Open jlowe opened 1 year ago
Relates to rapidsai/cudf#13814. We should consider leveraging cudaHostRegister to accelerate data transfers to the GPU on platforms where cudaHostRegister is relatively cheap.
This will accelerate places where we cannot get pinned memory.
Relates to rapidsai/cudf#13814. We should consider leveraging cudaHostRegister to accelerate data transfers to the GPU on platforms where cudaHostRegister is relatively cheap.