1 ; NOTE: Assertions have been autogenerated by utils/update_analyze_test_checks.py UTC_ARGS: --filter "LV: Found an estimated cost of [0-9]+ for VF [0-9]+ For instruction:\s*%v. = load i32, ptr %in."
2 ; RUN: opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+sse2 --debug-only=loop-vectorize < %s 2>&1 | FileCheck %s --check-prefix=SSE2
3 ; RUN: opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx --debug-only=loop-vectorize < %s 2>&1 | FileCheck %s --check-prefix=AVX1
4 ; RUN: opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx2 --debug-only=loop-vectorize < %s 2>&1 | FileCheck %s --check-prefix=AVX2
5 ; RUN: opt -passes=loop-vectorize -vectorizer-maximize-bandwidth -S -mattr=+avx512vl --debug-only=loop-vectorize < %s 2>&1 | FileCheck %s --check-prefix=AVX512
8 target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
9 target triple = "x86_64-unknown-linux-gnu"
11 @A = global [1024 x i32] zeroinitializer, align 128
12 @B = global [1024 x i8] zeroinitializer, align 128
16 ; SSE2: LV: Found an estimated cost of 1 for VF 1 For instruction: %v0 = load i32, ptr %in0, align 4
17 ; SSE2: LV: Found an estimated cost of 1 for VF 1 For instruction: %v1 = load i32, ptr %in1, align 4
18 ; SSE2: LV: Found an estimated cost of 1 for VF 1 For instruction: %v0 = load i32, ptr %in0, align 4
19 ; SSE2: LV: Found an estimated cost of 1 for VF 1 For instruction: %v1 = load i32, ptr %in1, align 4
20 ; SSE2: LV: Found an estimated cost of 14 for VF 2 For instruction: %v0 = load i32, ptr %in0, align 4
21 ; SSE2: LV: Found an estimated cost of 0 for VF 2 For instruction: %v1 = load i32, ptr %in1, align 4
22 ; SSE2: LV: Found an estimated cost of 30 for VF 4 For instruction: %v0 = load i32, ptr %in0, align 4
23 ; SSE2: LV: Found an estimated cost of 0 for VF 4 For instruction: %v1 = load i32, ptr %in1, align 4
24 ; SSE2: LV: Found an estimated cost of 60 for VF 8 For instruction: %v0 = load i32, ptr %in0, align 4
25 ; SSE2: LV: Found an estimated cost of 0 for VF 8 For instruction: %v1 = load i32, ptr %in1, align 4
26 ; SSE2: LV: Found an estimated cost of 120 for VF 16 For instruction: %v0 = load i32, ptr %in0, align 4
27 ; SSE2: LV: Found an estimated cost of 0 for VF 16 For instruction: %v1 = load i32, ptr %in1, align 4
30 ; AVX1: LV: Found an estimated cost of 1 for VF 1 For instruction: %v0 = load i32, ptr %in0, align 4
31 ; AVX1: LV: Found an estimated cost of 1 for VF 1 For instruction: %v1 = load i32, ptr %in1, align 4
32 ; AVX1: LV: Found an estimated cost of 1 for VF 1 For instruction: %v0 = load i32, ptr %in0, align 4
33 ; AVX1: LV: Found an estimated cost of 1 for VF 1 For instruction: %v1 = load i32, ptr %in1, align 4
34 ; AVX1: LV: Found an estimated cost of 10 for VF 2 For instruction: %v0 = load i32, ptr %in0, align 4
35 ; AVX1: LV: Found an estimated cost of 0 for VF 2 For instruction: %v1 = load i32, ptr %in1, align 4
36 ; AVX1: LV: Found an estimated cost of 20 for VF 4 For instruction: %v0 = load i32, ptr %in0, align 4
37 ; AVX1: LV: Found an estimated cost of 0 for VF 4 For instruction: %v1 = load i32, ptr %in1, align 4
38 ; AVX1: LV: Found an estimated cost of 42 for VF 8 For instruction: %v0 = load i32, ptr %in0, align 4
39 ; AVX1: LV: Found an estimated cost of 0 for VF 8 For instruction: %v1 = load i32, ptr %in1, align 4
40 ; AVX1: LV: Found an estimated cost of 84 for VF 16 For instruction: %v0 = load i32, ptr %in0, align 4
41 ; AVX1: LV: Found an estimated cost of 0 for VF 16 For instruction: %v1 = load i32, ptr %in1, align 4
42 ; AVX1: LV: Found an estimated cost of 168 for VF 32 For instruction: %v0 = load i32, ptr %in0, align 4
43 ; AVX1: LV: Found an estimated cost of 0 for VF 32 For instruction: %v1 = load i32, ptr %in1, align 4
46 ; AVX2: LV: Found an estimated cost of 1 for VF 1 For instruction: %v0 = load i32, ptr %in0, align 4
47 ; AVX2: LV: Found an estimated cost of 1 for VF 1 For instruction: %v1 = load i32, ptr %in1, align 4
48 ; AVX2: LV: Found an estimated cost of 1 for VF 1 For instruction: %v0 = load i32, ptr %in0, align 4
49 ; AVX2: LV: Found an estimated cost of 1 for VF 1 For instruction: %v1 = load i32, ptr %in1, align 4
50 ; AVX2: LV: Found an estimated cost of 3 for VF 2 For instruction: %v0 = load i32, ptr %in0, align 4
51 ; AVX2: LV: Found an estimated cost of 0 for VF 2 For instruction: %v1 = load i32, ptr %in1, align 4
52 ; AVX2: LV: Found an estimated cost of 6 for VF 4 For instruction: %v0 = load i32, ptr %in0, align 4
53 ; AVX2: LV: Found an estimated cost of 0 for VF 4 For instruction: %v1 = load i32, ptr %in1, align 4
54 ; AVX2: LV: Found an estimated cost of 12 for VF 8 For instruction: %v0 = load i32, ptr %in0, align 4
55 ; AVX2: LV: Found an estimated cost of 0 for VF 8 For instruction: %v1 = load i32, ptr %in1, align 4
56 ; AVX2: LV: Found an estimated cost of 24 for VF 16 For instruction: %v0 = load i32, ptr %in0, align 4
57 ; AVX2: LV: Found an estimated cost of 0 for VF 16 For instruction: %v1 = load i32, ptr %in1, align 4
58 ; AVX2: LV: Found an estimated cost of 50 for VF 32 For instruction: %v0 = load i32, ptr %in0, align 4
59 ; AVX2: LV: Found an estimated cost of 0 for VF 32 For instruction: %v1 = load i32, ptr %in1, align 4
61 ; AVX512-LABEL: 'test'
62 ; AVX512: LV: Found an estimated cost of 1 for VF 1 For instruction: %v0 = load i32, ptr %in0, align 4
63 ; AVX512: LV: Found an estimated cost of 1 for VF 1 For instruction: %v1 = load i32, ptr %in1, align 4
64 ; AVX512: LV: Found an estimated cost of 1 for VF 1 For instruction: %v0 = load i32, ptr %in0, align 4
65 ; AVX512: LV: Found an estimated cost of 1 for VF 1 For instruction: %v1 = load i32, ptr %in1, align 4
66 ; AVX512: LV: Found an estimated cost of 3 for VF 2 For instruction: %v0 = load i32, ptr %in0, align 4
67 ; AVX512: LV: Found an estimated cost of 0 for VF 2 For instruction: %v1 = load i32, ptr %in1, align 4
68 ; AVX512: LV: Found an estimated cost of 3 for VF 4 For instruction: %v0 = load i32, ptr %in0, align 4
69 ; AVX512: LV: Found an estimated cost of 0 for VF 4 For instruction: %v1 = load i32, ptr %in1, align 4
70 ; AVX512: LV: Found an estimated cost of 5 for VF 8 For instruction: %v0 = load i32, ptr %in0, align 4
71 ; AVX512: LV: Found an estimated cost of 0 for VF 8 For instruction: %v1 = load i32, ptr %in1, align 4
72 ; AVX512: LV: Found an estimated cost of 13 for VF 16 For instruction: %v0 = load i32, ptr %in0, align 4
73 ; AVX512: LV: Found an estimated cost of 0 for VF 16 For instruction: %v1 = load i32, ptr %in1, align 4
74 ; AVX512: LV: Found an estimated cost of 50 for VF 32 For instruction: %v0 = load i32, ptr %in0, align 4
75 ; AVX512: LV: Found an estimated cost of 0 for VF 32 For instruction: %v1 = load i32, ptr %in1, align 4
76 ; AVX512: LV: Found an estimated cost of 80 for VF 64 For instruction: %v0 = load i32, ptr %in0, align 4
77 ; AVX512: LV: Found an estimated cost of 80 for VF 64 For instruction: %v1 = load i32, ptr %in1, align 4
83 %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
85 %iv.0 = add nuw nsw i64 %iv, 0
86 %iv.1 = add nuw nsw i64 %iv, 1
88 %in0 = getelementptr inbounds [1024 x i32], ptr @A, i64 0, i64 %iv.0
89 %in1 = getelementptr inbounds [1024 x i32], ptr @A, i64 0, i64 %iv.1
91 %v0 = load i32, ptr %in0
92 %v1 = load i32, ptr %in1
94 %reduce.add.0 = add i32 %v0, %v1
95 %reduce.add.1 = add i32 %reduce.add.0, 0
96 %reduce.add.2 = add i32 %reduce.add.1, 0
98 %reduce.add.2.narrow = trunc i32 %reduce.add.2 to i8
100 %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0
101 store i8 %reduce.add.2.narrow, ptr %out
103 %iv.next = add nuw nsw i64 %iv, 4
104 %cmp = icmp ult i64 %iv.next, 1024
105 br i1 %cmp, label %for.body, label %for.cond.cleanup