Tolerate negative offset when matching sample profile.

In some cases (as illustrated in the unittest), lineno can be less than the heade_lineno because the function body are included from some other files. In this case, offset will be negative. This patch makes clang still able to match the profile to IR in this situation.

http://reviews.llvm.org/D13914

llvm-svn: 250873

4 files changed, 117 insertions, 10 deletions

diff --git a/llvm/lib/ProfileData/SampleProfReader.cpp b/llvm/lib/ProfileData/SampleProfReader.cpp
index d184969e046..899343f72f7 100644
--- a/llvm/lib/ProfileData/SampleProfReader.cpp
+++ b/llvm/lib/ProfileData/SampleProfReader.cpp
@@ -100,6 +100,12 @@ static bool ParseHead(const StringRef &Input, StringRef &FName,
   return true;
 }
 
+
+/// \brief Returns true if line offset \p L is legal (only has 16 bits).
+static bool isOffsetLegal(unsigned L) {
+  return (L & 0xffff) == L;
+}
+
 /// \brief Parse \p Input as line sample.
 ///
 /// \param Input input line.
@@ -124,7 +130,7 @@ static bool ParseLine(const StringRef &Input, bool &IsCallsite, uint32_t &Depth,
   StringRef Loc = Input.substr(Depth, n1 - Depth);
   size_t n2 = Loc.find('.');
   if (n2 == StringRef::npos) {
-    if (Loc.getAsInteger(10, LineOffset))
+    if (Loc.getAsInteger(10, LineOffset) || !isOffsetLegal(LineOffset))
       return false;
     Discriminator = 0;
   } else {
@@ -308,6 +314,10 @@ SampleProfileReaderBinary::readProfile(FunctionSamples &FProfile) {
     if (std::error_code EC = LineOffset.getError())
       return EC;
 
+    if (!isOffsetLegal(*LineOffset)) {
+      return std::error_code();
+    }
+
     auto Discriminator = readNumber<uint64_t>();
     if (std::error_code EC = Discriminator.getError())
       return EC;
diff --git a/llvm/lib/Transforms/IPO/SampleProfile.cpp b/llvm/lib/Transforms/IPO/SampleProfile.cpp
index e0691515c95..b797321b2f8 100644
--- a/llvm/lib/Transforms/IPO/SampleProfile.cpp
+++ b/llvm/lib/Transforms/IPO/SampleProfile.cpp
@@ -122,6 +122,7 @@ protected:
   void buildEdges(Function &F);
   bool propagateThroughEdges(Function &F);
   void computeDominanceAndLoopInfo(Function &F);
+  unsigned getOffset(unsigned L, unsigned H) const;
 
   /// \brief Map basic blocks to their computed weights.
   ///
@@ -174,6 +175,17 @@ protected:
 };
 }
 
+/// \brief Returns the offset of lineno \p L to head_lineno \p H
+///
+/// \param L  Lineno
+/// \param H  Header lineno of the function
+///
+/// \returns offset to the header lineno. 16 bits are used to represent offset.
+/// We assume that a single function will not exceed 65535 LOC.
+unsigned SampleProfileLoader::getOffset(unsigned L, unsigned H) const {
+  return (L - H) & 0xffff;
+}
+
 /// \brief Print the weight of edge \p E on stream \p OS.
 ///
 /// \param OS  Stream to emit the output to.
@@ -229,11 +241,9 @@ SampleProfileLoader::getInstWeight(const Instruction &Inst) const {
   const DILocation *DIL = DLoc;
   unsigned Lineno = DLoc.getLine();
   unsigned HeaderLineno = DIL->getScope()->getSubprogram()->getLine();
-  if (Lineno < HeaderLineno)
-    return std::error_code();
 
-  ErrorOr<uint64_t> R =
-      FS->findSamplesAt(Lineno - HeaderLineno, DIL->getDiscriminator());
+  ErrorOr<uint64_t> R = FS->findSamplesAt(getOffset(Lineno, HeaderLineno),
+                                          DIL->getDiscriminator());
   if (R)
     DEBUG(dbgs() << "    " << Lineno << "." << DIL->getDiscriminator() << ":"
                  << Inst << " (line offset: " << Lineno - HeaderLineno << "."
@@ -308,7 +318,7 @@ SampleProfileLoader::findCalleeFunctionSamples(const CallInst &Inst) const {
     return nullptr;
   }
   DISubprogram *SP = DIL->getScope()->getSubprogram();
-  if (!SP || DIL->getLine() < SP->getLine())
+  if (!SP)
     return nullptr;
 
   Function *CalleeFunc = Inst.getCalledFunction();
@@ -321,8 +331,9 @@ SampleProfileLoader::findCalleeFunctionSamples(const CallInst &Inst) const {
   if (FS == nullptr)
     return nullptr;
 
-  return FS->findFunctionSamplesAt(CallsiteLocation(
-      DIL->getLine() - SP->getLine(), DIL->getDiscriminator(), CalleeName));
+  return FS->findFunctionSamplesAt(
+      CallsiteLocation(getOffset(DIL->getLine(), SP->getLine()),
+                       DIL->getDiscriminator(), CalleeName));
 }
 
