1; RUN: opt < %s -passes=sample-profile -sample-profile-file=%S/Inputs/profsampleacc.extbinary.afdo -profile-summary-cutoff-hot=900000 -profile-sample-accurate -S | FileCheck %s --check-prefix=CALL_SUM_IS_HOT 2; RUN: opt < %s -passes=sample-profile -sample-profile-file=%S/Inputs/profsampleacc.extbinary.afdo -profile-summary-cutoff-hot=900000 -profile-sample-accurate -S | FileCheck %s --check-prefix=CALL_SUM_IS_HOT 3 4; RUN: llvm-profdata merge -sample -extbinary -prof-sym-list=%S/Inputs/profile-symbol-list.text %S/Inputs/profsampleacc.extbinary.afdo -o %t.symlist.afdo 5; RUN: opt < %s -passes=sample-profile -sample-profile-file=%t.symlist.afdo -profile-summary-cutoff-hot=600000 -profile-accurate-for-symsinlist -S | FileCheck %s --check-prefix=PROFSYMLIST 6; RUN: opt < %s -passes=sample-profile -sample-profile-file=%t.symlist.afdo -profile-accurate-for-symsinlist -profile-symbol-list-cutoff=2 -S | FileCheck %s --check-prefix=PSLCUTOFF2 7; RUN: opt < %s -passes=sample-profile -sample-profile-file=%t.symlist.afdo -profile-accurate-for-symsinlist -profile-symbol-list-cutoff=3 -S | FileCheck %s --check-prefix=PSLCUTOFF3 8; 9; If -profile-accurate-for-symsinlist and -profile-sample-accurate both present, 10; -profile-sample-accurate will override -profile-accurate-for-symsinlist. 11; RUN: opt < %s -passes=sample-profile -sample-profile-file=%S/Inputs/profsampleacc.extbinary.afdo -profile-summary-cutoff-hot=900000 -profile-sample-accurate -profile-accurate-for-symsinlist -S | FileCheck %s --check-prefix=CALL_SUM_IS_HOT 12; RUN: opt < %s -passes=sample-profile -sample-profile-file=%S/Inputs/profsampleacc.extbinary.afdo -profile-summary-cutoff-hot=900000 -profile-sample-accurate -profile-accurate-for-symsinlist -S | FileCheck %s --check-prefix=CALL_SUM_IS_HOT 13; 14; Original C++ test case 15; 16; #include <stdio.h> 17; 18; int sum(int x, int y) { 19; return x + y; 20; } 21; 22; int main() { 23; int s, i = 0; 24; while (i++ < 20000 * 20000) 25; if (i != 100) s = sum(i, s); else s = 30; 26; printf("sum is %d\n", s); 27; return 0; 28; } 29; 30@.str = private unnamed_addr constant [11 x i8] c"sum is %d\0A\00", align 1 31 32; Check _Z3sumii's function entry count will be 0 when 33; profile-sample-accurate is enabled. 34; CALL_SUM_IS_HOT: define i32 @_Z3sumii{{.*}}!prof ![[ZERO_ID:[0-9]+]] 35; 36; Check _Z3sumii's function entry count will be nonzero when 37; profile-sample-accurate is enabled because the callsite is warm and not 38; inlined so its function entry count is adjusted to nonzero. 39; CALL_SUM_IS_WARM: define i32 @_Z3sumii{{.*}}!prof ![[NONZERO_ID:[0-9]+]] 40; 41; Check _Z3sumii's function entry count will be initialized to -1 when 42; profile-accurate-for-profsymlist is enabled and _Z3sumii exists in the 43; profile symbol list because it also shows up in the profile as inline 44; instance. 45; PROFSYMLIST: define i32 @_Z3sumii{{.*}}!prof ![[UNKNOWN_ID:[0-9]+]] 46; 47; Function Attrs: nounwind uwtable 48define i32 @_Z3sumii(i32 %x, i32 %y) #0 !dbg !4 { 49entry: 50 %x.addr = alloca i32, align 4 51 %y.addr = alloca i32, align 4 52 store i32 %x, ptr %x.addr, align 4 53 store i32 %y, ptr %y.addr, align 4 54 %0 = load i32, ptr %x.addr, align 4, !dbg !11 55 %1 = load i32, ptr %y.addr, align 4, !dbg !11 56 %add = add nsw i32 %0, %1, !dbg !11 57 ret i32 %add, !dbg !11 58} 59 60; Check -profile-symbol-list-cutoff=3 will include _Z3toov into profile 61; symbol list and -profile-symbol-list-cutoff=2 will not. 62; PSLCUTOFF2: define i32 @_Z3toov{{.*}}!prof ![[TOO_ID:[0-9]+]] 63; PSLCUTOFF3: define i32 @_Z3toov{{.