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*store float %v7, ptr %out7"
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 i8] zeroinitializer, align 128
12 @B = global [1024 x float] zeroinitializer, align 128
16 ; SSE2: LV: Found an estimated cost of 1 for VF 1 For instruction: store float %v7, ptr %out7, align 4
17 ; SSE2: LV: Found an estimated cost of 1 for VF 1 For instruction: store float %v7, ptr %out7, align 4
18 ; SSE2: LV: Found an estimated cost of 24 for VF 2 For instruction: store float %v7, ptr %out7, align 4
19 ; SSE2: LV: Found an estimated cost of 56 for VF 4 For instruction: store float %v7, ptr %out7, align 4
20 ; SSE2: LV: Found an estimated cost of 112 for VF 8 For instruction: store float %v7, ptr %out7, align 4
23 ; AVX1: LV: Found an estimated cost of 1 for VF 1 For instruction: store float %v7, ptr %out7, align 4
24 ; AVX1: LV: Found an estimated cost of 1 for VF 1 For instruction: store float %v7, ptr %out7, align 4
25 ; AVX1: LV: Found an estimated cost of 24 for VF 2 For instruction: store float %v7, ptr %out7, align 4
26 ; AVX1: LV: Found an estimated cost of 56 for VF 4 For instruction: store float %v7, ptr %out7, align 4
27 ; AVX1: LV: Found an estimated cost of 120 for VF 8 For instruction: store float %v7, ptr %out7, align 4
28 ; AVX1: LV: Found an estimated cost of 240 for VF 16 For instruction: store float %v7, ptr %out7, align 4
31 ; AVX2: LV: Found an estimated cost of 1 for VF 1 For instruction: store float %v7, ptr %out7, align 4
32 ; AVX2: LV: Found an estimated cost of 1 for VF 1 For instruction: store float %v7, ptr %out7, align 4
33 ; AVX2: LV: Found an estimated cost of 24 for VF 2 For instruction: store float %v7, ptr %out7, align 4
34 ; AVX2: LV: Found an estimated cost of 56 for VF 4 For instruction: store float %v7, ptr %out7, align 4
35 ; AVX2: LV: Found an estimated cost of 120 for VF 8 For instruction: store float %v7, ptr %out7, align 4
36 ; AVX2: LV: Found an estimated cost of 240 for VF 16 For instruction: store float %v7, ptr %out7, align 4
38 ; AVX512-LABEL: 'test'
39 ; AVX512: LV: Found an estimated cost of 1 for VF 1 For instruction: store float %v7, ptr %out7, align 4
40 ; AVX512: LV: Found an estimated cost of 1 for VF 1 For instruction: store float %v7, ptr %out7, align 4
41 ; AVX512: LV: Found an estimated cost of 11 for VF 2 For instruction: store float %v7, ptr %out7, align 4
42 ; AVX512: LV: Found an estimated cost of 23 for VF 4 For instruction: store float %v7, ptr %out7, align 4
43 ; AVX512: LV: Found an estimated cost of 46 for VF 8 For instruction: store float %v7, ptr %out7, align 4
44 ; AVX512: LV: Found an estimated cost of 92 for VF 16 For instruction: store float %v7, ptr %out7, align 4
45 ; AVX512: LV: Found an estimated cost of 184 for VF 32 For instruction: store float %v7, ptr %out7, align 4
51 %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
53 %iv.0 = add nuw nsw i64 %iv, 0
54 %iv.1 = add nuw nsw i64 %iv, 1
55 %iv.2 = add nuw nsw i64 %iv, 2
56 %iv.3 = add nuw nsw i64 %iv, 3
57 %iv.4 = add nuw nsw i64 %iv, 4
58 %iv.5 = add nuw nsw i64 %iv, 5
59 %iv.6 = add nuw nsw i64 %iv, 6
60 %iv.7 = add nuw nsw i64 %iv, 7
62 %in = getelementptr inbounds [1024 x i8], ptr @A, i64 0, i64 %iv.0
63 %v.narrow = load i8, ptr %in
65 %v = uitofp i8 %v.narrow to float
67 %v0 = fadd float %v, 0.0
68 %v1 = fadd float %v, 1.0
69 %v2 = fadd float %v, 2.0
70 %v3 = fadd float %v, 3.0
71 %v4 = fadd float %v, 4.0
72 %v5 = fadd float %v, 5.0
73 %v6 = fadd float %v, 6.0
74 %v7 = fadd float %v, 7.0
76 %out0 = getelementptr inbounds [1024 x float], ptr @B, i64 0, i64 %iv.0
77 %out1 = getelementptr inbounds [1024 x float], ptr @B, i64 0, i64 %iv.1
78 %out2 = getelementptr inbounds [1024 x float], ptr @B, i64 0, i64 %iv.2
79 %out3 = getelementptr inbounds [1024 x float], ptr @B, i64 0, i64 %iv.3
80 %out4 = getelementptr inbounds [1024 x float], ptr @B, i64 0, i64 %iv.4
81 %out5 = getelementptr inbounds [1024 x float], ptr @B, i64 0, i64 %iv.5
82 %out6 = getelementptr inbounds [1024 x float], ptr @B, i64 0, i64 %iv.6
83 %out7 = getelementptr inbounds [1024 x float], ptr @B, i64 0, i64 %iv.7
85 store float %v0, ptr %out0
86 store float %v1, ptr %out1
87 store float %v2, ptr %out2
88 store float %v3, ptr %out3
89 store float %v4, ptr %out4
90 store float %v5, ptr %out5
91 store float %v6, ptr %out6
92 store float %v7, ptr %out7
94 %iv.next = add nuw nsw i64 %iv.0, 8
95 %cmp = icmp ult i64 %iv.next, 1024
96 br i1 %cmp, label %for.body, label %for.cond.cleanup