[llvm-branch-commits] Using matched block counts to measure discrepancy (PR #95486)
via llvm-branch-commits
llvm-branch-commits at lists.llvm.org
Thu Jun 13 16:51:50 PDT 2024
llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-llvm-transforms
Author: shaw young (shawbyoung)
<details>
<summary>Changes</summary>
Test Plan: tbd
---
Full diff: https://github.com/llvm/llvm-project/pull/95486.diff
2 Files Affected:
- (modified) bolt/lib/Profile/StaleProfileMatching.cpp (+29-8)
- (modified) llvm/include/llvm/Transforms/Utils/SampleProfileInference.h (-2)
``````````diff
diff --git a/bolt/lib/Profile/StaleProfileMatching.cpp b/bolt/lib/Profile/StaleProfileMatching.cpp
index 6588cf2c0ce66..cbd98f4d4769f 100644
--- a/bolt/lib/Profile/StaleProfileMatching.cpp
+++ b/bolt/lib/Profile/StaleProfileMatching.cpp
@@ -53,9 +53,9 @@ cl::opt<bool>
cl::opt<unsigned> MatchedProfileThreshold(
"matched-profile-threshold",
- cl::desc("Percentage threshold of matched execution counts at which stale "
+ cl::desc("Percentage threshold of matched basic blocks at which stale "
"profile inference is executed."),
- cl::init(5), cl::Hidden, cl::cat(BoltOptCategory));
+ cl::init(50), cl::Hidden, cl::cat(BoltOptCategory));
cl::opt<unsigned> StaleMatchingMaxFuncSize(
"stale-matching-max-func-size",
@@ -186,6 +186,17 @@ struct BlendedBlockHash {
uint8_t SuccHash{0};
};
+/// A data object containing function matching information.
+struct FunctionMatchingData {
+public:
+ /// The number of blocks matched exactly.
+ uint64_t MatchedExactBlocks{0};
+ /// The number of blocks matched loosely.
+ uint64_t MatchedLooseBlocks{0};
+ /// The number of execution counts matched.
+ uint64_t MatchedExecCounts{0};
+};
+
/// The object is used to identify and match basic blocks in a BinaryFunction
/// given their hashes computed on a binary built from several revisions behind
/// release.
@@ -400,7 +411,8 @@ createFlowFunction(const BinaryFunction::BasicBlockOrderType &BlockOrder) {
void matchWeightsByHashes(BinaryContext &BC,
const BinaryFunction::BasicBlockOrderType &BlockOrder,
const yaml::bolt::BinaryFunctionProfile &YamlBF,
- FlowFunction &Func) {
+ FlowFunction &Func,
+ FunctionMatchingData &FuncMatchingData) {
assert(Func.Blocks.size() == BlockOrder.size() + 1);
std::vector<FlowBlock *> Blocks;
@@ -440,9 +452,11 @@ void matchWeightsByHashes(BinaryContext &BC,
if (Matcher.isHighConfidenceMatch(BinHash, YamlHash)) {
++BC.Stats.NumMatchedBlocks;
BC.Stats.MatchedSampleCount += YamlBB.ExecCount;
- Func.MatchedExecCount += YamlBB.ExecCount;
+ FuncMatchingData.MatchedExecCounts += YamlBB.ExecCount;
+ FuncMatchingData.MatchedExactBlocks += 1;
LLVM_DEBUG(dbgs() << " exact match\n");
} else {
+ FuncMatchingData.MatchedLooseBlocks += 1;
LLVM_DEBUG(dbgs() << " loose match\n");
}
if (YamlBB.NumInstructions == BB->size())
@@ -582,11 +596,14 @@ void preprocessUnreachableBlocks(FlowFunction &Func) {
/// Decide if stale profile matching can be applied for a given function.
/// Currently we skip inference for (very) large instances and for instances
/// having "unexpected" control flow (e.g., having no sink basic blocks).
-bool canApplyInference(const FlowFunction &Func, const yaml::bolt::BinaryFunctionProfile &YamlBF) {
+bool canApplyInference(const FlowFunction &Func,
+ const yaml::bolt::BinaryFunctionProfile &YamlBF,
+ const FunctionMatchingData &FuncMatchingData) {
if (Func.Blocks.size() > opts::StaleMatchingMaxFuncSize)
return false;
- if (Func.MatchedExecCount / YamlBF.ExecCount >= opts::MatchedProfileThreshold)
+ if ((double)FuncMatchingData.MatchedExactBlocks / YamlBF.Blocks.size() >=
+ opts::MatchedProfileThreshold / 100.0)
return false;
bool HasExitBlocks = llvm::any_of(
@@ -735,18 +752,22 @@ bool YAMLProfileReader::inferStaleProfile(
const BinaryFunction::BasicBlockOrderType BlockOrder(
BF.getLayout().block_begin(), BF.getLayout().block_end());
+ // Create a containter for function matching data.
+ FunctionMatchingData FuncMatchingData;
+
// Create a wrapper flow function to use with the profile inference algorithm.
FlowFunction Func = createFlowFunction(BlockOrder);
// Match as many block/jump counts from the stale profile as possible
- matchWeightsByHashes(BF.getBinaryContext(), BlockOrder, YamlBF, Func);
+ matchWeightsByHashes(BF.getBinaryContext(), BlockOrder, YamlBF, Func,
+ FuncMatchingData);
// Adjust the flow function by marking unreachable blocks Unlikely so that
// they don't get any counts assigned.
preprocessUnreachableBlocks(Func);
// Check if profile inference can be applied for the instance.
- if (!canApplyInference(Func, YamlBF))
+ if (!canApplyInference(Func, YamlBF, FuncMatchingData))
return false;
// Apply the profile inference algorithm.
diff --git a/llvm/include/llvm/Transforms/Utils/SampleProfileInference.h b/llvm/include/llvm/Transforms/Utils/SampleProfileInference.h
index e7971ca1cb428..b4ea1ad840f9d 100644
--- a/llvm/include/llvm/Transforms/Utils/SampleProfileInference.h
+++ b/llvm/include/llvm/Transforms/Utils/SampleProfileInference.h
@@ -57,8 +57,6 @@ struct FlowFunction {
std::vector<FlowJump> Jumps;
/// The index of the entry block.
uint64_t Entry{0};
- // Matched execution count for the function.
- uint64_t MatchedExecCount{0};
};
/// Various thresholds and options controlling the behavior of the profile
``````````
</details>
https://github.com/llvm/llvm-project/pull/95486
More information about the llvm-branch-commits
mailing list