Refine instruction weight annotation algorithm for sample profiler.

[danielcdh]

Summary

This patch refined the instruction weight anootation algorithm:

1. Do not use dbg_value intrinsics for annotation.
2. Annotate cold calls if the call is inlined in profile, but not inlined before preparation. This indicates that the annotation preparation step found no sample for the inlined callsite, thus the call should be very cold.

[mgrang]

lib/Transforms/IPO/SampleProfile.cpp

460

Extraneous braces.

491

Extraneous braces.

[danielcdh]

formatting

[davidxl]

lib/Transforms/IPO/SampleProfile.cpp

489

move R=0 to newline.

489

Since this is the else branch which means R already is zero, why do we need to reset it?

[danielcdh]

lib/Transforms/IPO/SampleProfile.cpp

489

R is ErrorOr<uint64_t>, so else branch means that R is undefined instead of 0.

[danielcdh]

format

[davidxl]

lgtm

lib/Transforms/IPO/SampleProfile.cpp

 #include "llvm/Analysis/LoopInfo.h"
 #include "llvm/Analysis/PostDominators.h"
 #include "llvm/IR/Constants.h"
 #include "llvm/IR/DebugInfo.h"
 #include "llvm/IR/DiagnosticInfo.h"
 #include "llvm/IR/Dominators.h"
 #include "llvm/IR/Function.h"
 #include "llvm/IR/InstIterator.h"
+#include "llvm/IR/IntrinsicInst.h"
 #include "llvm/IR/Instructions.h"
 #include "llvm/IR/LLVMContext.h"
 #include "llvm/IR/MDBuilder.h"
 #include "llvm/IR/Metadata.h"
 #include "llvm/IR/Module.h"
 #include "llvm/Pass.h"
 #include "llvm/ProfileData/SampleProfReader.h"
 #include "llvm/Support/CommandLine.h"
SampleProfileLoader::getInstWeight(const Instruction &Inst) const {
   DebugLoc DLoc = Inst.getDebugLoc();
   if (!DLoc)
     return std::error_code();
 
   const FunctionSamples *FS = findFunctionSamples(Inst);
   if (!FS)
     return std::error_code();
 
+  // Ignore all dbg_value intrinsics.
+  const IntrinsicInst *II = dyn_cast<IntrinsicInst>(&Inst);
+  if (II && II->getIntrinsicID() == Intrinsic::dbg_value) {
+    return std::error_code();
+
   const DILocation *DIL = DLoc;
   unsigned Lineno = DLoc.getLine();
   unsigned HeaderLineno = DIL->getScope()->getSubprogram()->getLine();
 
   uint32_t LineOffset = getOffset(Lineno, HeaderLineno);
   uint32_t Discriminator = DIL->getDiscriminator();
   ErrorOr<uint64_t> R = FS->findSamplesAt(LineOffset, Discriminator);
   if (R) {
     bool FirstMark =
         CoverageTracker.markSamplesUsed(FS, LineOffset, Discriminator, R.get());
     if (FirstMark) {
       const Function *F = Inst.getParent()->getParent();
       LLVMContext &Ctx = F->getContext();
       emitOptimizationRemark(
           Ctx, DEBUG_TYPE, *F, DLoc,
           Twine("Applied ") + Twine(*R) + " samples from profile (offset: " +
               Twine(LineOffset) +
               ((Discriminator) ? Twine(".") + Twine(Discriminator) : "") + ")");
     }
     DEBUG(dbgs() << "    " << Lineno << "." << DIL->getDiscriminator() << ":"
                  << Inst << " (line offset: " << Lineno - HeaderLineno << "."
                  << DIL->getDiscriminator() << " - weight: " << R.get()
                  << ")\n");
+  } else {
+    // If a call instruction is inlined in profile, but not inlined here,
+    // it means that the inlined callsite has no sample, thus the call
+    // instruction should have 0 count.
+    const CallInst *CI = dyn_cast<CallInst>(&Inst);
+    if (CI && findCalleeFunctionSamples(*CI))
+      R = 0;
   }
   return R;
 }
 
 /// \brief Compute the weight of a basic block.
 ///
 /// The weight of basic block \p BB is the maximum weight of all the
 /// instructions in BB.
 ///