1; REQUIRES: asserts 2; Test for CSSPGO's SampleContextTracker to make sure context profile tree is promoted and merged properly 3; based on inline decision, so post inline counts are accurate. 4 5; Note that we need new pass manager to enable top-down processing for sample profile loader 6; 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 -debug-only=sample-context-tracker -o /dev/null 2>&1 | FileCheck %s --check-prefix=INLINE-ALL 7; RUN: opt < %s -passes=sample-profile -sample-profile-file=%S/Inputs/profile-context-tracker.prof -sample-profile-prioritized-inline=0 -sample-profile-inline-size=0 -debug-only=sample-context-tracker -o /dev/null 2>&1 | FileCheck %s --check-prefix=INLINE-HOT 8 9 10; Test we inlined the following in top-down order and promot rest not inlined context profile into base profile 11; main:3 @ _Z5funcAi 12; main:3 @ _Z5funcAi:1 @ _Z8funcLeafi 13; _Z5funcBi:1 @ _Z8funcLeafi 14; INLINE-ALL: Getting base profile for function: main 15; INLINE-ALL-NEXT: Merging context profile into base profile: main 16; INLINE-ALL-NEXT: Found context tree root to promote: external:12 @ main 17; INLINE-ALL-NEXT: Context promoted and merged to: main 18; INLINE-ALL-NEXT: Getting callee context for instr: %call = tail call i32 @_Z5funcBi 19; INLINE-ALL-NEXT: Callee context found: main:3.1 @ _Z5funcBi 20; INLINE-ALL-NEXT: Getting callee context for instr: %call1 = tail call i32 @_Z5funcAi 21; INLINE-ALL-NEXT: Callee context found: main:3 @ _Z5funcAi 22; INLINE-ALL-NEXT: Marking context profile as inlined: main:3 @ _Z5funcAi 23; INLINE-ALL-NEXT: Getting callee context for instr: %call.i = tail call i32 @_Z8funcLeafi 24; INLINE-ALL-NEXT: Callee context found: main:3 @ _Z5funcAi:1 @ _Z8funcLeafi 25; INLINE-ALL-NEXT: Marking context profile as inlined: main:3 @ _Z5funcAi:1 @ _Z8funcLeafi 26; INLINE-ALL-NEXT: Getting callee context for instr: %call.i1 = tail call i32 @_Z3fibi 27; INLINE-ALL-NEXT: Getting callee context for instr: %call5.i = tail call i32 @_Z3fibi 28; INLINE-ALL-NEXT: Getting base profile for function: _Z5funcAi 29; INLINE-ALL-NEXT: Merging context profile into base profile: _Z5funcAi 30; INLINE-ALL-NEXT: Getting base profile for function: _Z5funcBi 31; INLINE-ALL-NEXT: Merging context profile into base profile: _Z5funcBi 32; INLINE-ALL-NEXT: Found context tree root to promote: external:10 @ _Z5funcBi 33; INLINE-ALL-NEXT: Context promoted to: _Z5funcBi 34; INLINE-ALL-NEXT: Found context tree root to promote: main:3.1 @ _Z5funcBi 35; INLINE-ALL-NEXT: Context promoted and merged to: _Z5funcBi 36; INLINE-ALL-NEXT: Context promoted to: _Z5funcBi:1 @ _Z8funcLeafi 37; INLINE-ALL-NEXT: Found context tree root to promote: externalA:17 @ _Z5funcBi 38; INLINE-ALL-NEXT: Context promoted and merged to: _Z5funcBi 39; INLINE-ALL-NEXT: Getting callee context for instr: %call = tail call i32 @_Z8funcLeafi 40; INLINE-ALL-NEXT: Callee context found: _Z5funcBi:1 @ _Z8funcLeafi 41; INLINE-ALL-NEXT: Marking context profile as inlined: _Z5funcBi:1 @ _Z8funcLeafi 42; INLINE-ALL-NEXT: Getting callee context for instr: %call.i = tail call i32 @_Z3fibi 43; INLINE-ALL-NEXT: Getting callee context for instr: %call5.i = tail call i32 @_Z3fibi 44; INLINE-ALL-NEXT: Getting base profile for function: _Z8funcLeafi 45; INLINE-ALL-NEXT: Merging context profile into base profile: _Z8funcLeafi 