sleeepyjack commented on PR #2:
URL: https://github.com/apache/datasketches-cuda/pull/2#issuecomment-4504001845

   @tisonkun @leerho 
   That's an interesting question!
   
   `cuda_oxide` has been made public as an alpha release a week ago and I've 
been aware and involved in the project since a months or so. The nice thing 
about `cuda-oxide` is that it emits CUDA PTX directly from native Rust code - 
no FFI converting C/C++ as an intermediate whatsoever.
   
   The current `datasketches-cuda` project builds on top of our CUDA/C++ 
ecosystem (aka we're using `cuda::experimental::cuco::hyperloglog` which is 
implemented in C++).
   We are actively exploring how to make our existing CUDA/C++ ecosystem 
available to Rust without having to rewrite everything or having to go through 
FFI, but it's in very early stages.
   
   Since after all HLL is an "easy" algorithm, it should be straightforward to 
implement it natively in Rust and use `cuda-oxide` for GPU acceleration. For 
more complex sketch algorithms, we might want to centralize the GPU 
implementation for better maintainability.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]


---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to