Commits on Apr 10, 2024

1. Return device total global memory for MaxAllocSize …

   The SYCL specifiction inherits an OpenCL query for the maximum single
   allocaton that can be made, and requires that at least one-quarter of
   the device's reported global memory size can be allocated.
   
   CUDA doesn't really have such a limitation as far as I can tell, and
   will happily allocate anything up to the total size of global memory
   in one go. This means returning that size is a reasonable guess for
   MaxAllocSize.

   

   DuncanMcBain authored and kbenzie committed Apr 10, 2024
   Configuration menu

   • View commit details
   • Copy full SHA for c02d137
   • Browse repository at this point

   Copy the full SHA

   c02d137 View commit details

   Browse the repository at this point in the history