1; RUN: opt < %s -sample-profile -sample-profile-file=%S/Inputs/calls.prof | opt -analyze -branch-prob | FileCheck %s 2 3; Original C++ test case 4; 5; #include <stdio.h> 6; 7; int sum(int x, int y) { 8; return x + y; 9; } 10; 11; int main() { 12; int s, i = 0; 13; while (i++ < 20000 * 20000) 14; if (i != 100) s = sum(i, s); else s = 30; 15; printf("sum is %d\n", s); 16; return 0; 17; } 18; 19; Note that this test is missing the llvm.dbg.cu annotation. This emulates 20; the effect of the user having only used -fprofile-sample-use without 21; -gmlt when invoking the driver. In those cases, we need to track source 22; location information but we do not have to generate debug info in the 23; final binary. 24@.str = private unnamed_addr constant [11 x i8] c"sum is %d\0A\00", align 1 25 26; Function Attrs: nounwind uwtable 27define i32 @_Z3sumii(i32 %x, i32 %y) { 28entry: 29 %x.addr = alloca i32, align 4 30 %y.addr = alloca i32, align 4 31 store i32 %x, i32* %x.addr, align 4 32 store i32 %y, i32* %y.addr, align 4 33 %0 = load i32* %x.addr, align 4, !dbg !11 34 %1 = load i32* %y.addr, align 4, !dbg !11 35 %add = add nsw i32 %0, %1, !dbg !11 36 ret i32 %add, !dbg !11 37} 38 39; Function Attrs: uwtable 40define i32 @main() { 41entry: 42 %retval = alloca i32, align 4 43 %s = alloca i32, align 4 44 %i = alloca i32, align 4 45 store i32 0, i32* %retval 46 store i32 0, i32* %i, align 4, !dbg !12 47 br label %while.cond, !dbg !13 48 49while.cond: ; preds = %if.end, %entry 50 %0 = load i32* %i, align 4, !dbg !14 51 %inc = add nsw i32 %0, 1, !dbg !14 52 store i32 %inc, i32* %i, align 4, !dbg !14 53 %cmp = icmp slt i32 %0, 400000000, !dbg !14 54 br i1 %cmp, label %while.body, label %while.end, !dbg !14 55; CHECK: edge while.cond -> while.body probability is 5391 / 5392 = 99.9815% [HOT edge] 56; CHECK: edge while.cond -> while.end probability is 1 / 5392 = 0.018546% 57 58while.body: ; preds = %while.cond 59 %1 = load i32* %i, align 4, !dbg !16 60 %cmp1 = icmp ne i32 %1, 100, !dbg !16 61 br i1 %cmp1, label %if.then, label %if.else, !dbg !16 62; Without discriminator information, the profiler used to think that 63; both branches out of while.body had the same weight. In reality, 64; the edge while.body->if.then is taken most of the time. 65; 66; CHECK: edge while.body -> if.then probability is 5752 / 5753 = 99.9826% [HOT edge] 67; CHECK: edge while.body -> if.else probability is 1 / 5753 = 0.0173822% 68 69 70if.then: ; preds = %while.body 71 %2 = load i32* %i, align 4, !dbg !18 72 %3 = load i32* %s, align 4, !dbg !18 73 %call = call i32 @_Z3sumii(i32 %2, i32 %3), !dbg !18 74 store i32 %call, i32* %s, align 4, !dbg !18 75 br label %if.end, !dbg !18 76 77if.else: ; preds = %while.body 78 store i32 30, i32* %s, align 4, !dbg !20 79 br label %if.end 80 81if.end: ; preds = %if.else, %if.then 82 br label %while.cond, !dbg !22 83 84while.end: ; preds = %while.cond 85 %4 = load i32* %s, align 4, !dbg !24 86 %call2 = call i32 (i8*, ...)* @printf(i8* getelementptr inbounds ([11 x i8]* @.str, i32 0, i32 0), i32 %4), !dbg !24 87 ret i32 0, !dbg !25 88} 89 90declare i32 @printf(i8*, ...) #2 91 92!llvm.module.flags = !{!8, !9} 93!llvm.ident = !{!10} 94 95!0 = !{!"0x11\004\00clang version 3.5 \000\00\000\00\000", !1, !2, !2, !3, !2, !2} ; [ DW_TAG_compile_unit ] [./calls.cc] [DW_LANG_C_plus_plus] 96!1 = !{!"calls.cc", !"."} 97!2 = !{} 98!3 = !{!4, !7} 99!4 = !{!"0x2e\00sum\00sum\00\003\000\001\000\006\00256\000\003", !1, !5, !6, null, i32 (i32, i32)* @_Z3sumii, null, null, !2} ; [ DW_TAG_subprogram ] [line 3] [def] [sum] 100!5 = !{!"0x29", !1} ; [ DW_TAG_file_type ] [./calls.cc] 101!6 = !{!"0x15\00\000\000\000\000\000\000", i32 0, null, null, !2, null, null, null} ; [ DW_TAG_subroutine_type ] [line 0, size 0, align 0, offset 0] [from ] 102!7 = !{!"0x2e\00main\00main\00\007\000\001\000\006\00256\000\007", !1, !5, !6, null, i32 ()* @main, null, null, !2} ; [ DW_TAG_subprogram ] [line 7] [def] [main] 103!8 = !{i32 2, !"Dwarf Version", i32 4} 104!9 = !{i32 1, !"Debug Info Version", i32 2} 105!10 = !{!"clang version 3.5 "} 106!11 = !MDLocation(line: 4, scope: !4) 107!12 = !MDLocation(line: 8, scope: !7) 108!13 = !MDLocation(line: 9, scope: !7) 109!14 = !MDLocation(line: 9, scope: !15) 110!15 = !{!"0xb\001", !1, !7} ; [ DW_TAG_lexical_block ] [./calls.cc] 111!16 = !MDLocation(line: 10, scope: !17) 112!17 = !{!"0xb\0010\000\000", !1, !7} ; [ DW_TAG_lexical_block ] [./calls.cc] 113!18 = !MDLocation(line: 10, scope: !19) 114!19 = !{!"0xb\001", !1, !17} ; [ DW_TAG_lexical_block ] [./calls.cc] 115!20 = !MDLocation(line: 10, scope: !21) 116!21 = !{!"0xb\002", !1, !17} ; [ DW_TAG_lexical_block ] [./calls.cc] 117!22 = !MDLocation(line: 10, scope: !23) 118!23 = !{!"0xb\003", !1, !17} ; [ DW_TAG_lexical_block ] [./calls.cc] 119!24 = !MDLocation(line: 11, scope: !7) 120!25 = !MDLocation(line: 12, scope: !7) 121