-
Notifications
You must be signed in to change notification settings - Fork 40
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
Add benchmarking CI #419
Add benchmarking CI #419
Conversation
Do we need another codespeed instance for this? https://speed.juliagpu.org/ is only for CUDA.jl right now. I wonder if there isn't a better solution nowadays; other projects seem to be tackling this slightly differently:
|
a8ddb34
to
dadbf35
Compare
[only benchmarks] [skip docs]
I think the GemmKernels.jl solution would be the best fit since these cannot be run with Github Actions. I'll finish this up when I get more time. |
088d5ee
to
a5b02cf
Compare
The LuxDL.jl approach also runs on Buildkite: JuliaGPU/KernelAbstractions.jl#510 (comment) |
e087748
to
5251b6c
Compare
Moved over to #420 with a JuliaGPU branch in an attempt to figure out why it stopped working. Will continue development on that branch. Sorry for the noise. |
Copying over the benchmarks from CUDA.jl.
I'm not sure if I converted them properly. Some function seem to be missing Metal implementations (like
reverse
).The final (and biggest) problem is how inconsistent these results have been. Simply rerunning the benchmarks on the same code gives some huge performance differences. There is always at least one benchmark that is >20% slower or faster.
See #418 (comment)
Todo: