1; Test for CSSPGO's SampleContextTracker to make sure context profile tree is promoted and merged properly
2; based on inline decision, so post inline counts are accurate.
3
4; RUN: llvm-profdata merge --sample --extbinary %S/Inputs/profile-context-tracker.prof -o %t
5
6; Note that we need new pass manager to enable top-down processing for sample profile loader
7; Test we inlined the following in top-down order and entry counts accurate reflects post-inline base profile
8;   main:3 @ _Z5funcAi
9;   main:3 @ _Z5funcAi:1 @ _Z8funcLeafi
10;   _Z5funcBi:1 @ _Z8funcLeafi
11; RUN: opt < %s -passes=sample-profile -sample-profile-file=%S/Inputs/profile-context-tracker.prof -sample-profile-inline-size -sample-profile-prioritized-inline=0 -profile-sample-accurate -S | FileCheck %s --check-prefix=INLINE-ALL
12; RUN: opt < %s -passes=sample-profile -sample-profile-file=%t -sample-profile-inline-size -sample-profile-prioritized-inline=0 -profile-sample-accurate -S | FileCheck %s --check-prefix=INLINE-ALL
13; RUN: opt < %s -passes=sample-profile -sample-profile-file=%S/Inputs/profile-context-tracker.prof -sample-profile-inline-size -sample-profile-cold-inline-threshold=200 -profile-sample-accurate -S | FileCheck %s --check-prefix=INLINE-ALL
14; RUN: opt < %s -passes=sample-profile -sample-profile-file=%t -sample-profile-inline-size -sample-profile-cold-inline-threshold=200 -profile-sample-accurate -S | FileCheck %s --check-prefix=INLINE-ALL
15;
16; Test we inlined the following in top-down order and entry counts accurate reflects post-inline base profile
17;   _Z5funcAi:1 @ _Z8funcLeafi
18;   _Z5funcBi:1 @ _Z8funcLeafi
19
20
21@factor = dso_local global i32 3, align 4, !dbg !0
22
23define dso_local i32 @main() local_unnamed_addr #0 !dbg !18 {
24; INLINE-ALL: @main{{.*}}!prof ![[MAIN_PROF:[0-9]+]]
25; INLINE-HOT: @main{{.*}}!prof ![[MAIN_PROF:[0-9]+]]
26entry:
27  br label %for.body, !dbg !25
28
29for.cond.cleanup:                                 ; preds = %for.body
30  ret i32 %add3, !dbg !27
31
32for.body:                                         ; preds = %for.body, %entry
33  %x.011 = phi i32 [ 300000, %entry ], [ %dec, %for.body ]
34  %r.010 = phi i32 [ 0, %entry ], [ %add3, %for.body ]
35  %call = tail call i32 @_Z5funcBi(i32 %x.011), !dbg !32
36; _Z5funcBi is marked noinline
37; INLINE-ALL: call i32 @_Z5funcBi
38; INLINE-HOT: call i32 @_Z5funcBi
39  %add = add nuw nsw i32 %x.011, 1, !dbg !31
40  %call1 = tail call i32 @_Z5funcAi(i32 %add), !dbg !28
41; INLINE-ALL-NOT: call i32 @_Z5funcAi
42; INLINE-HOT: call i32 @_Z5funcAi
43  %add2 = add i32 %call, %r.010, !dbg !34
44  %add3 = add i32 %add2, %call1, !dbg !35
45  %dec = add nsw i32 %x.011, -1, !dbg !36
46  %cmp = icmp eq i32 %x.011, 0, !dbg !38
47  br i1 %cmp, label %for.cond.cleanup, label %for.body, !dbg !25
48}
49
50define dso_local i32 @_Z5funcAi(i32 %x) local_unnamed_addr #1 !dbg !40 {
51; _Z5funcAi is inlined, so outline remainder should have zero counts
52; INLINE-ALL: @_Z5funcAi{{.*}}!prof ![[FUNCA_PROF:[0-9]+]]
53; INLINE-HOT: @_Z5funcAi{{.*}}!prof ![[FUNCA_PROF:[0-9]+]]
54entry:
55  %add = add nsw i32 %x, 100000, !dbg !44
56; _Z8funcLeafi is already inlined on main->_Z5funcAi->_Z8funcLeafi,
57; so it should not be inlined on _Z5funcAi->_Z8funcLeafi based on updated
58; (merged and promoted) context profile
59; INLINE-ALL: call i32 @_Z8funcLeafi
60; INLINE-HOT-NOT: call i32 @_Z8funcLeafi
61  %call = tail call i32 @_Z8funcLeafi(i32 %add), !dbg !45
62  ret i32 %call, !dbg !46
63}
64
65define dso_local i32 @_Z8funcLeafi(i32 %x) local_unnamed_addr #1 !dbg !54 {
66; main->_Z5funcAi->_Z8funcLeafi is inlined, and _Z5funcBi->_Z8funcLeafi is also
67; inlined, so outline remainder should have empty profile
68; INLINE-ALL: @_Z8funcLeafi{{.*}}!prof ![[LEAF_PROF:[0-9]+]]
69; INLINE-HOT: @_Z8funcLeafi{{.*}}!prof ![[LEAF_PROF:[0-9]+]]
70entry:
71  %cmp = icmp sgt i32 %x, 0, !dbg !57
72  br i1 %cmp, label %while.body, label %while.cond2.preheader, !dbg !59
73
74while.cond2.preheader:                            ; preds = %entry
75  %cmp313 = icmp slt i32 %x, 0, !dbg !60
76  br i1 %cmp313, label %while.body4, label %if.end, !dbg !63
77
78while.body:                                       ; preds = %while.body, %entry
79  %x.addr.016 = phi i32 [ %sub, %while.body ], [ %x, %entry ]
80  %tmp = load volatile i32, i32* @factor, align 4, !dbg !64
81  %call = tail call i32 @_Z3fibi(i32 %tmp), !dbg !67
82  %sub = sub nsw i32 %x.addr.016, %call, !dbg !68
83  %cmp1 = icmp sgt i32 %sub, 0, !dbg !69
84  br i1 %cmp1, label %while.body, label %if.end, !dbg !71
85
86while.body4:                                      ; preds = %while.body4, %while.cond2.preheader
87  %x.addr.114 = phi i32 [ %add, %while.body4 ], [ %x, %while.cond2.preheader ]
88  %tmp1 = load volatile i32, i32* @factor, align 4, !dbg !72
89  %call5 = tail call i32 @_Z3fibi(i32 %tmp1), !dbg !74
90  %add = add nsw i32 %call5, %x.addr.114, !dbg !75
91  %cmp3 = icmp slt i32 %add, 0, !dbg !60
92  br i1 %cmp3, label %while.body4, label %if.end, !dbg !63
93
94if.end:                                           ; preds = %while.body4, %while.body, %while.cond2.preheader
95  %x.addr.2 = phi i32 [ 0, %while.cond2.preheader ], [ %sub, %while.body ], [ %add, %while.body4 ]
96  ret i32 %x.addr.2, !dbg !76
97}
98
99define dso_local i32 @_Z5funcBi(i32 %x) local_unnamed_addr #0 !dbg !47 {
100; _Z5funcBi is marked noinline, so outline remainder has promoted context profile
101; INLINE-ALL: @_Z5funcBi{{.*}}!prof ![[FUNCB_PROF:[0-9]+]]
102; INLINE-HOT: @_Z5funcBi{{.*}}!prof ![[FUNCB_PROF:[0-9]+]]
103entry:
104  %sub = add nsw i32 %x, -100000, !dbg !51
105  %call = tail call i32 @_Z8funcLeafi(i32 %sub), !dbg !52
106; _Z5funcBi is not inlined into main, so we main->_Z5funcBi->_Z8funcLeafi
107; should be inlined based on promoted context profile
108; INLINE-ALL-NOT: call i32 @_Z8funcLeafi
109; INLINE-HOT-NOT: call i32 @_Z8funcLeafi
110  ret i32 %call, !dbg !53
111}
112
113; INLINE-ALL-DAG: [[MAIN_PROF]] = !{!"function_entry_count", i64 1}
114; INLINE-ALL-DAG: [[FUNCA_PROF]] = !{!"function_entry_count", i64 0}
115; INLINE-ALL-DAG-SAME: [[LEAF_PROF]] = !{!"function_entry_count", i64 0}
116; INLINE-ALL-DAG: [[FUNCB_PROF]] = !{!"function_entry_count", i64 13}