46 47; Test we inlined the following in top-down order and promot rest not inlined context profile into base profile 48; _Z5funcAi:1 @ _Z8funcLeafi 49; _Z5funcBi:1 @ _Z8funcLeafi 50; INLINE-HOT: Getting base profile for function: main 51; INLINE-HOT-NEXT: Merging context profile into base profile: main 52; INLINE-HOT-NEXT: Found context tree root to promote: external:12 @ main 53; INLINE-HOT-NEXT: Context promoted and merged to: main 54; INLINE-HOT-NEXT: Getting callee context for instr: %call = tail call i32 @_Z5funcBi 55; INLINE-HOT-NEXT: Callee context found: main:3.1 @ _Z5funcBi 56; INLINE-HOT-NEXT: Getting callee context for instr: %call1 = tail call i32 @_Z5funcAi 57; INLINE-HOT-NEXT: Callee context found: main:3 @ _Z5funcAi 58; INLINE-HOT-NEXT: Getting base profile for function: _Z5funcAi 59; INLINE-HOT-NEXT: Merging context profile into base profile: _Z5funcAi 60; INLINE-HOT-NEXT: Found context tree root to promote: main:3 @ _Z5funcAi 61; INLINE-HOT-NEXT: Context promoted to: _Z5funcAi 62; INLINE-HOT-NEXT: Context promoted to: _Z5funcAi:1 @ _Z8funcLeafi 63; INLINE-HOT-NEXT: Getting callee context for instr: %call = tail call i32 @_Z8funcLeafi(i32 %add), !dbg !50 64; INLINE-HOT-NEXT: Callee context found: _Z5funcAi:1 @ _Z8funcLeafi 65; INLINE-HOT-NEXT: Marking context profile as inlined: _Z5funcAi:1 @ _Z8funcLeafi 66; INLINE-HOT-NEXT: Getting callee context for instr: %call.i = tail call i32 @_Z3fibi(i32 %tmp.i) #2, !dbg !62 67; INLINE-HOT-NEXT: Getting callee context for instr: %call5.i = tail call i32 @_Z3fibi(i32 %tmp1.i) #2, !dbg !69 68; INLINE-HOT-NEXT: Getting base profile for function: _Z5funcBi 69; INLINE-HOT-NEXT: Merging context profile into base profile: _Z5funcBi 70; INLINE-HOT-NEXT: Found context tree root to promote: external:10 @ _Z5funcBi 71; INLINE-HOT-NEXT: Context promoted to: _Z5funcBi 72; INLINE-HOT-NEXT: Found context tree root to promote: main:3.1 @ _Z5funcBi 73; INLINE-HOT-NEXT: Context promoted and merged to: _Z5funcBi 74; INLINE-HOT-NEXT: Context promoted to: _Z5funcBi:1 @ _Z8funcLeafi 75; INLINE-HOT-NEXT: Found context tree root to promote: externalA:17 @ _Z5funcBi 76; INLINE-HOT-NEXT: Context promoted and merged to: _Z5funcBi 77; INLINE-HOT-NEXT: Getting callee context for instr: %call = tail call i32 @_Z8funcLeafi 78; INLINE-HOT-NEXT: Callee context found: _Z5funcBi:1 @ _Z8funcLeafi 79; INLINE-HOT-NEXT: Marking context profile as inlined: _Z5funcBi:1 @ _Z8funcLeafi 80; INLINE-HOT-NEXT: Getting callee context for instr: %call.i = tail call i32 @_Z3fibi 81; INLINE-HOT-NEXT: Getting callee context for instr: %call5.i = tail call i32 @_Z3fibi 82; INLINE-HOT-NEXT: Getting base profile for function: _Z8funcLeafi 83; INLINE-HOT-NEXT: Merging context profile into base profile: _Z8funcLeafi 84 85 86@factor = dso_local global i32 3, align 4, !dbg !0 87 88define dso_local i32 @main() local_unnamed_addr #0 !dbg !18 { 89entry: 90 br label %for.body, !dbg !25 91 92for.cond.cleanup: ; preds = %for.body 93 ret i32 %add3, !dbg !27 94 95for.body: ; preds = %for.body, %entry 96 %x.011 = phi i32 [ 300000, %entry ], [ %dec, %for.body ] 97 %r.010 = phi i32 [ 0, %entry ], [ %add3, %for.body ] 98 %call = tail call i32 @_Z5funcBi(i32 %x.011), !dbg !32 99 %add = add nuw nsw i32 %x.011, 1, !dbg !31 100 %call1 = tail call i32 @_Z5funcAi(i32 %add), !dbg !28 