Use go slices #49
Open
Use go slices #49
+370
−191
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
kamstrup
reviewed
Jan 3, 2023
kamstrup
approved these changes
Jun 2, 2023
|
|
||
| func TestDecompressTooLarge(t *testing.T) { | ||
| src := []byte{40, 181, 47, 253, 228, 122, 118, 105, 67, 140, 234, 85, 20, 159, 67} | ||
| _, err := Decompress(nil, src) |
There was a problem hiding this comment.
From the test name I gather that the error here is that the decompressed size is too larger for the dst buf (nil)? It would be a bit easier to read if the dst buf was non-nil, like maybe 1 byte or something
| zr.sizes.dstPos = 0 | ||
|
|
||
| inHdr := (*reflect.SliceHeader)(unsafe.Pointer(&zr.inBuf)) | ||
| outHdr := (*reflect.SliceHeader)(unsafe.Pointer(&dst)) |
There was a problem hiding this comment.
you use the noescape() trick in compressInternal(), but not for dst here... I guess there is no chance stack allocated buffers could be used in this context?
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Outdated suggestions cannot be applied.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.
Fixes #48, #33
I've rewritten how the CGO wrapper is done to achieve two things:
Here's benchmark results against master (run on M1 using go 1.18):
CPU time
Throughput
Now you will notice that many results (particularly ones working with tiny buffers) are reporting being up to 20% slower, turns out this is because the CGO pointers checks are now taking significant amount of time, then again we're talking a few nanoseconds and this is completely negligible with larger buffers, so IMO this isn't that bad.
I've also made the Reader write directly into the provided buffer (if it's large enough), and those benchmarks show the biggest gain - about 5% faster when using large buffers. The ability to use the go slice directly could be also used in the Writer, but let's leave that for another PR.
I've also re-run the benchmarks with
GODEBUG=cgocheck=0, and the results definitely look even better:Had to use a gist, github didn't like this long PR description - https://gist.github.com/mhr3/84f58f62353ef3b9db30288df00fa2b3