xref: /llvm-project/llvm/test/Transforms/LoopVectorize/X86/vectorization-remarks-profitable.ll (revision 7d7577256b76e4293f455b8093504d5f7044ab4b)
1; RUN: opt < %s -passes=loop-vectorize -pass-remarks-missed='loop-vectorize' -mtriple=x86_64-unknown-linux -S 2>&1 | FileCheck %s
2
3; Verify analysis remarks are generated when interleaving is not beneficial.
4; CHECK: remark: vectorization-remarks-profitable.c:5:17: the cost-model indicates that vectorization is not beneficial
5; CHECK: remark: vectorization-remarks-profitable.c:5:17: the cost-model indicates that interleaving is not beneficial and is explicitly disabled or interleave count is set to 1
6; CHECK: remark: vectorization-remarks-profitable.c:11:3: the cost-model indicates that vectorization is not beneficial
7; CHECK: remark: vectorization-remarks-profitable.c:11:3: the cost-model indicates that interleaving is not beneficial
8
9; First loop.
10;  #pragma clang loop interleave(disable) unroll(disable)
11;  for(int i = 0; i < n; i++) {
12;    out[i] = *in[i];
13;  }
14
15; Second loop.
16;  #pragma clang loop unroll(disable)
17;  for(int i = 0; i < n; i++) {
18;    out[i] = *in[i];
19;  }
20
21; ModuleID = 'vectorization-remarks-profitable.ll'
22target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
23target triple = "x86_64-apple-macosx10.10.0"
24
25; Function Attrs: nounwind uwtable
26define void @do_not_interleave(ptr noalias nocapture readonly %in, ptr noalias nocapture %out, i32 %size) #0 !dbg !4 {
27entry:
28  %cmp.4 = icmp eq i32 %size, 0, !dbg !10
29  br i1 %cmp.4, label %for.end, label %for.body.preheader, !dbg !11
30
31for.body.preheader:                               ; preds = %entry
32  br label %for.body, !dbg !12
33
34for.body:                                         ; preds = %for.body.preheader, %for.body
35  %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %for.body.preheader ]
36  %arrayidx = getelementptr inbounds ptr, ptr %in, i64 %indvars.iv, !dbg !12
37  %0 = load ptr, ptr %arrayidx, align 8, !dbg !12
38  %1 = load i32, ptr %0, align 4, !dbg !13
39  %arrayidx2 = getelementptr inbounds float, ptr %out, i64 %indvars.iv, !dbg !14
40  store i32 %1, ptr %arrayidx2, align 4, !dbg !15
41  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1, !dbg !11
42  %lftr.wideiv = trunc i64 %indvars.iv.next to i32, !dbg !11
43  %exitcond = icmp eq i32 %lftr.wideiv, %size, !dbg !11
44  br i1 %exitcond, label %for.end.loopexit, label %for.body, !dbg !11, !llvm.loop !16
45
46for.end.loopexit:                                 ; preds = %for.body
47  br label %for.end, !dbg !19
48
49for.end:                                          ; preds = %for.end.loopexit, %entry
50  ret void, !dbg !19
51}
52
53; Function Attrs: nounwind uwtable
54define void @interleave_not_profitable(ptr noalias nocapture readonly %in, ptr noalias nocapture %out, i32 %size) #0 !dbg !6 {
55entry:
56  %cmp.4 = icmp eq i32 %size, 0, !dbg !20
57  br i1 %cmp.4, label %for.end, label %for.body, !dbg !21
58
59for.body:                                         ; preds = %entry, %for.body
60  %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]
61  %arrayidx = getelementptr inbounds ptr, ptr %in, i64 %indvars.iv, !dbg !22
62  %0 = load ptr, ptr %arrayidx, align 8, !dbg !22
63  %1 = load i32, ptr %0, align 4, !dbg !23
64  %arrayidx2 = getelementptr inbounds float, ptr %out, i64 %indvars.iv, !dbg !24
65  store i32 %1, ptr %arrayidx2, align 4, !dbg !25
66  %indvars.iv.next = add nuw nsw i64 %indvars.iv, 1, !dbg !21
67  %lftr.wideiv = trunc i64 %indvars.iv.next to i32, !dbg !21
68  %exitcond = icmp eq i32 %lftr.wideiv, %size, !dbg !21
69  br i1 %exitcond, label %for.end, label %for.body, !dbg !21, !llvm.loop !26
70
71for.end:                                          ; preds = %for.body, %entry
72  ret void, !dbg !27
73}
74
75attributes #0 = { nounwind uwtable "disable-tail-calls"="false" "less-precise-fpmad"="false" "frame-pointer"="all" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+mmx,+sse,+sse2" "unsafe-fp-math"="false" "use-soft-float"="false" }
76
77!llvm.dbg.cu = !{!0}
78!llvm.module.flags = !{!7, !8}
79!llvm.ident = !{!9}
80
81!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 3.8.0 (trunk 250016)", isOptimized: false, runtimeVersion: 0, emissionKind: LineTablesOnly, enums: !2)
82!1 = !DIFile(filename: "vectorization-remarks-profitable.c", directory: "")
83!2 = !{}
84!4 = distinct !DISubprogram(name: "do_not_interleave", scope: !1, file: !1, line: 1, type: !5, isLocal: false, isDefinition: true, scopeLine: 1, flags: DIFlagPrototyped, isOptimized: false, unit: !0, retainedNodes: !2)
85!5 = !DISubroutineType(types: !2)
86!6 = distinct !DISubprogram(name: "interleave_not_profitable", scope: !1, file: !1, line: 8, type: !5, isLocal: false, isDefinition: true, scopeLine: 8, flags: DIFlagPrototyped, isOptimized: false, unit: !0, retainedNodes: !2)
87!7 = !{i32 2, !"Dwarf Version", i32 4}
88!8 = !{i32 2, !"Debug Info Version", i32 3}
89!9 = !{!"clang version 3.8.0 (trunk 250016)"}
90!10 = !DILocation(line: 4, column: 23, scope: !4)
91!11 = !DILocation(line: 4, column: 3, scope: !4)
92!12 = !DILocation(line: 5, column: 17, scope: !4)
93!13 = !DILocation(line: 5, column: 16, scope: !4)
94!14 = !DILocation(line: 5, column: 7, scope: !4)
95!15 = !DILocation(line: 5, column: 14, scope: !4)
96!16 = distinct !{!16, !17, !18}
97!17 = !{!"llvm.loop.interleave.count", i32 1}
98!18 = !{!"llvm.loop.unroll.disable"}
99!19 = !DILocation(line: 6, column: 1, scope: !4)
100!20 = !DILocation(line: 11, column: 23, scope: !6)
101!21 = !DILocation(line: 11, column: 3, scope: !6)
102!22 = !DILocation(line: 12, column: 17, scope: !6)
103!23 = !DILocation(line: 12, column: 16, scope: !6)
104!24 = !DILocation(line: 12, column: 7, scope: !6)
105!25 = !DILocation(line: 12, column: 14, scope: !6)
106!26 = distinct !{!26, !18}
107!27 = !DILocation(line: 13, column: 1, scope: !6)
108
109