101 %add2 = add i32 %call, %r.010, !dbg !34 102 %add3 = add i32 %add2, %call1, !dbg !35 103 %dec = add nsw i32 %x.011, -1, !dbg !36 104 %cmp = icmp eq i32 %x.011, 0, !dbg !38 105 br i1 %cmp, label %for.cond.cleanup, label %for.body, !dbg !25 106} 107 108define dso_local i32 @_Z5funcAi(i32 %x) local_unnamed_addr #1 !dbg !40 { 109entry: 110 %add = add nsw i32 %x, 100000, !dbg !44 111 %call = tail call i32 @_Z8funcLeafi(i32 %add), !dbg !45 112 ret i32 %call, !dbg !46 113} 114 115define dso_local i32 @_Z8funcLeafi(i32 %x) local_unnamed_addr #1 !dbg !54 { 116entry: 117 %cmp = icmp sgt i32 %x, 0, !dbg !57 118 br i1 %cmp, label %while.body, label %while.cond2.preheader, !dbg !59 119 120while.cond2.preheader: ; preds = %entry 121 %cmp313 = icmp slt i32 %x, 0, !dbg !60 122 br i1 %cmp313, label %while.body4, label %if.end, !dbg !63 123 124while.body: ; preds = %while.body, %entry 125 %x.addr.016 = phi i32 [ %sub, %while.body ], [ %x, %entry ] 126 %tmp = load volatile i32, i32* @factor, align 4, !dbg !64 127 %call = tail call i32 @_Z3fibi(i32 %tmp), !dbg !67 128 %sub = sub nsw i32 %x.addr.016, %call, !dbg !68 129 %cmp1 = icmp sgt i32 %sub, 0, !dbg !69 130 br i1 %cmp1, label %while.body, label %if.end, !dbg !71 131 132while.body4: ; preds = %while.body4, %while.cond2.preheader 133 %x.addr.114 = phi i32 [ %add, %while.body4 ], [ %x, %while.cond2.preheader ] 134 %tmp1 = load volatile i32, i32* @factor, align 4, !dbg !72 135 %call5 = tail call i32 @_Z3fibi(i32 %tmp1), !dbg !74 136 %add = add nsw i32 %call5, %x.addr.114, !dbg !75 137 %cmp3 = icmp slt i32 %add, 0, !dbg !60 138 br i1 %cmp3, label %while.body4, label %if.end, !dbg !63 139 140if.end: ; preds = %while.body4, %while.body, %while.cond2.preheader 141 %x.addr.2 = phi i32 [ 0, %while.cond2.preheader ], [ %sub, %while.body ], [ %add, %while.body4 ] 142 ret i32 %x.addr.2, !dbg !76 143} 144 145define dso_local i32 @_Z5funcBi(i32 %x) local_unnamed_addr #0 !dbg !47 { 146entry: 147 %sub = add nsw i32 %x, -100000, !dbg !51 148 %call = tail call i32 @_Z8funcLeafi(i32 %sub), !dbg !52 149 ret i32 %call, !dbg !53 150} 151 152declare i32 @_Z3fibi(i32) 153 154attributes #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" } 155attributes #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" } 156 157!llvm.dbg.cu = !{!2} 158!llvm.module.flags = !{!14, !15, !16} 159!llvm.ident = !{!17} 160 161!0 = !DIGlobalVariableExpression(var: !1, expr: !DIExpression()) 162!1 = distinct !DIGlobalVariable(name: "factor", scope: !2, file: !3, line: 21, type: !13, isLocal: false, isDefinition: true) 163!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) 164!3 = !DIFile(filename: "merged.cpp", directory: "/local/autofdo") 165!4 = !{} 166!5 = !{!6, !10, !11} 167!6 = !DISubprogram(name: "funcA", linkageName: "_Z5funcAi", scope: !3, file: !3, line: 6, type: !7, flags: DIFlagPrototyped, spFlags: DISPFlagOptimized, retainedNodes: !4) 168!7 = !DISubroutineType(types: !8) 169!8 = !{!9, !9} 170!9 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed) 171!10 = !DISubprogram(name: "funcB", linkageName: "_Z5funcBi", scope: !3, file: !3, line: 7, type: !7, flags: DIFlagPrototyped, spFlags: DISPFlagOptimized, retainedNodes: !4) 172!11 = !DISubprogram(name: "funcLeaf", linkageName: "_Z8funcLeafi", scope: !3, file: !3, line: 22, type: !7, flags: DIFlagPrototyped, spFlags: DISPFlagOptimized, retainedNodes: !4) 173!12 = !