1; RUN: opt < %s -sample-profile -sample-profile-file=%S/Inputs/entry_counts_cold.prof -S -profile-sample-accurate | FileCheck %s
2; ModuleID = 'temp.bc'
3; This is a test similar to entry_counts_cold.ll. The key differences are:
4; - we want profile-sample-accurate. Normally, that would trigger resetting function entry counts to 0, and then update
5; them based on the sample profile (SampleProfile::runOnFunction).
6; - we don't have debug info for the function. This means we can't match sample counts in the profile to locations in this function.
7; When we attempt to update its profile info (at the end of SampleProfile::runOnModule - llvm::updateProfileCallee), we'll call
8; updateProfWeight with a 0 priorEntryCount, which will result in a division by 0.
9;
10; We're using @bar and the call to @baz.
11;
12; Currently, the test just ensures opt doesn't ICE
13source_filename = "temp.c"
14target datalayout = "e-m:o-i64:64-f80:128-n8:16:32:64-S128"
15target triple = "x86_64-apple-macosx10.14.0"
16
17; Function Attrs: nounwind ssp uwtable
18; CHECK: define i32 @top({{.*}} !prof [[TOP:![0-9]+]]
19define i32 @top(i32* %p) #0 !dbg !8 {
20entry:
21  %p.addr = alloca i32*, align 8
22  store i32* %p, i32** %p.addr, align 8, !tbaa !15
23  call void @llvm.dbg.declare(metadata i32** %p.addr, metadata !14, metadata !DIExpression()), !dbg !19
24  %0 = load i32*, i32** %p.addr, align 8, !dbg !20, !tbaa !15
25  %call = call i32 @foo(i32* %0), !dbg !21
26; foo is inlined
27; CHECK-NOT: call i32 @foo
28; CHECK: call i32 @bar
29  %1 = load i32*, i32** %p.addr, align 8, !dbg !22, !tbaa !15
30  %2 = load i32, i32* %1, align 4, !dbg !24, !tbaa !25
31  %tobool = icmp ne i32 %2, 0, !dbg !24
32  br i1 %tobool, label %if.then, label %if.end, !dbg !27
33
34if.then:                                          ; preds = %entry
35  %3 = load i32*, i32** %p.addr, align 8, !dbg !28, !tbaa !15
36; bar is not inlined
37; CHECK: call i32 @bar
38  %call1 = call i32 @bar(i32* %3), !dbg !29
39  br label %if.end, !dbg !29
40
41if.end:                                           ; preds = %if.then, %entry
42  ret i32 0, !dbg !30
43}
44
45; Function Attrs: nounwind readnone speculatable
46declare void @llvm.dbg.declare(metadata, metadata, metadata) #1
47
48; Function Attrs: nounwind ssp uwtable
49; CHECK: define i32 @foo({{.*}} !prof [[FOO:![0-9]+]]
50define i32 @foo(i32* %p) #0 !dbg !31 {
51entry:
52  %p.addr = alloca i32*, align 8
53  %a = alloca i32, align 4
54  store i32* %p, i32** %p.addr, align 8, !tbaa !15
55  call void @llvm.dbg.declare(metadata i32** %p.addr, metadata !33, metadata !DIExpression()), !dbg !35
56  %0 = bitcast i32* %a to i8*, !dbg !36
57  call void @llvm.lifetime.start.p0i8(i64 4, i8* %0) #4, !dbg !36
58  call void @llvm.dbg.declare(metadata i32* %a, metadata !34, metadata !DIExpression()), !dbg !37
59  %1 = load i32*, i32** %p.addr, align 8, !dbg !38, !tbaa !15
60  %arrayidx = getelementptr inbounds i32, i32* %1, i64 3, !dbg !38
61  %2 = load i32, i32* %arrayidx, align 4, !dbg !38, !tbaa !25
62  %3 = load i32*, i32** %p.addr, align 8, !dbg !39, !tbaa !15
63  %arrayidx1 = getelementptr inbounds i32, i32* %3, i64 2, !dbg !39
64  %4 = load i32, i32* %arrayidx1, align 4, !dbg !40, !tbaa !25
