[llvm] [PGO] Sampled instrumentation in PGO to speed up instrumentation binary (PR #69535)

[via llvm-commits]

================
@@ -412,30 +424,91 @@ PreservedAnalyses InstrProfiling::run(Module &M, ModuleAnalysisManager &AM) {
   return PreservedAnalyses::none();
 }
 
+// Perform instrumentation sampling.
+// We transform:
+//   Increment_Instruction;
+// to:
+//   if (__llvm_profile_sampling__ <= SampleDuration) {
+//     Increment_Instruction;
+//   }
+//   __llvm_profile_sampling__ += 1;
+//
+// "__llvm_profile_sampling__" is a thread-local global shared by all PGO
+// instrumentation variables (value-instrumentation and edge instrumentation).
+// It has a unsigned short type and will wrapper around when overflow.
+//
+// Note that, the code snippet after the transformation can still be
+// counter promoted. But I don't see a reason for that because the
+// counter updated should be sparse. That's the reason we disable
+// counter promotion by default when sampling is enabled.
+// This can be overwritten by the internal option.
+//
+void InstrProfiling::doSampling(Instruction *I) {
+  if (!isSamplingEnabled())
+    return;
+  int SampleDuration = SampledInstrumentDuration.getValue();
+  unsigned WrapToZeroValue = USHRT_MAX + 1;
+  assert(SampleDuration < USHRT_MAX);
+  auto *Int16Ty = Type::getInt16Ty(M->getContext());
+  auto *CountVar =
+      M->getGlobalVariable(INSTR_PROF_QUOTE(INSTR_PROF_PROFILE_SAMPLING_VAR));
+  assert(CountVar && "CountVar not set properly");
+  IRBuilder<> CondBuilder(I);
+  auto *LoadCountVar = CondBuilder.CreateLoad(Int16Ty, CountVar);
+  auto *DurationCond = CondBuilder.CreateICmpULE(
+      LoadCountVar, CondBuilder.getInt16(SampleDuration));
+  MDBuilder MDB(I->getContext());
+  MDNode *BranchWeight =
+      MDB.createBranchWeights(SampleDuration, WrapToZeroValue - SampleDuration);
+  Instruction *ThenTerm = SplitBlockAndInsertIfThen(
+      DurationCond, I, /* Unreacheable */ false, BranchWeight);
+  IRBuilder<> IncBuilder(I);
+  auto *NewVal = IncBuilder.CreateAdd(LoadCountVar, IncBuilder.getInt16(1));
----------------
xur-llvm wrote:

Not sure I get the question: this depends on the wrap-around of unsigned add. I thought this is a c/c++ standard.

https://github.com/llvm/llvm-project/pull/69535