[SampleFDO] Stop letting findCalleeFunctionSamples return unrelated profiles for invoke instructions

[wmi]

Summary

We see a warning of "No debug information found in function foo: Function profile not used" in a case. The function foo is called by an invoke instruction. It has no debug information because it has attribute((nodebug)) in the definition. It shouldn't have profile instance in the sample profile but compiler thinks it does, that turns out to be a compiler bug in findCalleeFunctionSamples. The bug is exposed when sample-profile-merge-inlinee is enabled recently.

Currently in findCalleeFunctionSamples, CalleeName is unset and is empty for invoke instruction. For empty CalleeName, findFunctionSamplesAt will treat the call as an indirect call and will return any inline instance profile at the same location as the instruction. That leads to a wrong profile being returned to function foo.

The patch set CalleeName when the instruction is an invoke.

[wenlei]

llvm/lib/Transforms/IPO/SampleProfile.cpp

823

So the original differentiation between invoke and call here isn't intentional, but purely an implementation oversight?

[wmi]

llvm/lib/Transforms/IPO/SampleProfile.cpp

823

I think so. I don't see fundamental difference between call and invoke with regard to sampleFDO.

[wenlei]

Thanks. LGTM.

[hoyFB]

llvm/lib/Transforms/IPO/SampleProfile.cpp

819

Does it return here for invokes that don't have debug info?

The fix looks good to me. We should not treat non-call instructions are indirect calls.

[wmi]

llvm/lib/Transforms/IPO/SampleProfile.cpp

819

No it doesn't. The invoke instruction has debug information. It is the definition of the function being invoked doesn't have debug information.

[hoyFB]

llvm/lib/Transforms/IPO/SampleProfile.cpp

819

Oh I see. Thanks for the fix!

llvm/lib/Transforms/IPO/SampleProfile.cpp

 ///
 /// \param Inst Call/Invoke instruction to query.
 ///
 /// \returns The FunctionSamples pointer to the inlined instance.
 const FunctionSamples *
 SampleProfileLoader::findCalleeFunctionSamples(const CallBase &Inst) const {
   const DILocation *DIL = Inst.getDebugLoc();
   if (!DIL) {
     return nullptr;
   }
 
   StringRef CalleeName;
-  if (const CallInst *CI = dyn_cast<CallInst>(&Inst))
-    if (Function *Callee = CI->getCalledFunction())
-      CalleeName = Callee->getName();
+  if (Function *Callee = Inst.getCalledFunction())
 
   const FunctionSamples *FS = findFunctionSamples(Inst);
   if (FS == nullptr)
     return nullptr;
 
   return FS->findFunctionSamplesAt(LineLocation(FunctionSamples::getOffset(DIL),
                                                 DIL->getBaseDiscriminator()),
                                    CalleeName);

llvm/test/Transforms/SampleProfile/nodebug-error.ll

+; Test the profile _Z3sumii won't be mistakenly annotated to foo
+; when '-sample-profile-merge-inlinee' is enabled. When the mistake
+; happens, there will be a "No debug information found" warning message.
+; RUN: opt < %s -passes=sample-profile \
+; RUN:   -sample-profile-file=%S/Inputs/inline-mergeprof.prof \
+; RUN:   -sample-profile-merge-inlinee=true -S 2>&1| FileCheck %s
+
+@.str = private unnamed_addr constant [11 x i8] c"sum is %d\0A\00", align 1
+declare void @__cxa_call_unexpected(i8*)
+declare i32 @__gxx_personality_v0(...)
+declare i32 @_Z3subii(i32 %x, i32 %y)
+
+define i32 @main() "use-sample-profile" nounwind uwtable ssp personality i8* bitcast (i32 (...)* @__gxx_personality_v0 to i8*) !dbg !6 {
+entry:
+  %retval = alloca i32, align 4
+  %s = alloca i32, align 4
+  %i = alloca i32, align 4
+  %tmp = load i32, i32* %i, align 4, !dbg !8
+  %tmp1 = load i32, i32* %s, align 4, !dbg !8
+  %call = invoke i32 @foo(i32 %tmp, i32 %tmp1)
+          to label %cont unwind label %lpad, !dbg !8
+; CHECK-NOT: warning: No debug information found in function foo
+; CHECK: invoke i32 @foo
+cont:
+  store i32 %call, i32* %s, align 4, !dbg !8
+  ret i32 0, !dbg !11
+lpad:
+  %lptmp0 = landingpad { i8*, i32 }
+          filter [0 x i8*] zeroinitializer
+  %lptmp1 = extractvalue { i8*, i32 } %lptmp0, 0
+  tail call void @__cxa_call_unexpected(i8* %lptmp1) noreturn nounwind
+  unreachable
+}
+
+define i32 @foo(i32 %x, i32 %y) #0 {
+entry:
+  %add = add nsw i32 %x, %y
+  ret i32 %add
+}
+
+attributes #0 = { "use-sample-profile" }
+
+declare i32 @printf(i8*, ...)
+
+!llvm.dbg.cu = !{!0}
+!llvm.module.flags = !{!3, !4}
+!llvm.ident = !{!5}
+
+!0 = distinct !DICompileUnit(language: DW_LANG_C_plus_plus, file: !1, producer: "clang version 3.5 ", isOptimized: false, runtimeVersion: 0, emissionKind: NoDebug, enums: !2, retainedTypes: !2, globals: !2, imports: !2)
+!1 = !DIFile(filename: "calls.cc", directory: ".")
+!2 = !{}
+!3 = !{i32 2, !"Dwarf Version", i32 4}
+!4 = !{i32 1, !"Debug Info Version", i32 3}
+!5 = !{!"clang version 3.5 "}
+!6 = distinct !DISubprogram(name: "main", scope: !1, file: !1, line: 7, type: !7, scopeLine: 7, virtualIndex: 6, flags: DIFlagPrototyped, spFlags: DISPFlagDefinition, unit: !0, retainedNodes: !2)
+!7 = !DISubroutineType(types: !2)
+!8 = !DILocation(line: 10, scope: !9)
+!9 = !DILexicalBlockFile(scope: !10, file: !1, discriminator: 2)
+!10 = distinct !DILexicalBlock(scope: !6, file: !1, line: 10)
+!11 = !DILocation(line: 12, scope: !6)