|
inlineprivate |
Helper function, obtains a block of memory of size sz
from CUDA on device device
.
If CUDA is disabled, uses standard l_malloc().
[in] | sz | size of the requested allocation in bytes |
[in] | device | device on which the allocation should take place |
device
or nullptr
if the allocation failed. References cuda_handle_error, syten::l_malloc(), syten::Cuda::select_device(), and syten::Cuda::synchronise().
Referenced by alloc(), and get_from_upper().