Fix quadratic slowdown in AST matcher parent map generation #87824
Conversation
|
Thank you for submitting a Pull Request (PR) to the LLVM Project! This PR will be automatically labeled and the relevant teams will be If you wish to, you can add reviewers by using the "Reviewers" section on this page. If this is not working for you, it is probably because you do not have write If you have received no comments on your PR for a week, you can request a review If you have further questions, they may be answered by the LLVM GitHub User Guide. You can also ask questions in a comment on this PR, on the LLVM Discord or on the forums. |
llvmbot
added
clang
|
@llvm/pr-subscribers-clang Author: None (higher-performance) ChangesAvoids the need to linearly re-scan all seen parent nodes to check for duplicates, which previously caused a slowdown for ancestry checks in Clang AST matchers. Fixes: #86881 Full diff: https://github.com/llvm/llvm-project/pull/87824.diff 1 Files Affected:
diff --git a/clang/lib/AST/ParentMapContext.cpp b/clang/lib/AST/ParentMapContext.cpp
index 21cfd5b1de6e9d..3369df38754485 100644
--- a/clang/lib/AST/ParentMapContext.cpp
+++ b/clang/lib/AST/ParentMapContext.cpp
@@ -61,7 +61,28 @@ class ParentMapContext::ParentMap {
template <typename, typename...> friend struct ::MatchParents;
/// Contains parents of a node.
- using ParentVector = llvm::SmallVector<DynTypedNode, 2>;
+ class ParentVector {
+ public:
+ ParentVector() = default;
+ explicit ParentVector(size_t n, const DynTypedNode &value) {
+ Items.reserve(n);
+ for (; n > 0; --n) {
+ push_back(value);
+ }
+ }
+ bool contains(const DynTypedNode &value) {
+ return Seen.contains(value);
+ }
+ void push_back(const DynTypedNode &value) {
+ if (!value.getMemoizationData() || Seen.insert(value).second) {
+ Items.push_back(value);
+ }
+ }
+ llvm::ArrayRef<DynTypedNode> view() const { return Items; }
+ private:
+ llvm::SmallVector<DynTypedNode, 2> Items;
+ llvm::SmallDenseSet<DynTypedNode, 2> Seen;
+ };
/// Maps from a node to its parents. This is used for nodes that have
/// pointer identity only, which are more common and we can save space by
@@ -99,7 +120,7 @@ class ParentMapContext::ParentMap {
return llvm::ArrayRef<DynTypedNode>();
}
if (const auto *V = I->second.template dyn_cast<ParentVector *>()) {
- return llvm::ArrayRef(*V);
+ return V->view();
}
return getSingleDynTypedNodeFromParentMap(I->second);
}
@@ -252,7 +273,7 @@ class ParentMapContext::ParentMap {
const auto *S = It->second.dyn_cast<const Stmt *>();
if (!S) {
if (auto *Vec = It->second.dyn_cast<ParentVector *>())
- return llvm::ArrayRef(*Vec);
+ return Vec->view();
return getSingleDynTypedNodeFromParentMap(It->second);
}
const auto *P = dyn_cast<Expr>(S);
|
|
@shafik @AaronBallman if you could kindly review this I'd appreciate it! |
Avoids the need to linearly re-scan all seen parent nodes to check for duplicates, which previously caused a slowdown for ancestry checks in Clang AST matchers. Fixes #86881.
|
Added a release note as well; feel free to edit it if needed (or let me know if I should change anything)! |
|
Looks like the build passed as well :) |
|
@higher-performance Congratulations on having your first Pull Request (PR) merged into the LLVM Project! Your changes will be combined with recent changes from other authors, then tested Please check whether problems have been caused by your change specifically, as How to do this, and the rest of the post-merge process, is covered in detail here. If your change does cause a problem, it may be reverted, or you can revert it yourself. If you don't get any reports, no action is required from you. Your changes are working as expected, well done! |
|
Hello, |
…f nodes have been seen (#129934) This mitigates a regression introduced in #87824. The mitigation here is to store pointers the deduplicated AST nodes, rather than copies of the nodes themselves. This allows a pointer-optimized set to be used and saves a lot of memory because `clang::DynTypedNode` is ~5 times larger than a pointer. Fixes #129808.
…f nodes have been seen (llvm#129934) This mitigates a regression introduced in llvm#87824. The mitigation here is to store pointers the deduplicated AST nodes, rather than copies of the nodes themselves. This allows a pointer-optimized set to be used and saves a lot of memory because `clang::DynTypedNode` is ~5 times larger than a pointer. Fixes llvm#129808. (cherry picked from commit 8c7f0ea)
…f nodes have been seen (llvm#129934) This mitigates a regression introduced in llvm#87824. The mitigation here is to store pointers the deduplicated AST nodes, rather than copies of the nodes themselves. This allows a pointer-optimized set to be used and saves a lot of memory because `clang::DynTypedNode` is ~5 times larger than a pointer. Fixes llvm#129808.
Avoids the need to linearly re-scan all seen parent nodes to check for duplicates, which previously caused a slowdown for ancestry checks in Clang AST matchers.
Fixes: #86881