199 lines · plain
1;; Test for different function processing orders affecting inlining in sample profile loader.2 3;; There is an SCC _Z5funcAi -> _Z8funcLeafi -> _Z5funcAi in the program.4;; With -use-profiled-call-graph=0, the top-down processing order of5;; that SCC is (_Z8funcLeafi, _Z5funcAi), which is determinined based on6;; the static call graph. With -use-profiled-call-graph=1, call edges7;; from profile are considered, thus the order becomes (_Z5funcAi, _Z8funcLeafi)8;; which leads to _Z8funcLeafi inlined into _Z5funcAi.9; RUN: opt < %s -passes=sample-profile -use-profiled-call-graph=1 -sample-profile-file=%S/Inputs/profile-context-order.prof -S | FileCheck %s -check-prefix=INLINE10; RUN: opt < %s -passes=sample-profile -use-profiled-call-graph=0 -sample-profile-file=%S/Inputs/profile-context-order.prof -S | FileCheck %s -check-prefix=NOINLINE11 12;; There is an indirect call _Z5funcAi -> _Z3fibi in the program.13;; With -use-profiled-call-graph=0, the processing order computed14;; based on the static call graph is (_Z3fibi, _Z5funcAi). With 15;; -use-profiled-call-graph=1, the indirect call edge from profile is16;; considered, thus the order becomes (_Z5funcAi, _Z3fibi) which leads to17;; _Z3fibi inlined into _Z5funcAi.18; RUN: opt < %s -passes=sample-profile -use-profiled-call-graph=1 -sample-profile-file=%S/Inputs/profile-context-order.prof -S | FileCheck %s -check-prefix=ICALL-INLINE19 20;; When a cycle is formed by profiled edges between _Z5funcBi and _Z8funcLeafi,21;; the function processing order matters. Without considering call edge weights22;; _Z8funcLeafi can be processed before _Z5funcBi, thus leads to suboptimal23;; inlining.24; RUN: opt < %s -passes=sample-profile -use-profiled-call-graph=1 -sort-profiled-scc-member=0 -sample-profile-file=%S/Inputs/profile-context-order-scc.prof -S | FileCheck %s -check-prefix=NOINLINEB25; RUN: opt < %s -passes=sample-profile -use-profiled-call-graph=1 -sort-profiled-scc-member=1 -sample-profile-file=%S/Inputs/profile-context-order-scc.prof -S | FileCheck %s -check-prefix=INLINEB26 27 28@factor = dso_local global i32 3, align 4, !dbg !029@fp = dso_local global ptr null, align 830 31; INLINE: define dso_local i32 @_Z5funcAi32; INLINE-NOT: call i32 @_Z8funcLeafi33; NOINLINE: define dso_local i32 @_Z5funcAi34; NOINLINE: call i32 @_Z8funcLeafi35; ICALL-INLINE: define dso_local i32 @_Z5funcAi36; ICALL-INLINE: call i32 @_Z3foo37; INLINEB: define dso_local i32 @_Z5funcBi38; INLINEB-NOT: call i32 @_Z8funcLeafi39; NOINLINEB: define dso_local i32 @_Z5funcBi40; NOINLINEB: call i32 @_Z8funcLeafi41define dso_local i32 @_Z5funcAi(i32 %x) local_unnamed_addr #0 !dbg !40 {42entry:43 %add = add nsw i32 %x, 100000, !dbg !4444 %0 = load ptr, ptr @fp, align 845 %call = call i32 %0(i32 8), !dbg !4546 %call1 = tail call i32 @_Z8funcLeafi(i32 %add), !dbg !4647 ret i32 %call, !dbg !4648}49 50define dso_local i32 @main() local_unnamed_addr #0 !dbg !18 {51entry:52 store ptr @_Z3fibi, ptr @fp, align 8, !dbg !2553 br label %for.body, !dbg !2554 55for.cond.cleanup: ; preds = %for.body56 ret i32 %add3, !dbg !2757 58for.body: ; preds = %for.body, %entry59 %x.011 = phi i32 [ 300000, %entry ], [ %dec, %for.body ]60 %r.010 = phi i32 [ 0, %entry ], [ %add3, %for.body ]61 %call = tail call i32 @_Z5funcBi(i32 %x.011), !dbg !3262 %add = add nuw nsw i32 %x.011, 1, !dbg !3163 %call1 = tail call i32 @_Z5funcAi(i32 %add), !dbg !2864 %add2 = add i32 %call, %r.010, !dbg !3465 %add3 = add i32 %add2, %call1, !dbg !3566 %dec = add nsw i32 %x.011, -1, !dbg !3667 %cmp = icmp eq i32 %x.011, 0, !dbg !3868 br i1 %cmp, label %for.cond.cleanup, label %for.body, !dbg !2569}70 71; INLINE: define dso_local i32 @_Z8funcLeafi72; NOINLINE: define dso_local i32 @_Z8funcLeafi73; ICALL-INLINE: define dso_local i32 @_Z8funcLeafi74; ICALL-NOINLINE: define dso_local i32 @_Z8funcLeafi75define dso_local i32 @_Z8funcLeafi(i32 %x) local_unnamed_addr #1 !dbg !54 {76entry:77 %cmp = icmp sgt i32 %x, 0, !dbg !5778 br i1 %cmp, label %while.body, label %while.cond2.preheader, !dbg !5979 80while.cond2.preheader: ; preds = %entry81 %cmp313 = icmp slt i32 %x, 0, !dbg !6082 br i1 %cmp313, label %while.body4, label %if.end, !dbg !6383 84while.body: ; preds = %while.body, %entry85 %x.addr.016 = phi i32 [ %sub, %while.body ], [ %x, %entry ]86 %tmp = load volatile i32, ptr @factor, align 4, !dbg !6487 %call = tail call i32 @_Z5funcAi(i32 %tmp), !dbg !6788 %sub = sub nsw i32 %x.addr.016, %call, !dbg !6889 %cmp1 = icmp sgt i32 %sub, 0, !dbg !6990 br i1 %cmp1, label %while.body, label %if.end, !dbg !7191 92while.body4: ; preds = %while.body4, %while.cond2.preheader93 %x.addr.114 = phi i32 [ %add, %while.body4 ], [ %x, %while.cond2.preheader ]94 %tmp1 = load volatile i32, ptr @factor, align 4, !dbg !7295 %call5 = tail call i32 @_Z5funcBi(i32 %tmp1), !dbg !7496 %add = add nsw i32 %call5, %x.addr.114, !dbg !7597 %cmp3 = icmp slt i32 %add, 0, !dbg !6098 br i1 %cmp3, label %while.body4, label %if.end, !dbg !6399 100if.end: ; preds = %while.body4, %while.body, %while.cond2.preheader101 %x.addr.2 = phi i32 [ 0, %while.cond2.preheader ], [ %sub, %while.body ], [ %add, %while.body4 ]102 ret i32 %x.addr.2, !dbg !76103}104 105define dso_local i32 @_Z5funcBi(i32 %x) local_unnamed_addr #0 !dbg !47 {106entry:107 %sub = add nsw i32 %x, -100000, !dbg !51108 %call = tail call i32 @_Z8funcLeafi(i32 %sub), !dbg !52109 ret i32 %call, !dbg !53110}111 112define dso_local i32 @_Z3fibi(i32 %x) local_unnamed_addr #1 !dbg !77 {113entry:114 %sub = add nsw i32 %x, -100000, !dbg !78115 %call = tail call i32 @_Z3foo(i32 %sub), !dbg !78116 ret i32 %sub, !dbg !78117}118 119declare i32 @_Z3foo(i32)120 121attributes #0 = { nofree noinline nounwind uwtable "use-sample-profile" }122attributes #1 = { nofree nounwind uwtable "use-sample-profile" }123 124!llvm.dbg.cu = !{!2}125!llvm.module.flags = !{!14, !15, !16}126!llvm.ident = !{!17}127 128!0 = !DIGlobalVariableExpression(var: !1, expr: !DIExpression())129!1 = distinct !DIGlobalVariable(name: "factor", scope: !2, file: !3, line: 21, type: !13, isLocal: false, isDefinition: true)130!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)131!3 = !DIFile(filename: "merged.cpp", directory: "/local/autofdo")132!4 = !{}133!5 = !{!6, !10, !11}134!6 = !DISubprogram(name: "funcA", linkageName: "_Z5funcAi", scope: !3, file: !3, line: 6, type: !7, flags: DIFlagPrototyped, spFlags: DISPFlagOptimized, retainedNodes: !4)135!7 = !DISubroutineType(types: !8)136!8 = !{!9, !9}137!9 = !DIBasicType(name: "int", size: 32, encoding: DW_ATE_signed)138!10 = !DISubprogram(name: "funcB", linkageName: "_Z5funcBi", scope: !3, file: !3, line: 7, type: !7, flags: DIFlagPrototyped, spFlags: DISPFlagOptimized, retainedNodes: !4)139!11 = !DISubprogram(name: "funcLeaf", linkageName: "_Z8funcLeafi", scope: !3, file: !3, line: 22, type: !7, flags: DIFlagPrototyped, spFlags: DISPFlagOptimized, retainedNodes: !4)140!12 = !