[zeromq-dev] Potential timing issues on certain Xeon large multi-socket environments

[Steven McCoy]

I just saw this question on Stack Exchange which shows more enlightening
caveats with multi-core environments and the wonderful time stamp counter:

http://serverfault.com/q/474151/42385


This would clearly be erroneous with the default code path in OpenPGM on
Linux unless running with RDTSC_FREQUENCY override:

https://code.google.com/p/openpgm/source/browse/trunk/openpgm/pgm/time.c#375


Of interest would be how other platforms would function as only Windows
appears to have support for a per-CPU clock rate.

-- 
Steve-o
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.zeromq.org/pipermail/zeromq-dev/attachments/20130131/1ae91d14/attachment.htm>