1;; Tests memprof profile matching when the leaf frame is missing in the 2;; profile. In this case the call to operator new was inlined before 3;; matching and we are able to match the next call frame up the inlined 4;; context. 5 6;; Avoid failures on big-endian systems that can't read the profile properly 7; REQUIRES: x86_64-linux 8 9;; # To generate below LLVM IR for use in matching. 10;; $ clang++ -gmlt -fdebug-info-for-profiling -S memprof_missing_leaf.cc \ 11;; -O2 -emit-llvm 12;; 13;; where memprof_missing_leaf.cc is as follows: 14;; 15;; #include <new> 16;; 17;; // Use musttail to simulate a missing leaf debug frame in the profiled binary. 18;; // Note we don't currently match onto explicit ::operator new calls, which is 19;; // why the non-musttail case uses implicit new (which doesn't support musttail). 20;; // Note that changes in the code below which affect relative line number 21;; // offsets of calls from their parent function can affect callsite matching in 22;; // the LLVM IR. 23;; #ifndef USE_MUSTTAIL 24;; #define USE_MUSTTAIL 0 25;; #endif 26;; 27;; // clang::musttail requires that the argument signature matches that of the caller. 28;; void *bar(std::size_t s) { 29;; #if USE_MUSTTAIL 30;; [[clang::musttail]] return ::operator new (s); 31;; #else 32;; return new char[s]; 33;; #endif 34;; } 35;; 36;; int main() { 37;; char *a = (char *)bar(1); 38;; delete a; 39;; return 0; 40;;} 41 42; RUN: split-file %s %t 43; RUN: llvm-profdata merge %t/memprof_missing_leaf.yaml -o %t/memprof_missing_leaf.memprofdata 44; RUN: opt < %t/memprof_missing_leaf.ll -passes='memprof-use<profile-filename=%t/memprof_missing_leaf.memprofdata>' -S | FileCheck %s 45 46;--- memprof_missing_leaf.yaml 47--- 48HeapProfileRecords: 49 - GUID: main 50 AllocSites: 51 - Callstack: 52 - { Function: main, LineOffset: 1, Column: 21, IsInlineFrame: false } 53 MemInfoBlock: 54 AllocCount: 1 55 TotalSize: 1 56 TotalLifetime: 0 57 TotalLifetimeAccessDensity: 0 58 CallSites: [] 59... 60;--- memprof_missing_leaf.ll 61; CHECK: call {{.*}} @_Znam{{.*}} #[[ATTR:[0-9]+]] 62; CHECK: attributes #[[ATTR]] = {{.*}} "memprof"="notcold" 63 64; ModuleID = '<stdin>' 65source_filename = "memprof_missing_leaf.cc" 66target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128" 67target triple = "x86_64-unknown-linux-gnu" 68 69; Function Attrs: nobuiltin allocsize(0) 70declare noundef nonnull ptr @_Znam(i64 noundef) #0 71 72; Function Attrs: mustprogress norecurse uwtable 73define dso_local noundef i32 @main() #1 !dbg !8 { 74entry: 75 %s.addr.i = alloca i64, align 8 76 %retval = alloca i32, align 4 77 %a = alloca ptr, align 8 78 store i32 0, ptr %retval, align 4 79 store i64 1, ptr %s.addr.i, align 8, !tbaa !11 80 %0 = load i64, ptr %s.addr.i, align 8, !dbg !15, !tbaa !11 81 %call.i = call noalias noundef nonnull ptr @_Znam(i64 noundef %0) #3, !dbg !18 82 store ptr %call.i, ptr %a, align 8, !dbg !19, !tbaa !20 83 %1 = load ptr, ptr %a, align 8, !dbg !22, !tbaa !20 84 %isnull = icmp eq ptr %1, null, !dbg !23 85 br i1 %isnull, label %delete.end, label %delete.notnull, !dbg !23 86 87delete.notnull: ; preds = %entry 88 call void @_ZdlPv(ptr noundef %1) #4, !dbg !23 89 br label %delete.end, !dbg !23 90 91delete.end: ; preds = %delete.notnull, %entry 92 ret i32 0, !dbg !24 93} 94 95; Function Attrs: nobuiltin nounwind 96declare void @_ZdlPv(ptr noundef) #2 97 98attributes #0 = { nobuiltin allocsize(0) "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" } 99attributes #1 = { mustprogress norecurse uwtable "min-legal-vector-width"="0" "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" } 100attributes #2 = { nobuiltin nounwind "no-trapping-math"="true" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+cmov,+cx8,+fxsr,+mmx,+sse,+sse2,+x87" "tune-cpu"="generic" } 101attributes #3 = { builtin allocsize(0) } 102attributes #4 = { builtin nounwind } 103 104!llvm.dbg.cu = !{!0} 105!llvm.module.flags = !{!2, !3, !4, !5, !6, !7} 106 107!0 = distinct !DICompileUnit(language: DW_LANG_C_plus_plus_14, file: !1, producer: "clang version 18.0.0 (git@github.com:llvm/llvm-project.git 71bf052ec90e77cb4aa66505d47cbc4b6016ac1d)", isOptimized: true, runtimeVersion: 0, emissionKind: LineTablesOnly, splitDebugInlining: false, debugInfoForProfiling: true, nameTableKind: None) 108!1 = !DIFile(filename: "memprof_missing_leaf.cc", directory: ".", checksumkind: CSK_MD5, checksum: "f1445a8699406a6b826128704d257677") 109!2 = !{i32 7, !"Dwarf Version", i32 5} 110!3 = !{i32 2, !"Debug Info Version", i32 3} 111!4 = !{i32 1, !"wchar_size", i32 4} 112!5 = !{i32 8, !"PIC Level", i32 2} 113!6 = !{i32 7, !"PIE Level", i32 2} 114!7 = !{i32 7, !"uwtable", i32 2} 115!8 = distinct !DISubprogram(name: "main", scope: !1, file: !1, line: 15, type: !9, scopeLine: 15, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0) 116!9 = !DISubroutineType(types: !10) 117!10 = !{} 118!11 = !{!12, !12, i64 0} 119!12 = !{!"long", !13, i64 0} 120!13 = !{!"omnipotent char", !14, i64 0} 121!14 = !{!"Simple C++ TBAA"} 122!15 = !DILocation(line: 11, column: 19, scope: !16, inlinedAt: !17) 123!16 = distinct !DISubprogram(name: "bar", linkageName: "_Z3barm", scope: !1, file: !1, line: 7, type: !9, scopeLine: 7, flags: DIFlagPrototyped | DIFlagAllCallsDescribed, spFlags: DISPFlagDefinition | DISPFlagOptimized, unit: !0) 124!17 = distinct !DILocation(line: 16, column: 21, scope: !8) 125!18 = !DILocation(line: 11, column: 10, scope: !16, inlinedAt: !17) 126!19 = !DILocation(line: 16, column: 9, scope: !8) 127!20 = !{!21, !21, i64 0} 128!21 = !{!"any pointer", !13, i64 0} 129!22 = !DILocation(line: 17, column: 10, scope: !8) 130!23 = !DILocation(line: 17, column: 3, scope: !8) 131!24 = !DILocation(line: 18, column: 3, scope: !8) 132