{!0}141!13 = !DIDerivedType(tag: DW_TAG_volatile_type, baseType: !9)142!14 = !{i32 7, !"Dwarf Version", i32 4}143!15 = !{i32 2, !"Debug Info Version", i32 3}144!16 = !{i32 1, !"wchar_size", i32 4}145!17 = !{!"clang version 11.0.0"}146!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)147!19 = !DISubroutineType(types: !20)148!20 = !{!9}149!21 = !{!22, !23}150!22 = !DILocalVariable(name: "r", scope: !18, file: !3, line: 12, type: !9)151!23 = !DILocalVariable(name: "x", scope: !24, file: !3, line: 13, type: !9)152!24 = distinct !DILexicalBlock(scope: !18, file: !3, line: 13, column: 3)153!25 = !DILocation(line: 13, column: 3, scope: !26)154!26 = !DILexicalBlockFile(scope: !24, file: !3, discriminator: 2)155!27 = !DILocation(line: 17, column: 3, scope: !18)156!28 = !DILocation(line: 14, column: 10, scope: !29)157!29 = distinct !DILexicalBlock(scope: !30, file: !3, line: 13, column: 37)158!30 = distinct !DILexicalBlock(scope: !24, file: !3, line: 13, column: 3)159!31 = !DILocation(line: 14, column: 29, scope: !29)160!32 = !DILocation(line: 14, column: 21, scope: !33)161!33 = !DILexicalBlockFile(scope: !29, file: !3, discriminator: 2)162!34 = !DILocation(line: 14, column: 19, scope: !29)163!35 = !DILocation(line: 14, column: 7, scope: !29)164!36 = !DILocation(line: 13, column: 33, scope: !37)165!37 = !DILexicalBlockFile(scope: !30, file: !3, discriminator: 6)166!38 = !DILocation(line: 13, column: 26, scope: !39)167!39 = !DILexicalBlockFile(scope: !30, file: !3, discriminator: 2)168!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)169!44 = !DILocation(line: 26, column: 22, scope: !40)170!45 = !DILocation(line: 28, column: 11, scope: !40)171!46 = !DILocation(line: 27, column: 3, scope: !40)172!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)173!51 = !DILocation(line: 33, column: 22, scope: !47)174!52 = !DILocation(line: 33, column: 11, scope: !47)175!53 = !DILocation(line: 35, column: 3, scope: !47)176!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)177!57 = !DILocation(line: 49, column: 9, scope: !58)178!58 = distinct !DILexicalBlock(scope: !54, file: !3, line: 49, column: 7)179!59 = !DILocation(line: 49, column: 7, scope: !54)180!60 = !DILocation(line: 58, column: 14, scope: !61)181!61 = !DILexicalBlockFile(scope: !62, file: !3, discriminator: 2)182!62 = distinct !DILexicalBlock(scope: !58, file: !3, line: 56, column: 8)183!63 = !DILocation(line: 58, column: 5, scope: !61)184!64 = !DILocation(line: 52, column: 16, scope: !65)185!65 = distinct !DILexicalBlock(scope: !66, file: !3, line: 51, column: 19)186!66 = distinct !DILexicalBlock(scope: !58, file: !3, line: 49, column: 14)187!67 = !DILocation(line: 52, column: 12, scope: !65)188!68 = !DILocation(line: 52, column: 9, scope: !65)189!69 = !DILocation(line: 51, column: 14, scope: !70)190!70 = !DILexicalBlockFile(scope: !66, file: !3, discriminator: 2)191!71 = !DILocation(line: 51, column: 5, scope: !70)192!72 = !DILocation(line: 59, column: 16, scope: !73)193!73 = distinct !DILexicalBlock(scope: !62, file: !3, line: 58, column: 19)194!74 = !DILocation(line: 59, column: 12, scope: !73)195!75 = !DILocation(line: 59, column: 9, scope: !73)196!76 = !DILocation(line: 63, column: 3, scope: !54)197!77 = distinct !DISubprogram(name: "funcB", linkageName: "_Z3fibi", scope: !3, file: !3, line: 32, type: !7, scopeLine: 32, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !2)198!78 = !DILocation(line: 33, column: 22, scope: !77)199