I’m hosting an open source project on GitLab.com, and am currently using the shared runners to run my tests. Thank you, GitLab, for the free compute resources.
As I add more tests, I’m catching a lot of bugs, but unfortunately the tests take longer and longer to run. Some of that can be helped by getting rid of obsolete tests, optimizing things, etc., but that’s human-time intensive and only goes so far.
But a lot of CI loads are embarrassingly parallelizable. Testing with a local runner, switching from “prove -j1” to “prove -j2” significantly cuts the time, with only a slight increase in total CPU time. So it’d be really nice of GitLab if you all were to provide some runners with 2 vCPUs, maybe selectable by tag for jobs that are optimized for them.