Use wrappers for the GPU buffer copy/allocations - Redmine #3318
All device buffer operations should be done via the wrapper functions. This is to ensure that the code can be reused between CUDA/OpenCL/SYCL. Also, the wrappers have the consistency checks, which may be helpful while debugging.
(from redmine: issue id 3318, created on 2020-01-22 by artemzhmurov)
- Relations:
- child #3319 (closed)
- child #3320 (closed)
- child #3321 (closed)
- parent #3311 (closed)
Edited by Artem Zhmurov