LibRapid / librapid

A highly optimised C++ library for mathematical applications and neural networks.
http://librapid.rtfd.io
MIT License
164 stars 10 forks source link

OpenCL GeneralArrayView #253

Open Pencilcaseman opened 1 year ago

Pencilcaseman commented 1 year ago

245

The current GeneralArrayView assignment implementation does not support OpenCL or CUDA. To implement this, the dataSourceExtractor implementation must return a struct containing the view's data, offset and stride. The OpenCL/CUDA kernels must then use this information to access the corresponding index in the data. This will require #245 since we need to runtime-generate the kernels to avoid having too many of them being compiled at once.

This also restricts the OpenCL/CUDA support in Python