[mpich-discuss] mpi_gather slow on a single node

[Keith.Bannister at csiro.au]

On 9 Dec 2016, at 9:38 am, Keith Bannister <ban115 at csiro.au> wrote:

> Moreover, that latency should be much smaller if it is using shared memory. If I understand it right, the average throughput = message_size/latency = 6.5 Gbits/sec, which resembles a network throughput to me. I’m sure the memory bus on this machine can sustain much more than that.

Ah, my mistake on this point. Looking at the osu_gather.c code, the correct calculation should be throughput = message_size * nranks / latency = 78 Gbits/sec. So the throughput is OK, but the latency variability is still pretty wild.

K




_______________________________________________
discuss mailing list     discuss at mpich.org
To manage subscription options or unsubscribe:
https://lists.mpich.org/mailman/listinfo/discuss