NVIDA’s high-performance HPC SDK is a real gem and an excellent tool for GPU optimized development. I have been designing AI algorithms for C++ parallel environments and CUDA Fortran for performance benchmarking. The CUDA Math API, Tensor Cores, and cuTensor are tremendously handy for AI developers in cases where real-time performance is essential on edge systems. Linear algebra algorithms on cuSPARSE are very efficient. The Open MPI with CUDA aware libraries is a formidable environment. I did not get a chance to explore the portability question.
This event should help improve productivity for AI and ML developers.
https://info.nvidia.com/simplifying-hpc-workflows-with-containers-panel-discussion-reg-page.html
#AI #highperformancecomputing #gpucomputing #MPI #openMPI #nvidia #CUDA #HPC
No comments:
Post a Comment
Note: Only a member of this blog may post a comment.