65  %add = add nsw i32 %4, %2, !dbg !40
66  store i32 %add, i32* %arrayidx1, align 4, !dbg !40, !tbaa !25
67  %5 = load i32*, i32** %p.addr, align 8, !dbg !41, !tbaa !15
68  %call = call i32 @bar(i32* %5), !dbg !42
69  store i32 %call, i32* %a, align 4, !dbg !43, !tbaa !25
70  %6 = load i32, i32* %a, align 4, !dbg !44, !tbaa !25
71  %add2 = add nsw i32 %6, 1, !dbg !45
72  %7 = bitcast i32* %a to i8*, !dbg !46
73  call void @llvm.lifetime.end.p0i8(i64 4, i8* %7) #4, !dbg !46
74  ret i32 %add2, !dbg !47
75}
76
77; Function Attrs: nounwind ssp uwtable
78; CHECK: define i32 @bar(i32* %p) #0 !prof [[BAR:![0-9]+]] {
79define i32 @bar(i32* %p) #0 {
80entry:
81  %p.addr = alloca i32*, align 8
82  store i32* %p, i32** %p.addr, align 8, !tbaa !15
83  call void @llvm.dbg.declare(metadata i32** %p.addr, metadata !50, metadata !DIExpression()), !dbg !51
84  ; CHECK: call void (...) @baz(), !dbg !{{[0-9]+}}
85  call void (...) @baz(), !dbg !52, !prof !100
86  %0 = load i32*, i32** %p.addr, align 8, !dbg !53, !tbaa !15
87  %arrayidx = getelementptr inbounds i32, i32* %0, i64 2, !dbg !53
88  %1 = load i32, i32* %arrayidx, align 4, !dbg !53, !tbaa !25
89  %2 = load i32*, i32** %p.addr, align 8, !dbg !54, !tbaa !15
90  %arrayidx1 = getelementptr inbounds i32, i32* %2, i64 1, !dbg !54
91  %3 = load i32, i32* %arrayidx1, align 4, !dbg !55, !tbaa !25
92  %add = add nsw i32 %3, %1, !dbg !55
93  store i32 %add, i32* %arrayidx1, align 4, !dbg !55, !tbaa !25
94  %4 = load i32*, i32** %p.addr, align 8, !dbg !56, !tbaa !15
95  %arrayidx2 = getelementptr inbounds i32, i32* %4, i64 3, !dbg !56
96  %5 = load i32, i32* %arrayidx2, align 4, !dbg !56, !tbaa !25
97  ret i32 %5, !dbg !57
98}
99
100; Function Attrs: argmemonly nounwind
101declare void @llvm.lifetime.start.p0i8(i64, i8* nocapture) #2
102
103; Function Attrs: argmemonly nounwind
104declare void @llvm.lifetime.end.p0i8(i64, i8* nocapture) #2
105
106declare void @baz(...) #3
107
108attributes #0 = { nounwind ssp uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "frame-pointer"="all" "no-infs-fp-math"="false" "no-jump-tables"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="penryn" "target-features"="+cx16,+fxsr,+mmx,+sahf,+sse,+sse2,+sse3,+sse4.1,+ssse3,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" "use-sample-profile" }
109attributes #1 = { nounwind readnone speculatable }
110attributes #2 = { argmemonly nounwind }
111attributes #3 = { "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "less-precise-fpmad"="false" "frame-pointer"="all" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "no-signed-zeros-fp-math"="false" "no-trapping-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="penryn" "target-features"="+cx16,+fxsr,+mmx,+sahf,+sse,+sse2,+sse3,+sse4.1,+ssse3,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" }
112attributes #4 = { nounwind }
113
114!llvm.dbg.cu = !{!0}
115!llvm.module.flags = !{!3, !4, !5, !6}
116!llvm.ident = !{!7}
117
118; CHECK: [[TOP]] = !{!"function_entry_count", i64 101}
119; CHECK: [[FOO]] = !{!"function_entry_count", i64 151}
120
121!0 = distinct !DICompileUnit(language: DW_LANG_C99, file: !1, producer: "clang version 8.0.0", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !2, nameTableKind: GNU)
122!1 = !DIFile(filename: "temp.c", directory: "llvm/test/Transforms/SampleProfile")
123!2 = !{}
124!3 = !{i32 2, !"Dwarf Version", i32 4}
125!4 = !{i32 2, !"Debug Info Version", i32 3}
126!5 = !{i32 1, !"wchar_size", i32 4}
127!6 = !{i32 7, !"PIC Level", i32 2}
128!7 = !{!"clang version 8.0.0"}
129!8 = distinct !DISubprogram(name: "top", scope: !1, file: !1, line: 5, type: !9, scopeLine: 5, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !13)
130!9 = !DISubroutineType(types: !10)
131!10 = !{!11, !12}
132!11 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
133!12 = !DIDerivedType(tag: DW_TAG_pointer_type, baseType: !11, size: 64)
134!13 = !{!14}
135!14 = !DILocalVariable(name: "p", arg: 1, scope: !8, file: !1, line: 5, type: !12)
136!15 = !{!16, !16, i64 0}
137!16 = !{!"any pointer", !17, i64 0}
138!17 = !{!"omnipotent char", !18, i64 0}
139!18 = !{!"Simple C/C++ TBAA"}
140!19 = !DILocation(line: 5, column: 14, scope: !8)
141!20 = !DILocation(line: 6, column: 7, scope: !8)
142!21 = !DILocation(line: 6, column: 3, scope: !8)
143!22 = !DILocation(line: 7, column: 8, scope: !23)
144!23 = distinct !DILexicalBlock(scope: !8, file: !1, line: 7, column: 7)
145!24 = !DILocation(line: 7, column: 7, scope: !23)
146!25 = !{!26, !26, i64 0}
147!26 = !{!"int", !17, i64 0}
148!27 = !DILocation(line: 7, column: 7, scope: !8)
149!28 = !DILocation(line: 8, column: 9, scope: !23)
150!29 = !DILocation(line: 8, column: 5, scope: !23)
151!30 = !DILocation(line: 9, column: 3, scope: !8)
152!31 = distinct !DISubprogram(name: "foo", scope: !1, file: !1, line: 12, type: !9, scopeLine: 12, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !32)
153!32 = !{!33, !34}
154!33 = !DILocalVariable(name: "p", arg: 1, scope: !31, file: !1, line: 12, type: !12)
155!34 = !DILocalVariable(name: "a", scope: !31, file: !1, line: 13, type: !11)
156!35 = !DILocation(line: 12, column: 14, scope: !31)
157!36 = !DILocation(line: 13, column: 3, scope: !31)
158!37 = !DILocation(line: 13, column: 7, scope: !31)
159!38 = !DILocation(line: 14, column: 11, scope: !31)
160!39 = !DILocation(line: 14, column: 3, scope: !31)
161!40 = !DILocation(line: 14, column: 8, scope: !31)
162!41 = !DILocation(line: 15, column: 11, scope: !31)
163!42 = !DILocation(line: 15, column: 7, scope: !31)
164!43 = !DILocation(line: 15, column: 5, scope: !31)
165!44 = !DILocation(line: 16, column: 10, scope: !31)
166!45 = !DILocation(line: 16, column: 11, scope: !31)
167!46 = !DILocation(line: 17, column: 1, scope: !31)
168!47 = !DILocation(line: 16, column: 3, scope: !31)
169!48 = distinct !DISubprogram(name: "bar", scope: !1, file: !1, line: 19, type: !9, scopeLine: 19, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0, retainedNodes: !49)
170!49 = !{!50}
171!50 = !DILocalVariable(name: "p", arg: 1, scope: !48, file: !1, line: 19, type: !12)
172!51 = !DILocation(line: 19, column: 15, scope: !48)
173!52 = !DILocation(line: 20, column: 3, scope: !48)
174!53 = !DILocation(line: 21, column: 11, scope: !48)
175!54 = !DILocation(line: 21, column: 3, scope: !48)
176!55 = !DILocation(line: 21, column: 8, scope: !48)
177!56 = !DILocation(line: 22, column: 10, scope: !48)
178!57 = !DILocation(line: 22, column: 3, scope: !48)
179!100 = !{!"branch_weights", i32 5}
180