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 double %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 double] zeroinitializer, align 128
16 ; SSE2: LV: Found an estimated cost of 1 for VF 1 For instruction: store double %v7, ptr %out7, align 8
17 ; SSE2: LV: Found an estimated cost of 24 for VF 2 For instruction: store double %v7, ptr %out7, align 8
18 ; SSE2: LV: Found an estimated cost of 48 for VF 4 For instruction: store double %v7, ptr %out7, align 8
21 ; AVX1: LV: Found an estimated cost of 1 for VF 1 For instruction: store double %v7, ptr %out7, align 8
22 ; AVX1: LV: Found an estimated cost of 24 for VF 2 For instruction: store double %v7, ptr %out7, align 8
23 ; AVX1: LV: Found an estimated cost of 56 for VF 4 For instruction: store double %v7, ptr %out7, align 8
24 ; AVX1: LV: Found an estimated cost of 112 for VF 8 For instruction: store double %v7, ptr %out7, align 8
27 ; AVX2: LV: Found an estimated cost of 1 for VF 1 For instruction: store double %v7, ptr %out7, align 8
28 ; AVX2: LV: Found an estimated cost of 24 for VF 2 For instruction: store double %v7, ptr %out7, align 8
29 ; AVX2: LV: Found an estimated cost of 56 for VF 4 For instruction: store double %v7, ptr %out7, align 8
30 ; AVX2: LV: Found an estimated cost of 112 for VF 8 For instruction: store double %v7, ptr %out7, align 8
32 ; AVX512-LABEL: 'test'
33 ; AVX512: LV: Found an estimated cost of 1 for VF 1 For instruction: store double %v7, ptr %out7, align 8
34 ; AVX512: LV: Found an estimated cost of 23 for VF 2 For instruction: store double %v7, ptr %out7, align 8
35 ; AVX512: LV: Found an estimated cost of 46 for VF 4 For instruction: store double %v7, ptr %out7, align 8
36 ; AVX512: LV: Found an estimated cost of 80 for VF 8 For instruction: store double %v7, ptr %out7, align 8
42 %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
44 %iv.0 = add nuw nsw i64 %iv, 0
45 %iv.1 = add nuw nsw i64 %iv, 1
46 %iv.2 = add nuw nsw i64 %iv, 2
47 %iv.3 = add nuw nsw i64 %iv, 3
48 %iv.4 = add nuw nsw i64 %iv, 4
49 %iv.5 = add nuw nsw i64 %iv, 5
50 %iv.6 = add nuw nsw i64 %iv, 6
51 %iv.7 = add nuw nsw i64 %iv, 7
53 %in = getelementptr inbounds [1024 x i8], ptr @A, i64 0, i64 %iv.0
54 %v.narrow = load i8, ptr %in
56 %v = uitofp i8 %v.narrow to double
58 %v0 = fadd double %v, 0.0
59 %v1 = fadd double %v, 1.0
60 %v2 = fadd double %v, 2.0
61 %v3 = fadd double %v, 3.0
62 %v4 = fadd double %v, 4.0
63 %v5 = fadd double %v, 5.0
64 %v6 = fadd double %v, 6.0
65 %v7 = fadd double %v, 7.0
67 %out0 = getelementptr inbounds [1024 x double], ptr @B, i64 0, i64 %iv.0
68 %out1 = getelementptr inbounds [1024 x double], ptr @B, i64 0, i64 %iv.1
69 %out2 = getelementptr inbounds [1024 x double], ptr @B, i64 0, i64 %iv.2
70 %out3 = getelementptr inbounds [1024 x double], ptr @B, i64 0, i64 %iv.3
71 %out4 = getelementptr inbounds [1024 x double], ptr @B, i64 0, i64 %iv.4
72 %out5 = getelementptr inbounds [1024 x double], ptr @B, i64 0, i64 %iv.5
73 %out6 = getelementptr inbounds [1024 x double], ptr @B, i64 0, i64 %iv.6
74 %out7 = getelementptr inbounds [1024 x double], ptr @B, i64 0, i64 %iv.7
76 store double %v0, ptr %out0
77 store double %v1, ptr %out1
78 store double %v2, ptr %out2
79 store double %v3, ptr %out3
80 store double %v4, ptr %out4
81 store double %v5, ptr %out5
82 store double %v6, ptr %out6
83 store double %v7, ptr %out7
85 %iv.next = add nuw nsw i64 %iv.0, 8
86 %cmp = icmp ult i64 %iv.next, 1024
87 br i1 %cmp, label %for.body, label %for.cond.cleanup