Re: Passing gil::image and gil::kernel to CUDA

From: Mateusz Loskot (mateusz_at_[hidden])
Date: 2019-12-09 14:47:28

On Thu, 10 Oct 2019 at 19:45, Olzhas Zhumabek
<anonymous.from.applecity_at_[hidden]> wrote:
>
> I would like to attempt GPU implementation of some image processing
> algorithms for my university project. I would like to be able to call as
> much of original GIL as possible, minus the io extension.
>
> *Problem description*
> Memory is no longer uniform when heterogeneous system is used. CPU cannot
> write to GPU's on-board VRAM without experiencing slow down.
> [...]

Olzhas,

Here is an interesting reading from Joel de Guzman (author of Boost.Spirit)
on the issue of transferring data to/from GPU:
https://www.cycfi.com/2019/04/gpu-dsp-latency/

You may want to ceck the other numerous materials Joel
links from his post, including his DSP library
https://github.com/cycfi/Q
and
https://blog.theincredibleholk.org/blog/2012/11/29/a-look-at-gpu-memory-transfer/

Best regards,
-- 
Mateusz Loskot, http://mateusz.loskot.net