1; RUN: opt < %s -loop-vectorize -mtriple=x86_64-unknown-linux -S -pass-remarks='loop-vectorize' 2>&1 | FileCheck -check-prefix=VECTORIZED %s
2; RUN: opt < %s -loop-vectorize -force-vector-width=1 -force-vector-interleave=4 -mtriple=x86_64-unknown-linux -S -pass-remarks='loop-vectorize' 2>&1 | FileCheck -check-prefix=UNROLLED %s
3; RUN: opt < %s -loop-vectorize -force-vector-width=1 -force-vector-interleave=1 -mtriple=x86_64-unknown-linux -S -pass-remarks-analysis='loop-vectorize' 2>&1 | FileCheck -check-prefix=NONE %s
4
5; This code has all the !dbg annotations needed to track source line information,
6; but is missing the llvm.dbg.cu annotation. This prevents code generation from
7; emitting debug info in the final output.
8; RUN: llc -mtriple x86_64-pc-linux-gnu %s -o - | FileCheck -check-prefix=DEBUG-OUTPUT %s
9; DEBUG-OUTPUT-NOT: .loc
10; DEBUG-OUTPUT-NOT: {{.*}}.debug_info
11
12; VECTORIZED: remark: vectorization-remarks.c:17:8: vectorized loop (vectorization factor: 4, unrolling interleave factor: 1)
13; UNROLLED: remark: vectorization-remarks.c:17:8: unrolled with interleaving factor 4 (vectorization not beneficial)
14; NONE: remark: vectorization-remarks.c:17:8: loop not vectorized: vector width and interleave count are explicitly set to 1
15
16target datalayout = "e-m:e-i64:64-f80:128-n8:16:32:64-S128"
17
18define i32 @foo(i32 %n) #0 {
19entry:
20  %diff = alloca i32, align 4
21  %cb = alloca [16 x i8], align 16
22  %cc = alloca [16 x i8], align 16
23  store i32 0, i32* %diff, align 4, !dbg !10, !tbaa !11
24  br label %for.body, !dbg !15
25
26for.body:                                         ; preds = %for.body, %entry
27  %indvars.iv = phi i64 [ 0, %entry ], [ %indvars.iv.next, %for.body ]
28  %add8 = phi i32 [ 0, %entry ], [ %add, %for.body ], !dbg !19
29  %arrayidx = getelementptr inbounds [16 x i8], [16 x i8]* %cb, i64 0, i64 %indvars.iv, !dbg !19
30  %0 = load i8, i8* %arrayidx, align 1, !dbg !19, !tbaa !21
31  %conv = sext i8 %0 to i32, !dbg !19
32  %arrayidx2 = getelementptr inbounds [16 x i8], [16 x i8]* %cc, i64 0, i64 %indvars.iv, !dbg !19
33  %1 = load i8, i8* %arrayidx2, align 1, !dbg !19, !tbaa !21
34  %conv3 = sext i8 %1 to i32, !dbg !19
35  %sub = sub i32 %conv, %conv3, !dbg !19
36  %add = add nsw i32 %sub, %add8, !dbg !19
37  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1, !dbg !15
38  %exitcond = icmp eq i64 %indvars.iv.next, 16, !dbg !15
39  br i1 %exitcond, label %for.end, label %for.body, !dbg !15
40
41for.end:                                          ; preds = %for.body
42  store i32 %add, i32* %diff, align 4, !dbg !19, !tbaa !11
43  call void @ibar(i32* %diff) #2, !dbg !22
44  ret i32 0, !dbg !23
45}
46
47declare void @ibar(i32*) #1
48
49!llvm.module.flags = !{!7, !8}
50!llvm.ident = !{!9}
51
52!1 = !MDFile(filename: "vectorization-remarks.c", directory: ".")
53!2 = !{}
54!3 = !{!4}
55!4 = !MDSubprogram(name: "foo", line: 5, isLocal: false, isDefinition: true, virtualIndex: 6, flags: DIFlagPrototyped, isOptimized: true, scopeLine: 6, file: !1, scope: !5, type: !6, function: i32 (i32)* @foo, variables: !2)
56!5 = !MDFile(filename: "vectorization-remarks.c", directory: ".")
57!6 = !MDSubroutineType(types: !2)
58!7 = !{i32 2, !"Dwarf Version", i32 4}
59!8 = !{i32 1, !"Debug Info Version", i32 3}
60!9 = !{!"clang version 3.5.0 "}
61!10 = !MDLocation(line: 8, column: 3, scope: !4)
62!11 = !{!12, !12, i64 0}
63!12 = !{!"int", !13, i64 0}
64!13 = !{!"omnipotent char", !14, i64 0}
65!14 = !{!"Simple C/C++ TBAA"}
66!15 = !MDLocation(line: 17, column: 8, scope: !16)
67!16 = distinct !MDLexicalBlock(line: 17, column: 8, file: !1, scope: !17)
68!17 = distinct !MDLexicalBlock(line: 17, column: 8, file: !1, scope: !18)
69!18 = distinct !MDLexicalBlock(line: 17, column: 3, file: !1, scope: !4)
70!19 = !MDLocation(line: 18, column: 5, scope: !20)
71!20 = distinct !MDLexicalBlock(line: 17, column: 27, file: !1, scope: !18)
72!21 = !{!13, !13, i64 0}
73!22 = !MDLocation(line: 20, column: 3, scope: !4)
74!23 = !MDLocation(line: 21, column: 3, scope: !4)
75