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, float\* %in0, align 4"
2; RUN: opt  -passes=loop-vectorize  -vectorizer-maximize-bandwidth -S -mattr=+avx512bw --debug-only=loop-vectorize < %s 2>&1 | FileCheck %s --check-prefixes=CHECK,AVX512
3; REQUIRES: asserts
4
5target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
6target triple = "x86_64-unknown-linux-gnu"
7
8@A = global [1024 x float] zeroinitializer, align 128
9@B = global [1024 x i8] zeroinitializer, align 128
10
11define void @test() {
12entry:
13  br label %for.body
14
15for.body:
16  %iv = phi i64 [ 0, %entry ], [ %iv.next, %for.body ]
17
18  %iv.0 = add nuw nsw i64 %iv, 0
19  %iv.1 = add nuw nsw i64 %iv, 1
20
21  %in0 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.0
22  %in1 = getelementptr inbounds [1024 x float], ptr @A, i64 0, i64 %iv.1
23
24  %v0 = load float, ptr %in0
25  %v1 = load float, ptr %in1
26
27  %reduce.add.0 = fadd float %v0, %v1
28
29  %reduce.add.0.narrow = fptoui float %reduce.add.0 to i8
30
31  %out = getelementptr inbounds [1024 x i8], ptr @B, i64 0, i64 %iv.0
32  store i8 %reduce.add.0.narrow, ptr %out
33
34  %iv.next = add nuw nsw i64 %iv.0, 2
35  %cmp = icmp ult i64 %iv.next, 1024
36  br i1 %cmp, label %for.body, label %for.cond.cleanup
37
38for.cond.cleanup:
39  ret void
40}
41