Sean Hefty wrote:
Here is version 2 of the patch. Based on observations of tests, I believe
Steve Wise's comments are reasonable, so I removed the rdma_resolve_addr
retry and simply changed the timeout value. Feel free to use whichever one
of these patches you like best. However, I urge you to apply one of these,
since the programs fail in a busy large fabric.
Why not just make the retry and timeout values command line parameters and allow
adjusting both?
- Sean
I'm not sure we need the retry.
But making the timeout a parameter is good idea.
And fix all the rdmacm programs in perftest (I think there are two: bw
and lat).
Steve.
_______________________________________________
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