*}}!prof ![[TOO_ID:[0-9]+]] 64define i32 @_Z3toov(i32 %x, i32 %y) #0 { 65entry: 66 %add = add nsw i32 %x, %y 67 ret i32 %add 68} 69 70; Function Attrs: uwtable 71define i32 @main() #0 !dbg !7 { 72entry: 73 %retval = alloca i32, align 4 74 %s = alloca i32, align 4 75 %i = alloca i32, align 4 76 store i32 0, ptr %retval 77 store i32 0, ptr %i, align 4, !dbg !12 78 br label %while.cond, !dbg !13 79 80while.cond: ; preds = %if.end, %entry 81 %0 = load i32, ptr %i, align 4, !dbg !14 82 %inc = add nsw i32 %0, 1, !dbg !14 83 store i32 %inc, ptr %i, align 4, !dbg !14 84 %cmp = icmp slt i32 %0, 400000000, !dbg !14 85 br i1 %cmp, label %while.body, label %while.end, !dbg !14 86 87while.body: ; preds = %while.cond 88 %1 = load i32, ptr %i, align 4, !dbg !16 89 %cmp1 = icmp ne i32 %1, 100, !dbg !16 90 br i1 %cmp1, label %if.then, label %if.else, !dbg !16 91 92; With the hot cutoff being set to 600000, the inline instance of _Z3sumii 93; in main is neither hot nor cold. Check it won't be inlined when 94; profile-sample-accurate is enabled. 95; CALL_SUM_IS_WARM: if.then: 96; CALL_SUM_IS_WARM: call i32 @_Z3sumii 97; CALL_SUM_IS_WARM: if.else: 98; 99; With the hot cutoff being set to 900000, the inline instance of _Z3sumii 100; in main is hot. Check the callsite of _Z3sumii will be inlined when 101; profile-sample-accurate is enabled. 102; CALL_SUM_IS_HOT: if.then: 103; CALL_SUM_IS_HOT-NOT: call i32 @_Z3sumii 104; CALL_SUM_IS_HOT: if.else: 105; 106; Check _Z3sumii will be inlined when profile-accurate-for-profsymlist is 107; enabled 108; PROFSYMLIST: if.then: 109; PROFSYMLIST-NOT: call i32 @_Z3sumii 110; PROFSYMLIST: if.else: 111if.then: ; preds = %while.body 112 %2 = load i32, ptr %i, align 4, !dbg !18 113 %3 = load i32, ptr %s, align 4, !dbg !18 114 %call = call i32 @_Z3sumii(i32 %2, i32 %3), !dbg !18 115 store i32 %call, ptr %s, align 4, !dbg !18 116 br label %if.end, !dbg !18 117 118if.else: ; preds = %while.body 119 store i32 30, ptr %s, align 4, !dbg !20 120 br label %if.end 121 122if.end: ; preds = %if.else, %if.then 123 br label %while.cond, !dbg !22 124 125while.end: ; preds = %while.cond 126 %4 = load i32, ptr %s, align 4, !dbg !24 127 %call2 = call i32 (ptr, ...) @printf(ptr @.str, i32 %4), !dbg !24 128 ret i32 0, !dbg !25 129} 130 131declare i32 @printf(ptr, ...) #2 132 133attributes #0 = { "use-sample-profile" } 134 135!llvm.dbg.cu = !{!0} 136!llvm.module.flags = !{!8, !9} 137!llvm.ident = !{!10} 138 139; CALL_SUM_IS_HOT: ![[ZERO_ID]] = !{!"function_entry_count", i64 0} 140; CALL_SUM_IS_WARM: ![[NONZERO_ID]] = !{!"function_entry_count", i64 5179} 141; PROFSYMLIST: ![[UNKNOWN_ID]] = !{!"function_entry_count", i64 -1} 142; PSLCUTOFF2: ![[TOO_ID]] = !{!"function_entry_count", i64 -1} 143; PSLCUTOFF3: ![[TOO_ID]] = !{!"function_entry_count", i64 0} 144!0 = distinct !DICompileUnit(language: DW_LANG_C_plus_plus, producer: "clang version 3.5 ", isOptimized: false, emissionKind: NoDebug, file: !1, enums: !2, retainedTypes: !2, globals: !2, imports: !2) 145!1 = !DIFile(filename: "calls.cc", directory: ".") 146!2 = !{} 147!4 = distinct !DISubprogram(name: "sum", line: 3, isLocal: false, isDefinition: true, virtualIndex: 6, flags: DIFlagPrototyped, isOptimized: false, unit: !0, scopeLine: 3, file: !1, scope: !5, type: !6, retainedNodes: !2) 148!5 = !DIFile(filename: "calls.cc", directory: ".") 149!6 = !DISubroutineType(types: !2) 150!7 = distinct !DISubprogram(name: "main", line: 7, isLocal: false, isDefinition: true, virtualIndex: 6, flags: DIFlagPrototyped, isOptimized: false, unit: !0, scopeLine: 7, file: !1, scope: !5, type: !6, retainedNodes: !2) 151!8 = !{i32 2, !"Dwarf Version", i32 4} 152!9 = !{i32 1, !"Debug Info Version", i32 3} 153!10 = !{!"clang version 3.5 "} 154!11 = !DILocation(line: 4, scope: !4) 155!12 = !DILocation(line: 8, scope: !7) 156!13 = !DILocation(line: 9, scope: !7) 157!14 = !DILocation(line: 9, scope: !15) 158!15 = !DILexicalBlockFile(discriminator: 2, file: !1, scope: !7) 159!16 = !DILocation(line: 10, scope: !17) 160!17 = distinct !DILexicalBlock(line: 10, column: 0, file: !1, scope: !7) 161!18 = !DILocation(line: 10, scope: !19) 162!19 = !DILexicalBlockFile(discriminator: 2, file: !1, scope: !17) 163!20 = !DILocation(line: 10, scope: !21) 164!21 = !DILexicalBlockFile(discriminator: 4, file: !1, scope: !17) 165!22 = !DILocation(line: 10, scope: !23) 166!23 = !DILexicalBlockFile(discriminator: 6, file: !1, scope: !17) 167!24 = !DILocation(line: 11, scope: !7) 168!25 = !DILocation(line: 12, scope: !7) 169