bosilca

Honestly, the only one of y'all that brought something really meaningful and constructive to this discussion was @SoapGentoo. The definition that applies the best to our case is `If one...

It would be interesting to see a comparison with the current allocator.

I'm very confused about this issue. First because the `parsec_ce` was never meant for applications, but was created to isolate the communication library (be it MPI, LCI or UCX) itself...

It doesn't have to come back quickly. We have the epoch when the task was evicted, we can define a quiet period for the task during which it is not...

We have some runners available on github. I moved this on our github runner, let's see how it goes. I means obviously it will fail as we can't build without...

No msys2-ucrt64 runner from github picked this job up and there is no info on the timeout. Not sure how to move this forward.

I don't see any github runner having done anything here, I don't grok what caps are we reaching. The repo is public, I use a public runner and [here](https://docs.github.com/en/actions/using-github-hosted-runners/using-github-hosted-runners/about-github-hosted-runners#standard-github-hosted-runners-for-public-repositories) it...

I saw a notification with a lldb stack trace but I can't find it here. Weird! Anyway, it seemed to indicate that the `complete_stage` was pointing to an incorrect location.

Added support for `CUDA_VISIBLE_DEVICES`.

I'm not sure what you are referring to. Are you saying that if `CUDA_VISIBLE_DEVICES` is set CUDA only make the specified devices visible to parsec so we can ignore the...