Advanced HIP Workshop | Optimising compute with concurrent IO | Converting output to pinned memory

This video is part of the HIP workshop playlist. View full playlist here: • Advanced HIP Workshop
The 3D output array out_h on the host is allocated as pinned memory using hipHostMalloc to enable asynchronous copies. Silent failures were encountered when trying to use addresses within the pinned memory allocation as inputs to hipMemcpyAsync. Therefore I had to use hipMemcpy3DAsync to perform the copies asynchronously and pass in the host pointer returned by hipHostmalloc.