{!0} 174!13 = !DIDerivedType(tag: DW_TAG_volatile_type, baseType: !9) 175!14 = !{i32 7, !"Dwarf Version", i32 4} 176!15 = !{i32 2, !"Debug Info Version", i32 3} 177!16 = !{i32 1, !"wchar_size", i32 4} 178!17 = !{!"clang version 11.0.0"} 179!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) 180!19 = !DISubroutineType(types: !20) 181!20 = !{!9} 182!21 = !{!22, !23} 183!22 = !DILocalVariable(name: "r", scope: !18, file: !3, line: 12, type: !9) 184!23 = !DILocalVariable(name: "x", scope: !24, file: !3, line: 13, type: !9) 185!24 = distinct !DILexicalBlock(scope: !18, file: !3, line: 13, column: 3) 186!25 = !DILocation(line: 13, column: 3, scope: !26) 187!26 = !DILexicalBlockFile(scope: !24, file: !3, discriminator: 2) 188!27 = !DILocation(line: 17, column: 3, scope: !18) 189!28 = !DILocation(line: 14, column: 10, scope: !29) 190!29 = distinct !DILexicalBlock(scope: !30, file: !3, line: 13, column: 37) 191!30 = distinct !DILexicalBlock(scope: !24, file: !3, line: 13, column: 3) 192!31 = !DILocation(line: 14, column: 29, scope: !29) 193!32 = !DILocation(line: 14, column: 21, scope: !33) 194!33 = !DILexicalBlockFile(scope: !29, file: !3, discriminator: 2) 195!34 = !DILocation(line: 14, column: 19, scope: !29) 196!35 = !DILocation(line: 14, column: 7, scope: !29) 197!36 = !DILocation(line: 13, column: 33, scope: !37) 198!37 = !DILexicalBlockFile(scope: !30, file: !3, discriminator: 6) 199!38 = !DILocation(line: 13, column: 26, scope: !39) 200!39 = !DILexicalBlockFile(scope: !30, file: !3, discriminator: 2) 201!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) 202!44 = !DILocation(line: 27, column: 22, scope: !40) 203!45 = !DILocation(line: 27, column: 11, scope: !40) 204!46 = !DILocation(line: 29, column: 3, scope: !40) 205!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) 206!51 = !DILocation(line: 33, column: 22, scope: !47) 207!52 = !DILocation(line: 33, column: 11, scope: !47) 208!53 = !DILocation(line: 35, column: 3, scope: !47) 209!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) 210!57 = !DILocation(line: 49, column: 9, scope: !58) 211!58 = distinct !DILexicalBlock(scope: !54, file: !3, line: 49, column: 7) 212!59 = !DILocation(line: 49, column: 7, scope: !54) 213!60 = !DILocation(line: 58, column: 14, scope: !61) 214!61 = !DILexicalBlockFile(scope: !62, file: !3, discriminator: 2) 215!62 = distinct !DILexicalBlock(scope: !58, file: !3, line: 56, column: 8) 216!63 = !DILocation(line: 58, column: 5, scope: !61) 217!64 = !DILocation(line: 52, column: 16, scope: !65) 218!65 = distinct !DILexicalBlock(scope: !66, file: !3, line: 51, column: 19) 219!66 = distinct !DILexicalBlock(scope: !58, file: !3, line: 49, column: 14) 220!67 = !DILocation(line: 52, column: 12, scope: !65) 221!68 = !DILocation(line: 52, column: 9, scope: !65) 222!69 = !DILocation(line: 51, column: 14, scope: !70) 223!70 = !DILexicalBlockFile(scope: !66, file: !3, discriminator: 2) 224!71 = !DILocation(line: 51, column: 5, scope: !70) 225!72 = !DILocation(line: 59, column: 16, scope: !73) 226!73 = distinct !DILexicalBlock(scope: !62, file: !3, line: 58, column: 19) 227!74 = !DILocation(line: 59, column: 12, scope: !73) 228!75 = !DILocation(line: 59, column: 9, scope: !73) 229!76 = !DILocation(line: 63, column: 3, scope: !54) 230