[InstrProf] Support conditional counter updates

This patch adds support of conditional counter updates
in single byte counters mode to reduce the write contention
by first checking whether the counter is set before
overwriting it.

Author: gulfemsavrun

llvm/lib/Transforms/Instrumentation/InstrProfiling.cpp

@@ -132,6 +132,11 @@ cl::opt<bool> AtomicFirstCounter(
              "the entry counter)"),
     cl::init(false));
 
+cl::opt<bool> ConditionalCounterUpdate(
+    "conditional-counter-update",
+    cl::desc("Do conditional counter updates in single byte counters mode)"),
+    cl::init(false));
+
 // If the option is not specified, the default behavior about whether
 // counter promotion is done depends on how instrumentaiton lowering
 // pipeline is setup, i.e., the default value of true of this option
@@ -1213,6 +1218,18 @@ Value *InstrLowerer::getBitmapAddress(InstrProfMCDCTVBitmapUpdate *I) {
 void InstrLowerer::lowerCover(InstrProfCoverInst *CoverInstruction) {
   auto *Addr = getCounterAddress(CoverInstruction);
   IRBuilder<> Builder(CoverInstruction);
+  if (ConditionalCounterUpdate) {
+    Instruction *SplitBefore = CoverInstruction->getNextNode();
+    auto &Ctx = CoverInstruction->getParent()->getContext();
+    auto *Int8Ty = llvm::Type::getInt8Ty(Ctx);
+    Value *Load = Builder.CreateLoad(Int8Ty, Addr, "pgocount");
+    Value *Cmp = Builder.CreateICmpNE(Load, ConstantInt::get(Int8Ty, 0),
+                                      "pgocount.ifnonzero");
+    Instruction *ThenBranch =
+        SplitBlockAndInsertIfThen(Cmp, SplitBefore, false);
+    Builder.SetInsertPoint(ThenBranch);
+  }
+
   // We store zero to represent that this block is covered.
   Builder.CreateStore(Builder.getInt8(0), Addr);
   CoverInstruction->eraseFromParent();

llvm/test/Instrumentation/InstrProfiling/conditional-counter-updates.ll

@@ -0,0 +1,17 @@
+; RUN: opt < %s -S -passes=instrprof -conditional-counter-update | FileCheck %s
+
+target triple = "x86_64-unknown-linux-gnu"
+
+@__profn_foo = private constant [3 x i8] c"foo"
+
+; CHECK-LABEL: define void @foo
+; CHECK-NEXT: %pgocount = load i8, ptr @__profc_foo, align 1
+; CHECK-NEXT: %pgocount.ifnonzero = icmp ne i8 %pgocount, 0
+; CHECK-NEXT: br i1 %pgocount.ifnonzero, label %1, label %2
+; CHECK: 1:                                                ; preds = %0
+; CHECK-NEXT:  store i8 0, ptr @__profc_foo, align 1
+; CHECK-NEXT:  br label %2
+define void @foo() {
+  call void @llvm.instrprof.cover(ptr @__profn_foo, i64 0, i32 1, i32 0)
+  ret void
+}