Roland Mainz wrote: > Jonathan Chew wrote: >> Rafael Vanoni wrote: >>> Hey everyone >>> >>> Is the kernel aware of the status of the interconnect between different >>> NUMA nodes ? >> No, not currently. It just assumes that there is some interconnect >> between the nodes and may know the latency between them when the system >> is not loaded. > > Does that mean this latency value is "static" (e.g. won't change during > lifetime) ?
The latency values are determined during boot, and reconfigured in the event of a dynamic reconfiguration. > Would it be possible to make it "dynamic", e.g. update the > latency value based on both real latency and link saturation (or maybe > better wrap the read accesses to the latency value into a CPP macro to > have a centralised way to make dynamic calculations without having to > change lots of places in the codebase later) ? As Jonathan said previously on this thread, there's currently no support for link saturation and it is something of concern. > BTW: How does the topology code handle DR ? Take a look at http://src.opensolaris.org/source/xref/onnv/onnv-gate/usr/src/uts/common/os/lgrp.c#522. That's the platform independent portion of the code called when there's a DR event. -- Rafael Vanoni [EMAIL PROTECTED] http://blogs.sun.com/rv _______________________________________________ perf-discuss mailing list perf-discuss@opensolaris.org
