[llvm][IR] Extend BranchWeightMetadata to track provenance of weights (#86609)

This patch implements the changes to LLVM IR discussed in

https://discourse.llvm.org/t/rfc-update-branch-weights-metadata-to-allow-tracking-branch-weight-origins/75032

In this patch, we add an optional field to MD_prof metadata nodes for
branch weights, which can be used to distinguish weights added from
`llvm.expect*` intrinsics from those added via other methods, e.g.
from profiles or inserted by the compiler.

One of the major motivations, is for use with MisExpect diagnostics,
which need to know if branch_weight metadata originates from an
llvm.expect intrinsic. Without that information, we end up checking
branch weights multiple times in the case if ThinLTO + SampleProfiling,
leading to some inaccuracy in how we report MisExpect related
diagnostics to users.

Since we change the format of MD_prof metadata in a fundamental way, we
need to update code handling branch weights in a number of places.

We also update the lang ref for branch weights to reflect the change.

1 files changed, 13 insertions, 10 deletions

diff --git a/llvm/lib/Transforms/Utils/SimplifyCFG.cpp b/llvm/lib/Transforms/Utils/SimplifyCFG.cpp
index 292739b..107c8bb 100644
--- a/llvm/lib/Transforms/Utils/SimplifyCFG.cpp
+++ b/llvm/lib/Transforms/Utils/SimplifyCFG.cpp
@@ -861,26 +861,28 @@ static bool ValuesOverlap(std::vector<ValueEqualityComparisonCase> &C1,
 
 // Set branch weights on SwitchInst. This sets the metadata if there is at
 // least one non-zero weight.
-static void setBranchWeights(SwitchInst *SI, ArrayRef<uint32_t> Weights) {
+static void setBranchWeights(SwitchInst *SI, ArrayRef<uint32_t> Weights,
+                             bool IsExpected) {
   // Check that there is at least one non-zero weight. Otherwise, pass
   // nullptr to setMetadata which will erase the existing metadata.
   MDNode *N = nullptr;
   if (llvm::any_of(Weights, [](uint32_t W) { return W != 0; }))
-    N = MDBuilder(SI->getParent()->getContext()).createBranchWeights(Weights);
+    N = MDBuilder(SI->getParent()->getContext())
+            .createBranchWeights(Weights, IsExpected);
   SI->setMetadata(LLVMContext::MD_prof, N);
 }
 
 // Similar to the above, but for branch and select instructions that take
 // exactly 2 weights.
 static void setBranchWeights(Instruction *I, uint32_t TrueWeight,
-                             uint32_t FalseWeight) {
+                             uint32_t FalseWeight, bool IsExpected) {
   assert(isa<BranchInst>(I) || isa<SelectInst>(I));
   // Check that there is at least one non-zero weight. Otherwise, pass
   // nullptr to setMetadata which will erase the existing metadata.
   MDNode *N = nullptr;
   if (TrueWeight || FalseWeight)
     N = MDBuilder(I->getParent()->getContext())
-            .createBranchWeights(TrueWeight, FalseWeight);
+            .createBranchWeights(TrueWeight, FalseWeight, IsExpected);
   I->setMetadata(LLVMContext::MD_prof, N);
 }
 
@@ -1338,7 +1340,7 @@ bool SimplifyCFGOpt::PerformValueComparisonIntoPredecessorFolding(
 
     SmallVector<uint32_t, 8> MDWeights(Weights.begin(), Weights.end());
 
-    setBranchWeights(NewSI, MDWeights);
+    setBranchWeights(NewSI, MDWeights, /*IsExpected=*/false);
   }
 
   EraseTerminatorAndDCECond(PTI);
@@ -3831,7 +3833,7 @@ static bool performBranchToCommonDestFolding(BranchInst *BI, BranchInst *PBI,
     FitWeights(NewWeights);
 
     SmallVector<uint32_t, 8> MDWeights(NewWeights.begin(), NewWeights.end());
-    setBranchWeights(PBI, MDWeights[0], MDWeights[1]);
+    setBranchWeights(PBI, MDWeights[0], MDWeights[1], /*IsExpected=*/false);
 
     // TODO: If BB is reachable from all paths through PredBlock, then we
     // could replace PBI's branch probabilities with BI's.
@@ -4568,7 +4570,7 @@ static bool SimplifyCondBranchToCondBranch(BranchInst *PBI, BranchInst *BI,
     // Halve the weights if any of them cannot fit in an uint32_t
     FitWeights(NewWeights);
 
-    setBranchWeights(PBI, NewWeights[0], NewWeights[1]);
+    setBranchWeights(PBI, NewWeights[0], NewWeights[1], /*IsExpected=*/false);
   }
 
   // OtherDest may have phi nodes.  If so, add an entry from PBI's
@@ -4604,7 +4606,8 @@ static bool SimplifyCondBranchToCondBranch(BranchInst *PBI, BranchInst *BI,
 
         FitWeights(NewWeights);
 
-        setBranchWeights(NV, NewWeights[0], NewWeights[1]);
+        setBranchWeights(NV, NewWeights[0], NewWeights[1],
+                         /*IsExpected=*/false);
       }
     }
   }
@@ -4667,7 +4670,7 @@ bool SimplifyCFGOpt::SimplifyTerminatorOnSelect(Instruction *OldTerm,
       // Create a conditional branch sharing the condition of the select.
       BranchInst *NewBI = Builder.CreateCondBr(Cond, TrueBB, FalseBB);
       if (TrueWeight != FalseWeight)
-        setBranchWeights(NewBI, TrueWeight, FalseWeight);
+        setBranchWeights(NewBI, TrueWeight, FalseWeight, /*IsExpected=*/false);
     }
   } else if (KeepEdge1 && (KeepEdge2 || TrueBB == FalseBB)) {
     // Neither of the selected blocks were successors, so this
@@ -5617,7 +5620,7 @@ bool SimplifyCFGOpt::TurnSwitchRangeIntoICmp(SwitchInst *SI,
         TrueWeight /= 2;
         FalseWeight /= 2;
       }
-      setBranchWeights(NewBI, TrueWeight, FalseWeight);
+      setBranchWeights(NewBI, TrueWeight, FalseWeight, /*IsExpected=*/false);
     }
   }