JuliaGPU/Metal.jl

Port the opportunistic synchronization from CUDA.jl

Opened this issue · 1 comments

It would probably be good to port the opportunistic synchronization from CUDA.jl

Originally posted by @maleadt in #314 (comment)

EDIT by maleadt: Specifically, the tracking of which stream has recently modified a buffer, and whether to synchronize or not.

In addition, copyto! should be asynchronous when possible, e.g., when copying between MTLArrays. As suggested by @tgymnich, this could probably be done by always using the MTLBlitCommandEncoder.