Feature Description
OpenCL sucks.
We need a more robust, scalable, portable and sustainable approach to the GPU ecosystem, possibly with native support for different backends.
Implementation Ideas
In the spare time I'm working on a unified bridge between CUDA, Metal and OpenCL, allowing for the same code to execute on different backends, with minor changes to the kernels.
Feature Description
OpenCL sucks.
We need a more robust, scalable, portable and sustainable approach to the GPU ecosystem, possibly with native support for different backends.
Implementation Ideas
In the spare time I'm working on a unified bridge between CUDA, Metal and OpenCL, allowing for the same code to execute on different backends, with minor changes to the kernels.