-
Notifications
You must be signed in to change notification settings - Fork 30.1k
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
benchmark: add bench for zlib gzip + gunzip cycle #20034
Conversation
}); | ||
|
||
function main({ inputLen, duration, type }) { | ||
const buffer = Buffer.alloc(inputLen, fs.readFileSync(__filename)); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Nit: any reason for using fs.readFileSync(__filename)
instead of an arbitrary fill character? Is it for making the chunks to gzip/gunzip realistic?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@lpinca Yes, exactly that. If you have other suggestions, I’ll take them too :)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Nah, that's ok and I have no better ideas.
inputLen: [1024], | ||
duration: [5], | ||
type: ['string', 'buffer'] | ||
}); |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
These options have to be reflected in the benchmark test as well (parallel/test-benchmark-zlib.js
).
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
@BridgeAR they are no? See line 14.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
As far as I can tell, inputLen
and duration
have to be added to the test.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Oh, true.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Actually inputLen
and duration
are used in the test.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Right now we just have:
runBenchmark('zlib',
[
'method=deflate',
'n=1',
'options=true',
'type=Deflate'
]);
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
It seems this benchmark is like net benchmarks where throughput is calculated after x sec. I don't think this will be run in CI. Or I am not understanding the issue :)
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Right but we have tests that run these benchmarks to confirm that they work and also to confirm that we didn't break some functionality. See test/parallel/test-benchmark-zlib.js
.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I see now, I didn't read the original comment carefully. Sorry for the noise.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Done!
bench.end(readFromOutput * 8 / (1024 ** 3)); | ||
|
||
// Cut off writing the easy way. | ||
input.write = () => {}; |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Should not better write
be replaced instead of input.write
?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I’d find it a bit surprising to override something defined with function foo(){}
, tbh…?
Ping @addaleax |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Just to make sure it does not land as is.
Originally wrote this for some work that is going to take a while longer before it’s ready to be PR’ed, so it seems fine to start with this on its own.
c903705
to
5c2517f
Compare
test/parallel/test-benchmark-zlib.js
Outdated
'type=Deflate' | ||
'type=Deflate', | ||
'inputLen=1024', | ||
'duration=1' |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I'm -0 on this. It's heuristic (so still might fail on a really slow or clogged machine), probably safer to call bench.end(gb)
with
// Give result in GBit/s, like the net benchmarks do.
// Never return 0, so that tests won't fail on really slow machines.
const gb = (readFromOutput * 8 / (1024 ** 3)) || 1
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
ohh and #20128 - settings duration=1
will add a 1sec test to the suite.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Do we do this elsewhere? Otherwise it seems like it would falsify test results – what if the test is actually broken and returns 0?
Would you be be okay with the current code if it passes a stress test run?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Regarding the duration: Yeah, that’s not great. It’s not quite as bad as if it were a sequential
test though…
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Let me read through the test and harness a little bit more. AFAIK the test-benchmark-*
set are there "mostly" for syntax, API and regression testing. Designing correctness tests for the benchmarks might be out-of-scope.
Maybe we can just relax the Error: called end() with operation count <= 0
rule anyway 🤔
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Found it:
Line 197 in 48e90ed
if (!process.env.NODEJS_BENCHMARK_ZERO_ALLOWED && operations <= 0) { |
so calling:
runBenchmark('zlib',
[
'method=deflate',
'n=1',
'options=true',
'type=Deflate',
'inputLen=1024',
'duration=0.001'
],
{
'NODEJS_BENCHMARK_ZERO_ALLOWED': 1
});
should get the test to pass.
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
Thanks, done!
|
Landed in e797d5b |
Originally wrote this for some work that is going to take a while longer before it’s ready to be PR’ed, so it seems fine to start with this on its own. PR-URL: #20034 Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Luigi Pinca <luigipinca@gmail.com> Reviewed-By: Ruben Bridgewater <ruben@bridgewater.de>
Originally wrote this for some work that is going to take a while longer before it’s ready to be PR’ed, so it seems fine to start with this on its own. PR-URL: #20034 Reviewed-By: James M Snell <jasnell@gmail.com> Reviewed-By: Luigi Pinca <luigipinca@gmail.com> Reviewed-By: Ruben Bridgewater <ruben@bridgewater.de>
Originally wrote this for some work that is going to take a while
longer before it’s ready to be PR’ed, so it seems fine to start
with this on its own.
Checklist
make -j4 test
(UNIX), orvcbuild test
(Windows) passes