guetzli
guetzli copied to clipboard
CUDA and/or OpenCL usage for DCT and other functions?
Current execution time is horrible for any practical use case. I was wondering why don't we simply plug-in available functions from NVIDIA's NPP library? It wouldn't mean porting anything to CUDA simply plugging in available functions. Anyone working on GPU port?
CUDA is Nvidia™ only and cannot function without proprietary drivers, so I think it's not worth wasting time on it. Only OpenCL would make sense.
Currently I think CUDA is popular enough. And with specified version of CUDA driver, I think it is useful to do that.
@leafjungle is there a CUDA driver for AMD or Intel chips?
There is a CUDA and OpenCL version:
https://github.com/google/guetzli/pull/227
@pornel , yeah, that sounds great. I git clone the from http://github.com/ianhuang-777/guetzli. And I can not found cu_mem defined in anywhere of cuda or opencl.