feat(blockstm): cache pre-state to optimize value-based validation#25814
Closed
mmsqe wants to merge 8 commits intocosmos:mainfrom
Closed
feat(blockstm): cache pre-state to optimize value-based validation#25814mmsqe wants to merge 8 commits intocosmos:mainfrom
mmsqe wants to merge 8 commits intocosmos:mainfrom
Conversation
Codecov Report❌ Patch coverage is Additional details and impacted files@@ Coverage Diff @@
## main #25814 +/- ##
==========================================
- Coverage 68.85% 68.78% -0.08%
==========================================
Files 926 930 +4
Lines 60567 60956 +389
==========================================
+ Hits 41706 41931 +225
- Misses 18861 19025 +164
... and 3 files with indirect coverage changes 🚀 New features to boost your workflow:
|
yihuang
reviewed
Jan 26, 2026
| const minSnapshotKeysCap = 1024 | ||
|
|
||
| type mvIndexEntry[V any] struct { | ||
| key []byte |
Collaborator
There was a problem hiding this comment.
If we'll choose an unordered data structure, you may give sync.Map a try?
Collaborator
Author
There was a problem hiding this comment.
sec/op and B/op are similar on average, but seems syncmap allocate more (allocs/op geomean +10.08%).
Collaborator
There was a problem hiding this comment.
allocs/op don't matter if sec/op and B/op is the same right?
The main trade-off in all unordered data structure based approach, is the heavy iteration overhead.
* add Has operation validation that ignores version changes, only checking for existence state (Found/Deleted) * reduce memory overhead by packing HasDescriptor into ReadDescriptor * add negative caching for Has lookups to prevent repeated fallback to storage for non-existent keys * add readFound skip large value * handle snapshots * avoid storage.Get for Has() resolve fix
* add mvIndex sharded key->version-tree map * update mvData iterator/snapshot paths to use key snapshots * refactor MVIterator to iterate snapshot keys * cache mvIndex key snapshot
switch per-key version trees to an in-place, locked tidwall/btree to avoid btree.Copy on updates
* replace lock-holding btree iteration with a chunked cursor that releases locks between refills * use COW key snapshots so iterators stay serializable without blocking writers while waiting on dependencies
* avoid per-key COW snapshots: keep 0/1 item inline and promote to a locked B-tree on demand * avoid per-Set allocations on the single-item path
Contributor
|
This pull request has been automatically marked as stale because it has not had recent activity. It will be closed in 7 days-before-close if no further activity occurs. |
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
Description
Closes: #25777