In my asynchronous implementation ( https://github.com/e-ago/CoMD-CUDA-Async ) using runs with parameters like "-e -x 80 -y 80 -z 80" with up to 16 process I noticed that after the first atom exchange function, the size of the communication buffers in the force exchange function never change; however the size is calculated again during each force exchange function call.
Therefore, after the first atom exchange, I always used the same buffer size and without having any crash.
Can this improvement be used in the most general case? (i.e. all the type of executions?)
In my asynchronous implementation ( https://github.com/e-ago/CoMD-CUDA-Async ) using runs with parameters like "-e -x 80 -y 80 -z 80" with up to 16 process I noticed that after the first atom exchange function, the size of the communication buffers in the force exchange function never change; however the size is calculated again during each force exchange function call.
Therefore, after the first atom exchange, I always used the same buffer size and without having any crash.
Can this improvement be used in the most general case? (i.e. all the type of executions?)