117
118; INLINE-HOT-DAG: [[MAIN_PROF]] = !{!"function_entry_count", i64 1}
119; INLINE-HOT-DAG: [[FUNCA_PROF]] = !{!"function_entry_count", i64 12}
120; INLINE-HOT-DAG-SAME: [[LEAF_PROF]] = !{!"function_entry_count", i64 0}
121; INLINE-HOT-DAG: [[FUNCB_PROF]] = !{!"function_entry_count", i64 13}
122
123declare i32 @_Z3fibi(i32)
124
125attributes #0 = { nofree noinline norecurse nounwind uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "frame-pointer"="none" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "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"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" "use-sample-profile" }
126attributes #1 = { nofree norecurse nounwind uwtable "correctly-rounded-divide-sqrt-fp-math"="false" "disable-tail-calls"="false" "frame-pointer"="none" "less-precise-fpmad"="false" "min-legal-vector-width"="0" "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"="x86-64" "target-features"="+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "unsafe-fp-math"="false" "use-soft-float"="false" "use-sample-profile" }
127
128!llvm.dbg.cu = !{!2}
129!llvm.module.flags = !{!14, !15, !16}
130!llvm.ident = !{!17}
131
132!0 = !DIGlobalVariableExpression(var: !1, expr: !DIExpression())
133!1 = distinct !DIGlobalVariable(name: "factor", scope: !2, file: !3, line: 21, type: !13, isLocal: false, isDefinition: true)
134!2 = distinct !DICompileUnit(language: DW_LANG_C_plus_plus_14, file: !3, producer: "clang version 11.0.0", isOptimized: true, runtimeVersion: 0, emissionKind: FullDebug, enums: !4, retainedTypes: !5, globals: !12, splitDebugInlining: false, debugInfoForProfiling: true, nameTableKind: None)
135!3 = !DIFile(filename: "merged.cpp", directory: "/local/autofdo")
136!4 = !{}
137!5 = !{!6, !10, !11}
138!6 = !DISubprogram(name: "funcA", linkageName: "_Z5funcAi", scope: !3, file: !3, line: 6, type: !7, flags: DIFlagPrototyped, spFlags: DISPFlagOptimized, retainedNodes: !4)
139!7 = !DISubroutineType(types: !8)
140!8 = !{!9, !9}
141!9 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)
142!10 = !DISubprogram(name: "funcB", linkageName: "_Z5funcBi", scope: !3, file: !3, line: 7, type: !7, flags: DIFlagPrototyped, spFlags: DISPFlagOptimized, retainedNodes: !4)
143!11 = !DISubprogram(name: "funcLeaf", linkageName: "_Z8funcLeafi", scope: !3, file: !3, line: 22, type: !7, flags: DIFlagPrototyped, spFlags: DISPFlagOptimized, retainedNodes: !4)
144!12 = !{!0}
145!13 = !DIDerivedType(tag: DW_TAG_volatile_type, baseType: !9)
146!14 = !{i32 7, !"Dwarf Version", i32 4}
147!15 = !{i32 2, !"Debug Info Version", i32 3}
148!16 = !{i32 1, !"wchar_size", i32 4}
149!17 = !{!"clang version 11.0.0"}
150!18 = distinct !DISubprogram(name: "main", scope: !3, file: !3, line: 11, type: !19, scopeLine: 11, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !2, retainedNodes: !21)
151!19 = !DISubroutineType(types: !20)
152!20 = !{!9}
153!21 = !{!22, !23}
154!22 = !DILocalVariable(name: "r", scope: !18, file: !3, line: 12, type: !9)
155!23 = !DILocalVariable(name: "x", scope: !24, file: !3, line: 13, type: !9)
