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*%v0 = load float, ptr %in0"
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 float] 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 float, ptr %in0, align 4
17 ; SSE2: LV: Found an estimated cost of 21 for VF 2 For instruction: %v0 = load float, ptr %in0, align 4
18 ; SSE2: LV: Found an estimated cost of 49 for VF 4 For instruction: %v0 = load float, ptr %in0, align 4
19 ; SSE2: LV: Found an estimated cost of 98 for VF 8 For instruction: %v0 = load float, ptr %in0, align 4
22 ; AVX1: LV: Found an estimated cost of 1 for VF 1 For instruction: %v0 = load float, ptr %in0, align 4
23 ; AVX1: LV: Found an estimated cost of 23 for VF 2 For instruction: %v0 = load float, ptr %in0, align 4
24 ; AVX1: LV: Found an estimated cost of 49 for VF 4 For instruction: %v0 = load float, ptr %in0, align 4
25 ; AVX1: LV: Found an estimated cost of 105 for VF 8 For instruction: %v0 = load float, ptr %in0, align 4
26 ; AVX1: LV: Found an estimated cost of 210 for VF 16 For instruction: %v0 = load float, ptr %in0, align 4
29 ; AVX2: LV: Found an estimated cost of 1 for VF 1 For instruction: %v0 = load float, ptr %in0, align 4
30 ; AVX2: LV: Found an estimated cost of 21 for VF 2 For instruction: %v0 = load float, ptr %in0, align 4
31 ; AVX2: LV: Found an estimated cost of 49 for VF 4 For instruction: %v0 = load float, ptr %in0, align 4
32 ; AVX2: LV: Found an estimated cost of 105 for VF 8 For instruction: %v0 = load float, ptr %in0, align 4
33 ; AVX2: LV: Found an estimated cost of 210 for VF 16 For instruction: %v0 = load float, ptr %in0, align 4
35 ; AVX512-LABEL: 'test'
36 ; AVX512: LV: Found an estimated cost of 1 for VF 1 For instruction: %v0 = load float, ptr %in0, align 4
37 ; AVX512: LV: Found an estimated cost of 8 for VF 2 For instruction: %v0 = load float, ptr %in0, align 4
38 ; AVX512: LV: Found an estimated cost of 12 for VF 4 For instruction: %v0 = load float, ptr %in0, align 4
39 ; AVX512: LV: Found an estimated cost of 35 for VF 8 For instruction: %v0 = load float, ptr %in0, align 4
40 ; AVX512: LV: Found an estimated cost of 70 for VF 16 For instruction: %v0 = load float, ptr %in0, align 4
41 ; AVX512: LV: Found an estimated cost of 280 for VF 32 For instruction: %v0 = load float, ptr %in0, align 4
47 %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
49 %iv.0 = add nuw nsw i64 %iv, 0
50 %iv.1 = add nuw nsw i64 %iv, 1
51 %iv.2 = add nuw nsw i64 %iv, 2
52 %iv.3 = add nuw nsw i64 %iv, 3
53 %iv.4 = add nuw nsw i64 %iv, 4
54 %iv.5 = add nuw nsw i64 %iv, 5
55 %iv.6 = add nuw nsw i64 %iv, 6
57 %in0 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.0
58 %in1 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.1
59 %in2 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.2
60 %in3 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.3
61 %in4 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.4
62 %in5 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.5
63 %in6 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.6
65 %v0 = load float, ptr %in0
66 %v1 = load float, ptr %in1
67 %v2 = load float, ptr %in2
68 %v3 = load float, ptr %in3
69 %v4 = load float, ptr %in4
70 %v5 = load float, ptr %in5
71 %v6 = load float, ptr %in6
73 %reduce.add.0 = fadd float %v0, %v1
74 %reduce.add.1 = fadd float %reduce.add.0, %v2
75 %reduce.add.2 = fadd float %reduce.add.1, %v3
76 %reduce.add.3 = fadd float %reduce.add.2, %v4
77 %reduce.add.4 = fadd float %reduce.add.3, %v5
78 %reduce.add.5 = fadd float %reduce.add.4, %v6
80 %reduce.add.5.narrow = fptoui float %reduce.add.5 to i8
82 %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0
83 store i8 %reduce.add.5.narrow, ptr %out
85 %iv.next = add nuw nsw i64 %iv.0, 7
86 %cmp = icmp ult i64 %iv.next, 1024
87 br i1 %cmp, label %for.body, label %for.cond.cleanup