libcudacxx
libcudacxx copied to clipboard
NVIDIA
Support <random> in libcu++ and libnv++
Hello, is there a timeline for adding the <random>, <cuda/std/random>, and/or <cuda/random> headers to libcu++?
Likewise, is there any timeline for adding these headers to libnv++? (I couldn't find any repo for libnv++ on github so that's why I'm asking here instead).
I'd love to be able to run the thrust Monte-Carlo example using one of the three headers above instead of <thrust/random.h> if it were available.
We can attempt to put it on our roadmap. This has potential security concerns I imagine that would need to be answered.
Apologies for my ignorance but what might the potential security concerns be? And how do they come into play for <random>, <cuda/std/random>, or <cuda/random> but not for <thrust/random.h>?
Just to give you some more context for my request - I'm using CUDA and Thrust to run financial Monte Carlo simulations on GPU for my employer, and the thought of being able to do this with just the NVIDIA C++ Standard Library is very tantalizing. The big missing piece at the moment is <random>.
If it would be better to start this discussion through the sales channel instead, then please let me know. I just figured I'd try to raise the issue here first.
Lastly, am I right that there is no separate repository yet for libnv++ or nvc++? If so, then is it acceptable to raise issues for either of those two here in the meantime?
Hi, I just wanted to follow up on this issue. If this isn't on your official roadmap yet, then is something that you would accept community contributions for?
I still believe that adding support for one or all of the random libraries is going to be a huge win for people using the NVIDIA C++ Standard Library in HPC, especially in finance. Please let me know if there's any way that I can help contribute to it.