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

From: Olzhas Zhumabek (anonymous.from.applecity_at_[hidden])
Date: 2019-10-21 07:06:57

>
> All that being said, I still think this would be a nice project for GIL
> (perhaps even GSoC ?), and might yield interesting performance
> improvements for a number of frequent algorithms.

I would be happy to implement the project under your lead.  Though
I was thinking if we could implement core GPU facilities before the GSoC
period,
and then may be create an option for coding competency test to implement
something using those. I believe most of the building blocks for feature
detection
are there now minus scale space generation struct definitions. Besides, the
patent for
a popular feature detection algorithm (SIFT or SURF, don't really remember
the name),
is expiring just before the GSoC start date, so I guess we could implement
that
along with a bunch of other simple ones, and those are very straight
forward compared
to A-KAZE that I'm working on (it uses flux concepts, which are very
differential math heavy).
May be I could co-mentor, if it is possible.

Best regards,
Olzhas