1;; Tests memprof when contains loop unroll of allocation, where the unrolled 2;; allocations have the same context but different allocation types. 3 4;; Avoid failures on big-endian systems that can't read the profile properly 5; REQUIRES: x86_64-linux 6 7;; TODO: Use text profile inputs once that is available for memprof. 8;; # To update the Inputs below, run Inputs/update_memprof_inputs.sh. 9;; # To generate below LLVM IR for use in matching. 10;; $ clang++ -gmlt -fdebug-info-for-profiling -S %S/Inputs/memprof_loop_unroll_b.cc -emit-llvm 11 12; RUN: llvm-profdata merge %S/Inputs/memprof_loop_unroll.memprofraw --profiled-binary %S/Inputs/memprof_loop_unroll.exe -o %t.memprofdata 13;; Set the minimum lifetime threshold to 0 to ensure that one context is 14;; considered cold (the other will be notcold). 15; RUN: opt < %s -passes='memprof-use<profile-filename=%t.memprofdata>' -S -memprof-report-hinted-sizes -memprof-ave-lifetime-cold-threshold=0 2>&1 | FileCheck %s 16 17;; Conservatively annotate as not cold. We get two messages as there are two 18;; unrolled copies of the allocation. 19; CHECK: MemProf hinting: Total size for full allocation context hash {{.*}} and indistinguishable alloc type notcold: 4 20; CHECK: MemProf hinting: Total size for full allocation context hash {{.*}} and indistinguishable alloc type notcold: 4 21; CHECK: call {{.*}} @_Znam{{.*}} #[[ATTR:[0-9]+]] 22; CHECK: attributes #[[ATTR]] = { builtin allocsize(0) "memprof"="notcold" } 23; CHECK-NOT: stackIds: () 24 25; ModuleID = 'memprof_loop_unroll_b.cc' 26source_filename = "memprof_loop_unroll_b.cc" 27target datalayout = "e-m:e-p270:32:32-p271:32:32-p272:64:64-i64:64-i128:128-f80:128-n8:16:32:64-S128" 28target triple = "x86_64-unknown-linux-gnu" 29 30@a = external global [2 x ptr], align 16 31 32; Function Attrs: mustprogress noinline optnone uwtable 33define dso_local void @_Z3foov() #0 !dbg !10 { 34entry: 35 %i = alloca i32, align 4 36 store i32 0, ptr %i, align 4, !dbg !13 37 br label %for.cond, !dbg !14 38 39for.cond: ; preds = %for.inc, %entry 40 %0 = load i32, ptr %i, align 4, !dbg !15 41 %cmp = icmp slt i32 %0, 2, !dbg !17 42 br i1 %cmp, label %for.body, label %for.end, !dbg !18 43 44for.body: ; preds = %for.cond 45 %call = call noalias noundef nonnull ptr @_Znam(i64 noundef 4) #2, !dbg !19 46 %1 = load i32, ptr %i, align 4, !dbg !20 47 %idxprom = sext i32 %1 to i64, !dbg !21 48 %arrayidx = getelementptr inbounds [2 x ptr], ptr @a, i64 0, i64 %idxprom, !dbg !21 49 store ptr %call, ptr %arrayidx, align 8, !dbg !22 50 br label %for.inc, !dbg !23 51 52for.inc: ; preds = %for.body 53 %2 = load i32, ptr %i, align 4, !dbg !24 54 %inc = add nsw i32 %2, 1, !dbg !24 55 store i32 %inc, ptr %i, align 4, !dbg !24 56 br label %for.cond, !dbg !26, !llvm.loop !27 57 58for.end: ; preds = %for.cond 59 ret void, !dbg !30 60} 61 62; Function Attrs: nobuiltin allocsize(0) 63declare noundef nonnull ptr @_Znam(i64 noundef) #1 64 65attributes #0 = { mustprogress noinline optnone uwtable "frame-pointer"="all" "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" } 66attributes #1 = { nobuiltin allocsize(0) "frame-pointer"="all" "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" } 67attributes #2 = { builtin allocsize(0) } 68 69!llvm.dbg.cu = !{!0} 70!llvm.module.flags = !{!2, !3, !4, !5, !6, !7, !8} 71!llvm.ident = !{!9} 72 73!0 = distinct !DICompileUnit(language: DW_LANG_C_plus_plus_14, file: !1, producer: "clang version 18.0.0", isOptimized: false, runtimeVersion: 0, emissionKind: LineTablesOnly, splitDebugInlining: false, debugInfoForProfiling: true, nameTableKind: None) 74!1 = !DIFile(filename: "memprof_loop_unroll_b.cc", directory: "/", checksumkind: CSK_MD5, checksum: "00276e590d606451dc54f3ff5f4bba25") 75!2 = !{i32 7, !"Dwarf Version", i32 5} 76!3 = !{i32 2, !"Debug Info Version", i32 3} 77!4 = !{i32 1, !"wchar_size", i32 4} 78!5 = !{i32 8, !"PIC Level", i32 2} 79!6 = !{i32 7, !"PIE Level", i32 2} 80!7 = !{i32 7, !"uwtable", i32 2} 81!8 = !{i32 7, !"frame-pointer", i32 2} 82!9 = !{!"clang version 18.0.0"} 83!10 = distinct !DISubprogram(name: "foo", linkageName: "_Z3foov", scope: !1, file: !1, line: 4, type: !11, scopeLine: 4, flags: DIFlagPrototyped, spFlags: DISPFlagDefinition, unit: !0) 84!11 = !DISubroutineType(types: !12) 85!12 = !{} 86!13 = !DILocation(line: 5, column: 14, scope: !10) 87!14 = !DILocation(line: 5, column: 10, scope: !10) 88!15 = !DILocation(line: 5, column: 21, scope: !16) 89!16 = !DILexicalBlockFile(scope: !10, file: !1, discriminator: 2) 90!17 = !DILocation(line: 5, column: 23, scope: !16) 91!18 = !DILocation(line: 5, column: 5, scope: !16) 92!19 = !DILocation(line: 6, column: 16, scope: !10) 93!20 = !DILocation(line: 6, column: 11, scope: !10) 94!21 = !DILocation(line: 6, column: 9, scope: !10) 95!22 = !DILocation(line: 6, column: 14, scope: !10) 96!23 = !DILocation(line: 7, column: 5, scope: !10) 97!24 = !DILocation(line: 5, column: 28, scope: !25) 98!25 = !DILexicalBlockFile(scope: !10, file: !1, discriminator: 4) 99!26 = !DILocation(line: 5, column: 5, scope: !25) 100!27 = distinct !{!27, !28, !23, !29} 101!28 = !DILocation(line: 5, column: 5, scope: !10) 102!29 = !{!"llvm.loop.mustprogress"} 103!30 = !DILocation(line: 8, column: 1, scope: !10) 104