 /// \brief Get the FunctionSamples for an instruction.
@@ -345,10 +356,10 @@ SampleProfileLoader::findFunctionSamples(const Instruction &Inst) const {
   for (const DILocation *DIL = Inst.getDebugLoc(); DIL;
        DIL = DIL->getInlinedAt()) {
     DISubprogram *SP = DIL->getScope()->getSubprogram();
-    if (!SP || DIL->getLine() < SP->getLine())
+    if (!SP)
       return nullptr;
     if (!CalleeName.empty()) {
-      S.push_back(CallsiteLocation(DIL->getLine() - SP->getLine(),
+      S.push_back(CallsiteLocation(getOffset(DIL->getLine(), SP->getLine()),
                                    DIL->getDiscriminator(), CalleeName));
     }
     CalleeName = SP->getLinkageName();
diff --git a/llvm/test/Transforms/SampleProfile/Inputs/offset.prof b/llvm/test/Transforms/SampleProfile/Inputs/offset.prof
new file mode 100644
index 00000000000..b07ce3504fb
--- /dev/null
+++ b/llvm/test/Transforms/SampleProfile/Inputs/offset.prof
@@ -0,0 +1,4 @@
+_Z3fooi:300:1
+ 65532: 1000
+ 65533: 10
+ 65535: 990
diff --git a/llvm/test/Transforms/SampleProfile/offset.ll b/llvm/test/Transforms/SampleProfile/offset.ll
new file mode 100644
index 00000000000..151fc0941ed
--- /dev/null
+++ b/llvm/test/Transforms/SampleProfile/offset.ll
@@ -0,0 +1,82 @@
+; RUN: opt < %s -sample-profile -sample-profile-file=%S/Inputs/offset.prof | opt -analyze -branch-prob | FileCheck %s
+
+; Original C++ code for this test case:
+;
+; a.cc:
+; #1
+; #2
+; #3
+; #4
+; #5 int foo(int a) {
+; #6 #include "a.b"
+; #7}
+;
+; a.b:
+; #1 if (a > 0) {
+; #2   return 10;
+; #3 } else {
+; #4   return 20;
+; #5 }
+
+; Function Attrs: nounwind uwtable
+define i32 @_Z3fooi(i32 %a) #0 {
+entry:
+  %retval = alloca i32, align 4
+  %a.addr = alloca i32, align 4
+  store i32 %a, i32* %a.addr, align 4
+  call void @llvm.dbg.declare(metadata i32* %a.addr, metadata !11, metadata !12), !dbg !13
+  %0 = load i32, i32* %a.addr, align 4, !dbg !14
+  %cmp = icmp sgt i32 %0, 0, !dbg !18
+  br i1 %cmp, label %if.then, label %if.else, !dbg !19
+; CHECK: edge entry -> if.then probability is 0x0147ae14 / 0x80000000 = 1.00%
+; CHECK: edge entry -> if.else probability is 0x7eb851ec / 0x80000000 = 99.00% [HOT edge]
+
+if.then:                                          ; preds = %entry
+  store i32 10, i32* %retval, align 4, !dbg !20
+  br label %return, !dbg !20
+
+if.else:                                          ; preds = %entry
+  store i32 20, i32* %retval, align 4, !dbg !22
+  br label %return, !dbg !22
+
+return:                                           ; preds = %if.else, %if.then
+  %1 = load i32, i32* %retval, align 4, !dbg !24
+  ret i32 %1, !dbg !24
+}
+
+; Function Attrs: nounwind readnone
+declare void @llvm.dbg.declare(metadata, metadata, metadata) #1
+
+attributes #0 = { nounwind uwtable "disable-tail-calls"="false" "less-precise-fpmad"="false" "no-frame-pointer-elim"="true" "no-frame-pointer-elim-non-leaf" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "target-cpu"="x86-64" "target-features"="+fxsr,+mmx,+sse,+sse2" "unsafe-fp-math"="false" "use-soft-float"="false" }
+attributes #1 = { nounwind readnone }
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!8, !9}
+!llvm.ident = !{!10}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C_plus_plus, file: !1, producer: "clang version 3.8.0 (trunk 250750)", isOptimized: false, runtimeVersion: 0, emissionKind: 1, enums: !2, subprograms: !3)
+!1 = !DIFile(filename: "a.cc", directory: "/tmp")
+!2 = !{}
+!3 = !{!4}
+!4 = distinct !DISubprogram(name: "foo", linkageName: "_Z3fooi", scope: !1, file: !1, line: 5, type: !5, isLocal: false, isDefinition: true, scopeLine: 5, flags: DIFlagPrototyped, isOptimized: false, function: i32 (i32)* @_Z3fooi, variables: !2)
+!5 = !DISubroutineType(types: !6)
+!6 = !{!7, !7}
+!7 = !DIBasicType(name: "int", size: 32, align: 32, encoding: DW_ATE_signed)
+!8 = !{i32 2, !"Dwarf Version", i32 4}
+!9 = !{i32 2, !"Debug Info Version", i32 3}
+!10 = !{!"clang version 3.8.0 (trunk 250750)"}
+!11 = !DILocalVariable(name: "a", arg: 1, scope: !4, file: !1, line: 5, type: !7)
+!12 = !DIExpression()
+!13 = !DILocation(line: 5, column: 13, scope: !4)
+!14 = !DILocation(line: 1, column: 5, scope: !15)
+!15 = distinct !DILexicalBlock(scope: !17, file: !16, line: 1, column: 5)
+!16 = !DIFile(filename: "./a.b", directory: "/tmp")
+!17 = !DILexicalBlockFile(scope: !4, file: !16, discriminator: 0)
+!18 = !DILocation(line: 1, column: 7, scope: !15)
+!19 = !DILocation(line: 1, column: 5, scope: !17)
+!20 = !DILocation(line: 2, column: 3, scope: !21)
+!21 = distinct !DILexicalBlock(scope: !15, file: !16, line: 1, column: 12)
+!22 = !DILocation(line: 4, column: 3, scope: !23)
+!23 = distinct !DILexicalBlock(scope: !15, file: !16, line: 3, column: 8)
+!24 = !DILocation(line: 7, column: 1, scope: !25)
+!25 = !DILexicalBlockFile(scope: !4, file: !1, discriminator: 0)