fix: Reuse vector in LocalPartition #12002
Closed
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
facebook-github-bot
added
the
CLA Signed
✅ Deploy Preview for meta-velox canceled.
|
|
This pull request was exported from Phabricator. Differential Revision: D67742489 |
Yuhta
force-pushed
the
export-D67742489
branch
from
232444d to
d4fbc7e
Compare
|
This pull request was exported from Phabricator. Differential Revision: D67742489 |
xiaoxmeng
reviewed
Jan 10, 2025
velox/common/memory/ByteStream.h
Outdated
| } | ||
|
|
||
| void appendBool(bool value, int32_t count); | ||
|
|
||
| /// Fast path used by appending one null in vector serialization. | ||
| template <bool kValue> | ||
| void appendOneBool() { |
There was a problem hiding this comment.
Can we have a unit test for this? Thanks!
There was a problem hiding this comment.
I will remove the specialization in the new version, just make sure it's inlined should be enough
velox/common/memory/ByteStream.h
Outdated
| @@ -411,8 +430,10 @@ class ByteOutputStream { | |||
| // The total number of bytes allocated from 'arena_' in 'ranges_'. | |||
| int64_t allocatedBytes_{0}; | |||
|
|
|||
| // Pointer to the current element of 'ranges_'. | |||
| ByteRange* current_{nullptr}; | |||
| // Copy of the current element in 'ranges_'. This is copied to avoid memory | |||
There was a problem hiding this comment.
I am not very sure if we want to this optimization until we see it cause noticeable regression on actual query. Thanks!
There was a problem hiding this comment.
We do see a few percentage (~3%) improvements in the E2E query by removing the 2 extra hops.
Yuhta
added a commit
to Yuhta/velox
that referenced
this pull request
Jan 10, 2025
Summary: X-link: facebookincubator/nimble#122 More than 10% of the CPU are spent on the destruction of local partition output when the load is high. Also add some optimizations for serialization. Optimization on `ByteOutputStream::appendBool` does not show significant gain in the query in example (because they are a lot small batches), but it is net gain and would be significant in large batches, so I leave it in the code. Differential Revision: D67742489
Yuhta
force-pushed
the
export-D67742489
branch
from
d4fbc7e to
d377ecb
Compare
|
This pull request was exported from Phabricator. Differential Revision: D67742489 |
xiaoxmeng
approved these changes
Jan 10, 2025
| vectorPool.push(makeVector(), 3); | ||
| vectorPool.push(makeVector(), 1); | ||
| auto vector = vectorPool.pop(); | ||
| ASSERT_TRUE(vector); |
There was a problem hiding this comment.
nit: ASSERT_TRUE(vector != nullptr);
| ASSERT_TRUE(vector); | ||
| ASSERT_EQ(vector.get(), vectors[0]); | ||
| vector = vectorPool.pop(); | ||
| ASSERT_TRUE(vector); |
Summary: X-link: facebookincubator/nimble#122 More than 10% of the CPU are spent on the destruction of local partition output when the load is high. Also add some optimizations for serialization. Optimization on `ByteOutputStream::appendBool` does not show significant gain in the query in example (because they are a lot small batches), but it is net gain and would be significant in large batches, so I leave it in the code. Differential Revision: D67742489
Yuhta
force-pushed
the
export-D67742489
branch
from
d377ecb to
0891cb0
Compare
|
This pull request was exported from Phabricator. Differential Revision: D67742489 |
facebook-github-bot
pushed a commit
to facebookincubator/nimble
that referenced
this pull request
Jan 10, 2025
Summary: X-link: facebookincubator/velox#12002 Pull Request resolved: #122 More than 10% of the CPU are spent on the destruction of local partition output when the load is high. Also add some optimizations for serialization. Optimization on `ByteOutputStream::appendBool` does not show significant gain in the query in example (because they are a lot small batches), but it is net gain and would be significant in large batches, so I leave it in the code. Reviewed By: xiaoxmeng Differential Revision: D67742489 fbshipit-source-id: 8e70dd128f31caa7909ed7c1e2b4ac1e59d7c87d
|
This pull request has been merged in 9dcfd39. |
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.
Summary:
More than 10% of the CPU are spent on the destruction of local partition output when the load is high.
Also add some optimizations for serialization.
Differential Revision: D67742489