Four examples:
-
vec-kernel.cl: two very simple OpenCL kernels, used for the Atomese examples, below. These implement vector addition and vector multiplicattion. -
atomese-kernel.scmdemonstrates how to use Atomese to open a channel to an OpenCL device, and then send a compute kernel and some floating- point vector data to it. The results from the computation are pulled back into the AtomSpace. The data is operated on by the kernels defined invec-kernel.cl. -
accumulator.scmdemonstrates how to define a location in the AtomSpace which shadows a vector held on the GPU. That is, the vector "lives" on the GPU, and is downloaded back to to the system when it is "examined", e.g. to be printed to stdout. -
dot-product-bad.scmunder development; eventually meant to be a "realistic" example of a dot product. Doesn't work right now.