Skip to content

VTK-m can now defer de-allocations of CUDA memory.

Robert Maynard requested to merge robertmaynard/vtk-m:deferred_cuda_free into master

This is done as cudaFree blocks execution on all cuda streams. When doing operations such as moving virtuals to CUDA memory space we want to defer the cuda free as long as possible to not block other stream execution.

Merge request reports