search

JuliaAttic / CUDA.jl

DEPRECATED: old Julia programming interface for CUDA
64 stars 20 forks source link

Add a short note stating that development stopped. #19

Closed vchuravy closed 8 years ago

vchuravy commented 8 years ago

I think this is fair to say and about time, also see discussion in https://github.com/JuliaGPU/CUDA.jl/pull/16.

@maleadt You were thinking or ripping the high-level abstraction layers out of CUDA.jl and keeping the low-level functionality, right? I was wondering about how to best do that without causing pain to users that are currently using CUDA.jl.

cc: @lindahua @timholy

timholy commented 8 years ago

:+1:

maleadt commented 8 years ago

+1. Maybe print a deprecation warning upon import, say, add to the CUDA.jl base file:

Base.warn_once("CUDA.jl is deprecated, please use CUDArt.jl instead")

@vchuravy I'm going to take a stab at reorganizing the CUDA functionality as discussed in #16, but due to time restrictions most likely not before next month. Not to inflict too much pain on current users, I'm thinking of putting the low-level driver functionality in CUDAdrv.jl (rather than CUDA.jl), and incrementally move CUDArt.jl over to use that functionality. Hopefully, that shouldn't break any code.

vchuravy commented 8 years ago

@maleadt Sounds good. I created https://github.com/JuliaGPU/CUDAdrv.jl and gave you write access