Mentions légales du service

Skip to content

mpi/examples/benchs/abstract_sendrecv_bench: add feature for allocating memory on cuda or HIP node. May be used to check GPUdirect support.

Matthieu Kuhn requested to merge satanas-atos/starpu:check_gpudirect into master

Enable the selection of CUDA memory node for the sendrecv_bench MPI benchmark.

Usefull to check StarPU's GPUDirect support / option for example.

Merge request reports