Ralf Gommers
Ralf Gommers
@bwhitt7 I think a lot of this is covered now by your recent work. Would you be able to update this issue with the current state, and see if/when it...
That's a question for @ngoldbaum I think. It's not immediately clear to me whether there's something wrong with the way `atomic_load`/`atomic_store` is used here, or if the Intel compilers have...
> As an aside, is there really no way to capture all these build messages except to scroll up and copy? Output redirection with > doesn't seem to work. I...
> Do you want me to close this issue and open a new one with just the remaining problem(s)? Yes please, that would be helpful.
> I think a new ID is called for (CMake gives it `IntelLLVM`, This is what gh-10909 does. So that probably supercedes this PR.
It'd be great to get an exact reproducer for this @nickdesaulniers. Without knowing how you installed numpy, pytorch and whatever else is in your environment (venv or conda env?), this...
That benchmark result looks acceptable to me in principle - only larger overhead for
This is a large PR with potential performance implications and active discussion, so I removed the 1.16.0 milestone - we're not getting this one merged by tomorrow.
Thanks @kgryte for the detailed proposal. This function is heavily used and present everywhere, so it makes sense to add from that perspective. The main question I have at the...
My takeaway from the discussion in the community meeting was similar to the question @betatim asked above. Can actually be split into two: - Should we add a `length` keyword?...