Improve VariableWidthBlock deserialization #15883
Merged
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
pettyjamesm
force-pushed
the
further-improve-variablewidthblock-encoding
branch
2 times, most recently
from
19f13fa to
ade22a2
Compare
radek-kondziolka
approved these changes
Jan 30, 2023
Contributor
radek-kondziolka
left a comment
radek-kondziolka
left a comment
There was a problem hiding this comment.
LGTM but please run microbenchmarks
core/trino-spi/src/main/java/io/trino/spi/block/VariableWidthBlockEncoding.java
Outdated
Show resolved
Hide resolved
pettyjamesm
force-pushed
the
further-improve-variablewidthblock-encoding
branch
3 times, most recently
from
250f2fb to
cc6286f
Compare
pettyjamesm
force-pushed
the
further-improve-variablewidthblock-encoding
branch
from
cc6286f to
82b6d42
Compare
Member
Author
|
@raunaqmorarka - let me know if you have any questions when you get a chance to review this PR. |
Member
|
@radek-starburst could you please run this and post results from your system as well ? |
Contributor
|
My results: |
raunaqmorarka
approved these changes
Feb 9, 2023
core/trino-spi/src/main/java/io/trino/spi/block/VariableWidthBlockEncoding.java
Outdated
Show resolved
Hide resolved
Avoids allocating an unnecessary additional array allocation when deserializing VariableWidthBlocks, since we can use the final offsets array to temporarily store the serialized lengths values and transform them back into valid offsets in place. Benchmark (nulls) Mode Before After Units BenchmarkBlockSerde.deserializeSliceDirect 0 avgt 4.174 ± 0.455 3.700 ± 0.092 ns/op BenchmarkBlockSerde.deserializeSliceDirect .01 avgt 4.943 ± 0.340 4.575 ± 0.330 ns/op BenchmarkBlockSerde.deserializeSliceDirect .10 avgt 5.471 ± 0.329 5.252 ± 0.078 ns/op BenchmarkBlockSerde.deserializeSliceDirect .50 avgt 6.730 ± 0.815 3.592 ± 0.167 ns/op BenchmarkBlockSerde.deserializeSliceDirect .90 avgt 2.636 ± 0.265 2.646 ± 0.260 ns/op BenchmarkBlockSerde.deserializeSliceDirect .99 avgt 1.303 ± 0.068 1.708 ± 0.038 ns/op
pettyjamesm
force-pushed
the
further-improve-variablewidthblock-encoding
branch
from
82b6d42 to
3adf7c3
Compare
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 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
Follows up from #15760 to further optimize
VariableWidthBlockEncoding#readBlockto avoid allocating an unnecessary additional lengths array, since the final offsets array can temporarily store the serialized lengths of non-null positions and transform them back into valid offsets in-place.Avoiding the extra array allocation does not reduce CPU throughput, and generally also improve as a result of being able to trigger CMOV optimizations a special path for the "no-nulls" case.
Benchmarks:
Release notes
(x) This is not user-visible or docs only and no release notes are required.
( ) Release notes are required, please propose a release note for me.
( ) Release notes are required, with the following suggested text: