Omri Mor

> For external users, such as TTG, this is not the case. Yup, that would be a problem :). For my above suggestion, I'd add that the comm engine should...

Have you looked at performance at all? I think I've discussed with @bosilca ad nauseam regarding the implications regarding communication prioritization. I might pull a version of this into my...

> didn't get worse though That's relevant! How much did you scale and were you using `THREAD_MULTIPLE`? George's hypotheses regarding this are, if I recall correctly, along the lines that...

> Argh of course not, CI doesn't test the GET protocol. Test fail if run with `PARSEC_MCA_runtime_comm_get=1`. I still get `MPI_ERR_TRUNCATE` errors, so there is a problem with GET. `MPI_ERR_TRUNCATE`...

> The MPI backend is still student research quality, at best. "research quality" The LCI backend is better, in my _humble_ opinion, and is certainly better-documented. It still has a...

This should be assigned to @bosilca, per his request.

https://github.com/ICLDisco/parsec/blob/530601533fc80ad27d7c12cdfda56744746baac9/parsec/recursive.h#L55 Doesn't this line force recursive taskpools to execute on the CPU? I'm not entirely surprised that there are issues with recursive + GPU, since I don't think it was...

I feel like the problem with this approach is that priorities are (generally) an aspect of the algorithm that is to be used by the scheduler to make scheduling decisions,...

Duplicate of #269. I forgot that I'd fixed this in my branch and I guess hadn't ported it to mainline PaRSEC.

My brief understanding of this is that `parsecf.F90` declares `parsec_context_t` to be a type, compatible with C, with a single member, named `PTR`, of type `C_PTR`. That is to say,...