156!24 = distinct !DILexicalBlock(scope: !18, file: !3, line: 13, column: 3)
157!25 = !DILocation(line: 13, column: 3, scope: !26)
158!26 = !DILexicalBlockFile(scope: !24, file: !3, discriminator: 2)
159!27 = !DILocation(line: 17, column: 3, scope: !18)
160!28 = !DILocation(line: 14, column: 10, scope: !29)
161!29 = distinct !DILexicalBlock(scope: !30, file: !3, line: 13, column: 37)
162!30 = distinct !DILexicalBlock(scope: !24, file: !3, line: 13, column: 3)
163!31 = !DILocation(line: 14, column: 29, scope: !29)
164!32 = !DILocation(line: 14, column: 21, scope: !33)
165!33 = !DILexicalBlockFile(scope: !29, file: !3, discriminator: 2)
166!34 = !DILocation(line: 14, column: 19, scope: !29)
167!35 = !DILocation(line: 14, column: 7, scope: !29)
168!36 = !DILocation(line: 13, column: 33, scope: !37)
169!37 = !DILexicalBlockFile(scope: !30, file: !3, discriminator: 6)
170!38 = !DILocation(line: 13, column: 26, scope: !39)
171!39 = !DILexicalBlockFile(scope: !30, file: !3, discriminator: 2)
172!40 = distinct !DISubprogram(name: "funcA", linkageName: "_Z5funcAi", scope: !3, file: !3, line: 26, type: !7, scopeLine: 26, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !2)
173!44 = !DILocation(line: 27, column: 22, scope: !40)
174!45 = !DILocation(line: 27, column: 11, scope: !40)
175!46 = !DILocation(line: 29, column: 3, scope: !40)
176!47 = distinct !DISubprogram(name: "funcB", linkageName: "_Z5funcBi", scope: !3, file: !3, line: 32, type: !7, scopeLine: 32, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !2)
177!51 = !DILocation(line: 33, column: 22, scope: !47)
178!52 = !DILocation(line: 33, column: 11, scope: !47)
179!53 = !DILocation(line: 35, column: 3, scope: !47)
180!54 = distinct !DISubprogram(name: "funcLeaf", linkageName: "_Z8funcLeafi", scope: !3, file: !3, line: 48, type: !7, scopeLine: 48, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !2)
181!57 = !DILocation(line: 49, column: 9, scope: !58)
182!58 = distinct !DILexicalBlock(scope: !54, file: !3, line: 49, column: 7)
183!59 = !DILocation(line: 49, column: 7, scope: !54)
184!60 = !DILocation(line: 58, column: 14, scope: !61)
185!61 = !DILexicalBlockFile(scope: !62, file: !3, discriminator: 2)
186!62 = distinct !DILexicalBlock(scope: !58, file: !3, line: 56, column: 8)
187!63 = !DILocation(line: 58, column: 5, scope: !61)
188!64 = !DILocation(line: 52, column: 16, scope: !65)
189!65 = distinct !DILexicalBlock(scope: !66, file: !3, line: 51, column: 19)
190!66 = distinct !DILexicalBlock(scope: !58, file: !3, line: 49, column: 14)
191!67 = !DILocation(line: 52, column: 12, scope: !65)
192!68 = !DILocation(line: 52, column: 9, scope: !65)
193!69 = !DILocation(line: 51, column: 14, scope: !70)
194!70 = !DILexicalBlockFile(scope: !66, file: !3, discriminator: 2)
195!71 = !DILocation(line: 51, column: 5, scope: !70)
196!72 = !DILocation(line: 59, column: 16, scope: !73)
197!73 = distinct !DILexicalBlock(scope: !62, file: !3, line: 58, column: 19)
198!74 = !DILocation(line: 59, column: 12, scope: !73)
199!75 = !DILocation(line: 59, column: 9, scope: !73)
200!76 = !DILocation(line: 63, column: 3, scope: !54)
201