When CPU_TIME is enabled Cholla now prints out how much GPU memory is allocated at the end of each time step. To do this I added new functions, Reduce_size_t_Max an Print_GPU_Memory_Usage, to determine the maximum amount of memory used on any GPU and print that information out. The motivation was to determine how many cells we can fit on a given GPU with a given Cholla configuration easily. Early tests show that on Frontier we can fit about 457^3 cells per GPU in an MHD only build.
Summary
When
CPU_TIME
is enabled Cholla now prints out how much GPU memory is allocated at the end of each time step. To do this I added new functions,Reduce_size_t_Max
anPrint_GPU_Memory_Usage
, to determine the maximum amount of memory used on any GPU and print that information out. The motivation was to determine how many cells we can fit on a given GPU with a given Cholla configuration easily. Early tests show that on Frontier we can fit about 457^3 cells per GPU in an MHD only build.