mrnorman
Port MiniWeather over the CUDASTF programming model
This PR introduces a new version of the MiniWeather benchmark based on the CUDASTF programming model.
CUDASTF is shipped in NVIDIA's CCCL project, and implements task-parallelism as a C++ header only library.
This example shows how to leverage CUDA graphs to hide latencies on small problem sizes, or to scale parallel_for kernels over multiple devices of the same machine (e.g. a DGX platform).
Following https://github.com/mrnorman/miniWeather/pull/20, we might want to use a naming directory structure cpp_$(project_name). In this case project_name is cudastf.
Sorry for the long delay. I'm still not getting notified of PRs even though I'm "watching" this on github. I'll look at this soon, and thanks so much for submitting it.
Hello,
No problem, there is no hurry and we keep improving it !
Thanks, Cédric
De : Matt Norman @.> Envoyé : mercredi 15 janvier 2025 16:45 À : mrnorman/miniWeather @.> Cc : Cedric Augonnet @.>; Author @.> Objet : Re: [mrnorman/miniWeather] Port MiniWeather over the CUDASTF programming model (PR #19)
Sorry for the long delay. I'm still not getting notified of PRs even though I'm "watching" this on github. I'll look at this soon, and thanks so much for submitting it.
— Reply to this email directly, view it on GitHubhttps://github.com/mrnorman/miniWeather/pull/19#issuecomment-2593271699, or unsubscribehttps://github.com/notifications/unsubscribe-auth/BFWSSGSEMPIXFSKMAT4WHW32KZ7ADAVCNFSM6AAAAABQQTU3GOVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDKOJTGI3TCNRZHE. You are receiving this because you authored the thread.Message ID: @.***>