refine auto_growth allocator #35732
Merged
refine auto_growth allocator #35732
Conversation
This file contains 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
|
Thanks for your contribution! |
zhiqiu
force-pushed
the
dev/refine_allocator
branch
from
05e34d5
to
279478f
Compare
zhiqiu
force-pushed
the
dev/refine_allocator
branch
from
676dfa6
to
e562d20
Compare
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Outdated suggestions cannot be applied.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.
PR types
Performance optimization
PR changes
Others
Describe
refine auto_growth allocator
CUDA does address alignment itself, so in most case we do not need to manually perform address alignment.
refer: https://stackoverflow.com/questions/14082964/cuda-alignment-256bytes-seriously
The problem is that, the original implementation may alloc more 256 bytes due to AlignedAllocator and the 256 bytes may put into
free blocks. Which may introduce more memory fragmentation and increase the time to find fit block when allocing new allocation.before
after
The max number of free blocks in standalone_executor_test(single thread, no gc version, for consistance) decreases: 200->185