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 double, 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 double] 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 double, ptr %in0, align 8
17 ; SSE2: LV: Found an estimated cost of 18 for VF 2 For instruction: %v0 = load double, ptr %in0, align 8
18 ; SSE2: LV: Found an estimated cost of 36 for VF 4 For instruction: %v0 = load double, ptr %in0, align 8
21 ; AVX1: LV: Found an estimated cost of 1 for VF 1 For instruction: %v0 = load double, ptr %in0, align 8
22 ; AVX1: LV: Found an estimated cost of 18 for VF 2 For instruction: %v0 = load double, ptr %in0, align 8
23 ; AVX1: LV: Found an estimated cost of 42 for VF 4 For instruction: %v0 = load double, ptr %in0, align 8
24 ; AVX1: LV: Found an estimated cost of 84 for VF 8 For instruction: %v0 = load double, ptr %in0, align 8
27 ; AVX2: LV: Found an estimated cost of 1 for VF 1 For instruction: %v0 = load double, ptr %in0, align 8
28 ; AVX2: LV: Found an estimated cost of 9 for VF 2 For instruction: %v0 = load double, ptr %in0, align 8
29 ; AVX2: LV: Found an estimated cost of 24 for VF 4 For instruction: %v0 = load double, ptr %in0, align 8
30 ; AVX2: LV: Found an estimated cost of 48 for VF 8 For instruction: %v0 = load double, ptr %in0, align 8
32 ; AVX512-LABEL: 'test'
33 ; AVX512: LV: Found an estimated cost of 1 for VF 1 For instruction: %v0 = load double, ptr %in0, align 8
34 ; AVX512: LV: Found an estimated cost of 11 for VF 2 For instruction: %v0 = load double, ptr %in0, align 8
35 ; AVX512: LV: Found an estimated cost of 21 for VF 4 For instruction: %v0 = load double, ptr %in0, align 8
36 ; AVX512: LV: Found an estimated cost of 51 for VF 8 For instruction: %v0 = load double, ptr %in0, align 8
37 ; AVX512: LV: Found an estimated cost of 120 for VF 16 For instruction: %v0 = load double, ptr %in0, align 8
38 ; AVX512: LV: Found an estimated cost of 240 for VF 32 For instruction: %v0 = load double, ptr %in0, align 8
44 %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
46 %iv.0 = add nuw nsw i64 %iv, 0
47 %iv.1 = add nuw nsw i64 %iv, 1
48 %iv.2 = add nuw nsw i64 %iv, 2
49 %iv.3 = add nuw nsw i64 %iv, 3
50 %iv.4 = add nuw nsw i64 %iv, 4
51 %iv.5 = add nuw nsw i64 %iv, 5
53 %in0 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.0
54 %in1 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.1
55 %in2 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.2
56 %in3 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.3
57 %in4 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.4
58 %in5 = getelementptr inbounds [1024 x double], ptr @A, i64 0, i64 %iv.5
60 %v0 = load double, ptr %in0
61 %v1 = load double, ptr %in1
62 %v2 = load double, ptr %in2
63 %v3 = load double, ptr %in3
64 %v4 = load double, ptr %in4
65 %v5 = load double, ptr %in5
67 %reduce.add.0 = fadd double %v0, %v1
68 %reduce.add.1 = fadd double %reduce.add.0, %v2
69 %reduce.add.2 = fadd double %reduce.add.1, %v3
70 %reduce.add.3 = fadd double %reduce.add.2, %v4
71 %reduce.add.4 = fadd double %reduce.add.3, %v5
73 %reduce.add.4.narrow = fptoui double %reduce.add.4 to i8
75 %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0
76 store i8 %reduce.add.4.narrow, ptr %out
78 %iv.next = add nuw nsw i64 %iv.0, 6
79 %cmp = icmp ult i64 %iv.next, 1024
80 br i1 %cmp, label %for.body, label %for.cond.cleanup