Dag batch gathering scaling improvements #2318
Open
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
There is need to allocate rapidly memory which will be automatically freed when a new search iteration starts. This adds a custom allocator to manage memory.
Refactoring GatherMinibatch aims to improve the scaling to more threads. Task scheduling becomes faster. It allows task number scale up dynamically. These changes are aimed to be elo neutral when GPU is the bottleneck. CPU bottleneck gains a minor improvement. The plan is to base future changes to improve CPU performance.
This fixes a performance issue that I noticed when testing tinygyal vs master tinygyal match.
…ag_scaling_improvements
Menkib64
force-pushed
the
dag_scaling_improvements_v3
branch
from
3ce9244 to
3898767
Compare
Kernel might decide to suspend a thread when scheduling new tasks. This could result into a case where another writer could manage to write a newer task to the same bucket. Compare exchange avoids overwriting the newer pointer before the competiting reader has a chance to read it.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
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Outdated suggestions cannot be applied.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 is a major refactoring to the search implementation. The main focus is to reduce cost of scheduling task. There are many threading and memory management changes towards the goal.
I expect that there will be a few minor conflicts with #2312.