11, %8
80 %arrayidx17 = getelementptr inbounds [10240 x i32], [10240 x i32]* @B, i64 0, i64 %indvars.iv
81 store i32 %add15, i32* %arrayidx17, align 4
82 %indvars.iv.next = add nuw nsw i64 %indvars.iv, 5
83 %cmp = icmp slt i64 %indvars.iv.next, 1024
84 br i1 %cmp, label %for.body, label %for.cond.cleanup