Jonathan R. Madsen

> I'm not sure how we'd handle OMP, CUDA backend library shipping, though some libs like pytorch or tensorflow could likely be used as inspiration for that. Create conda-forge packaging.

> Conda and PyPI are two completely different ecosystems, and mixing them is not really recommended. @tylerjereddy Yes but when it comes to mixing C++ and Python, pip is not...

> I thnk the problem here is to create a one-line pip install for pykokkos which will install pykokkos-base as a dependency @namehta4 I am pretty sure that you could...

> SciPy and NumPy both mix C++ and Python (and Fortran for SciPy) and provide solutions in both ecosystems. Yes, but they don't have to deal with multiple backends causing...

I just did a `pip install pykokkos-base` and ran into an issue because it defaulted to enabling CUDA and I didn't have `nvcc` in my path. All I had to...

BTW, it should probably be noted in the chrome-tracing that this is effectively the newer version of that tool. Do we want to apply formatting to the external perfetto files...

Yeah, the problem here is that the parallelism is frighteningly terrible. With 12 threads, the total CPU utilization of ~160% and it takes 85+ minutes to complete with a binary...

Right, I am already using `BPatch::openBinary` as described above to get the address space when I am doing a binary rewrite — this is what was happening in the 85...

@wcohen It is not spending time in the same area as #1305 -- it is spending all of its time parsing (and waiting on locks ti parse): ```console Total: 30841...

There are performance and other issues noted in these which help explain what is going on here - https://github.com/dyninst/dyninst/issues/1335 - https://github.com/dyninst/dyninst/pull/1336