[Synth] Fix race condition and memory corruption in longest path analysis caching #9098
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
…ysis caching This commit addresses a race condition and memory corruption issue in the LongestPathAnalysis where concurrent access to cached results could lead to iterator invalidation and memory corruption. The issue occurred when the cachedResults DenseMap was modified during iteration, causing existing iterators and references to become invalid and point to corrupted memory. The fix changes the cachedResults map to store unique_ptr<SmallVector<OpenPath>> instead of SmallVector<OpenPath> directly. This ensures that the underlying SmallVector objects remain at stable memory locations even when the DenseMap is rehashed or modified, preventing iterator invalidation and memory corruption. Additionally, in markRegEndPoint, we now preemptively call getOrComputePaths for the endpoint before recording paths. This ensures all necessary paths are computed and cached before any operations that might trigger map modifications, eliminating the race condition. Additional unique_ptr is not ideal so eventually we might want to use BumpAllocator
seldridge
approved these changes
Oct 16, 2025
Comment on lines
-733
to
+734
| DenseMap<std::pair<Value, size_t>, SmallVector<OpenPath>> cachedResults; | ||
| DenseMap<std::pair<Value, size_t>, std::unique_ptr<SmallVector<OpenPath>>> | ||
| cachedResults; |
Member
There was a problem hiding this comment.
Unrelated, consider a struct over std::pair in a follow-on. This has the benefit of readability as well as possibly getting this onto one line. 😉
dtzSiFive
approved these changes
Oct 16, 2025
uenoku
force-pushed
the
dev/hidetou/fix-longest-path
branch
from
e5f0bb0 to
bfd6106
Compare
Member
Author
cacheResults is intended to cache results under the current module scope and the cache must be frozen if accessed from other threads (which are basically computing paths from parent scope). The issue was that cache was not pre-computed properly in the module scope. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This commit addresses a race condition and memory corruption issue in the LongestPathAnalysis where concurrent access to cached results could lead to iterator invalidation and memory corruption. The issue occurred when the cachedResults DenseMap was modified during iteration, causing existing iterators and references to become invalid and point to corrupted memory.
The fix changes the cachedResults map to store unique_ptr<SmallVector> instead of SmallVector directly. This ensures that the underlying SmallVector objects remain at stable memory locations even when the DenseMap is rehashed or modified, preventing iterator invalidation and memory corruption.
Additionally, in markRegEndPoint, we now preemptively call getOrComputePaths for the endpoint before recording paths. This ensures all necessary paths are computed and cached before any operations that might trigger map modifications, eliminating the race condition.
Additional unique_ptr is not ideal so eventually we might want to use BumpAllocator