CUDA Develop

[逄晓霖]

Hi everyone,

Our research group is currently developing new functional support for Vampire. I am responsible for the parallelization part of these functions, and the MPI version is almost complete.

I would like to know the current status of the CUDA branch. Are there any recent developments or recommended approaches for integrating GPU acceleration with Vampire?

It would also be great if I could get some suggestions or best practices from developers or other users who have experience with the CUDA branch.

Thank you！

[gabo...@gmail.com]

Since I don't have an NVIDIA GPU, I'm unable to take advantage the parallel computations with CUDA.

I remember reading in the post at [1] support for CUDA 10.

For your development, it might be helpful for others if you could support the newer CUDA Toolkits [2] for 11.x and 12.x .

In the CUDA branch makefile at [3,] or the past post at [4], you should see that the makefile currently only supports up to Compute Capability 7.5.

So for your development, it might be helpful to others if you could also support NVIDIA GPU with higher Compute Capability (8.0 - 12.0 seen at [5]).

[1] https://groups.google.com/g/vampire-users/c/t4bu2yuxZlI/m/8yvcLk6HAAAJ

[2] https://developer.nvidia.com/cuda-toolkit-archive

[3] https://github.com/richard-evans/vampire/blob/cuda/src/cuda/makefile

[4] https://groups.google.com/g/vampire-users/c/fCxSGtOpJv4/m/7nXMPjbsBQAJ

[5] https://developer.nvidia.com/cuda-gpus

Sincerely,

Gavin

VAMPIRE user

[Richard Evans]

Hi,

We are working on the CUDA version this summer, updating the various bits of CUDA code with new functionality and also the newer CUDA libraries. So, this should be much improved in the next few weeks. A new update is coming with quite a few changes, hopefully by the end of the year.

All the best,

Richard

[逄晓霖]

Hi ,

Thank you for the info and updates! Your suggestions are very helpful. 

And I'm looking forward to the upcoming CUDA-enhanced version of Vampire — and it would be great if a new manual comes along with it!

Best,
Xiaolin Pang