Or Gerlitz wrote:
OK,its always good to start with facts on the ground... before commiting this test, my original thinking was that for messages whose size=X is less then the IB Link level MTU it holds that:latency(X,UD) <= latency(X,UC) <= latency(X,RC) Running the latency test provided with the perftest package on my systems (*) I get the below results. Does anyone has insight why the --minimal-- and typical UD latency is 1us ( = 20%) worse then the --minimal-- and typical RC latency??? Or.
In all devices that support the memfree architecture (InfiniHost III-Ex running at memfree mode, InfiniHost III Lx and ConnectX) you will find the performance of UD comparable to the RC/UC. With the introduction of the memfree architecture, we are focused on development and performance optimization for this architecture. That is the main reason for memfree to achieve better performance.
-Dror _______________________________________________ general mailing list [email protected] http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
