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