search

brainboxdotcc / DPP

C++ Discord API Bot Library - D++ is Lightweight and scalable for small and huge bots!
https://dpp.dev/
Apache License 2.0
1.07k stars 163 forks source link

ci: allow runners to use mold, increased concurrency to 4 #1198

Closed Jaskowicz1 closed 3 months ago

Jaskowicz1 commented 3 months ago

This PR attempts to speed up the action runners by allowing them to use mold and also changing the concurrency to 3 instead of 2 (github action runners were updated to 4 cores a while ago).

The checklist does not apply here.

netlify[bot] commented 3 months ago

Deploy Preview for dpp-dev ready!

Name Link
Latest commit 4fa6cb83aac7af2a1506efb62278e43db275d48d
Latest deploy log https://app.netlify.com/sites/dpp-dev/deploys/6692e4fdfff7c500088ac581
Deploy Preview https://deploy-preview-1198--dpp-dev.netlify.app
Preview on mobile
Toggle QR Code...

QR Code

Use your smartphone camera to open QR code link.

To edit notification comments on pull requests, go to your Netlify site configuration.

braindigitalis commented 3 months ago

can we bump up the concurrency on gh runners to 4 then instead of 3? do we see any improvement?

Jaskowicz1 commented 3 months ago

can we bump up the concurrency on gh runners to 4 then instead of 3? do we see any improvement?

That was done in the recent push :)

Seems that we barely get an improvement across the board. A couple of runners are faster and a couple of them aren't any different.

Mishura4 commented 3 months ago

I mean mold will only affect linking which is a fraction of the time the runners spend on D++ mold really shines when LTO is enabled

Jaskowicz1 commented 3 months ago

@braindigitalis Ignore my last comment about "little results".

CodeQL is now ~3 minutes faster (comparing https://github.com/brainboxdotcc/DPP/actions/runs/9921615282/job/27409774387?pr=1197 with https://github.com/brainboxdotcc/DPP/actions/runs/9921876515/job/27410330135?pr=1198).

The Action Runners now builds on average 30 seconds to 2 minutes faster (seems like Clang15, g++13, g++14, and g++15, were nearly 2 minutes faster now).

Self hosted runner arm7hf didn't benefit at all, same with self hosted runner arm64.

OSX didn't have any change to concurrency, nor linker. Neither did Windows.

Jaskowicz1 commented 3 months ago

I mean mold will only affect linking which is a fraction of the time the runners spend on D++ mold really shines when LTO is enabled

true true, the reason it's dropped a good bit is the concurrency being upped to 4.

Jaskowicz1 commented 3 months ago

CodeQL is now down another minute! We're seeing 17 mins now!