Efficent parallel use of eval caches #10590
Open
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Outdated suggestions cannot be applied.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.
Motivation
This commit makes evaluation caches behave properly in presence of parallel
users. Transactions and SQL statements themselves are designed in such a way
that there cannot be race conditions while keeping transaction very small.
(Transactions used to last for the whole lifetime of the cache).
On top of #10570 it enables to significantly speed up evaluation of a large
number of attributes by splitting them across several calls to nix.
From rough measurements on 16 cores it goes from 12s to 5s for ~200 attributes in the same flake.
This also fixes a wide range of issues discovered during the implementation,
like the
insert or update
statements that would create a brand new rowid andsilently make all the children of the old rowid unreachable, invalidating
significant portions of the cache.
This code has been stress-tested a bit, so it should be devoid of obvious mistakes.
Context
Priorities and Process
Add 👍 to pull requests you find important.
The Nix maintainer team uses a GitHub project board to schedule and track reviews.