decoders: replace binary.Read with a version without reflection and allocations #141
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
vincentbernat
force-pushed
the
feature/no-binary-read
branch
from
7e5362b to
e65432c
Compare
|
By using the |
vincentbernat
changed the title
|
I may have introduced a regression, let me investigate! |
vincentbernat
force-pushed
the
feature/no-binary-read
branch
from
e65432c to
c797534
Compare
|
OK for now. I have fixed the bug. Previously, when we provide |
vincentbernat
added a commit
to akvorado/akvorado
that referenced
this pull request
Jan 19, 2023
``` benchmark old ns/op new ns/op delta BenchmarkDecodeEncodeNetflow/with_encoding-12 7664 7548 -1.51% BenchmarkDecodeEncodeNetflow/without_encoding-12 7273 6872 -5.51% BenchmarkDecodeEncodeSflow/with_encoding-12 14029 9905 -29.40% BenchmarkDecodeEncodeSflow/without_encoding-12 13784 9245 -32.93% benchmark old allocs new allocs delta BenchmarkDecodeEncodeNetflow/with_encoding-12 143 134 -6.29% BenchmarkDecodeEncodeNetflow/without_encoding-12 143 134 -6.29% BenchmarkDecodeEncodeSflow/with_encoding-12 295 137 -53.56% BenchmarkDecodeEncodeSflow/without_encoding-12 295 137 -53.56% benchmark old bytes new bytes delta BenchmarkDecodeEncodeNetflow/with_encoding-12 8272 8160 -1.35% BenchmarkDecodeEncodeNetflow/without_encoding-12 8272 8160 -1.35% BenchmarkDecodeEncodeSflow/with_encoding-12 9816 8216 -16.30% BenchmarkDecodeEncodeSflow/without_encoding-12 9816 8216 -16.30% ``` "old" is still before using the fork. See netsampler/goflow2#141 for details of the fix.
vincentbernat
force-pushed
the
feature/no-binary-read
branch
from
c797534 to
34e153f
Compare
Instead of allocating small slices, we rely on the fact that most call sites are providing a `bytes.Buffer` and use the `Next()` method. For sFlow decoding, in my case, we get a 33% speedup. A `bytes.Reader` would even be more efficient, but unfortunately, they don't have a `Next()` method. While Go should be smart enough to make the allocation of `bs` on the stack, it does not, even when `io.ReadFull()` is inlines. ``` decoders/utils/utils.go:23:13: make([]byte, n) escapes to heap ```
vincentbernat
force-pushed
the
feature/no-binary-read
branch
from
34e153f to
7562215
Compare
Merged
|
Added benchmark: |
|
Will integrate it with v2 as well. |
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.
See #140. Not really tested. Still need to get rid of
bytes.Buffer.