Open eyalroz opened 4 years ago
Going back and cleaning up my old repos and see that I must have missed this. My nvtx wrapper stuff has since been moved into the official NVTX repo here: https://github.com/NVIDIA/NVTX/tree/dev/cpp
Would love to collaborate. It's definitely on my to do list to add more of the features like thread naming and such.
Umm, well, now that this repository is officially adopted by NVIDIA, it seems like there's less of an opportunity to moving it and my API wrappers closer together - as I assume NVIDIA would not want any sort of inter-dependency... right?
Hello Jake,
Long time no speak. Actually, do I even have your email address? I think I don't. would you mind emailing me something?
Anyway... as I was looking for your address in your repos, I came upon this one. As you are probably aware, I am - unfortunately and fortunately - the maintainer of the modern-C++ wrappers for CUDA on the host side: cuda-api-wrappers. NVTX is a part of CUDA, and thus, already partially covered by the wrappers library.
... but then, looking at your documentation, I don't see much that regards use of the GPU. Is this intentional?
Anyway, to be honest, the
profiling.hpp
/profiling.cpp
file pair incuda-api-wrappers
, while working, is not the part I've invested the most time and care designing and implementing. Specifically, it doesn't cover NVTX3-specific features (such as domains and categories), and could otherwise use a bit of polish. To my credit I'll say that I do wrap some functionality which you don't - OS thread naming, without the user code needing to include any OS-specific headers.In terms of infrastructure - you seem to have serious unit test coverage; I have some example programs, which cover a lot, but not everything; but I seem to have a better worked-out
CMakeLists.txt
(I've learned some CMake voodoo since last year).Anyway, do you think there's potential for us to collaborate here?