Blender 2.79: OpenCL on par with CUDA

Re:Blender 2.79: OpenCL on par with CUDA by shawn-butler on Wednesday April 12, 2017 @11:11AM (#14098415)

You might want to take a look at SYCL for higher-level abstractions if that was your preference.

There is an open-source implementation to dig into: https://github.com/Xilinx/triSYCL

[ Parent | on HN ]

Re:Blender 2.79: OpenCL on par with CUDA by tombert on Wednesday April 12, 2017 @11:22AM (#14098537)

Actually, nowadays since I mostly do esoteric functional stuff, I mostly use the wonderful `accelerate` package for Haskell: https://www.stackage.org/package/accelerate

It abstracts away most of the difficult stuff.

[ Parent | on HN ]

Re:Blender 2.79: OpenCL on par with CUDA by Athas on Wednesday April 12, 2017 @12:48PM (#14099419)

Out of curiosity, what are you doing with Accelerate? I've been looking for examples of people using it in anger.

[ Parent | on HN ]

Re:Blender 2.79: OpenCL on par with CUDA by tombert on Wednesday April 12, 2017 @01:04PM (#14099563)

I can't get into the weeds of it because of some NDA chicanery, but there are a couple vector transformations that I have to do on a not-quite-big-data-but-also-too-big-for-regular-CPU set of data. The initial version of the project was doing some basic stuff in Python, but that proved to be slow so I rewrote most of it in Haskell and Accelerate.

[ Parent | on HN ]

Re:Blender 2.79: OpenCL on par with CUDA by rodburns on Monday April 17, 2017 @06:36AM (#14130142)

In addition to triSYCL we also have an implementation of SYCL called ComputeCpp http://computecpp.codeplay.com/

[ Parent | on HN ]