From: Jesse B. <jb...@sg...> - 2002-02-22 19:14:08
|
On Fri, Feb 22, 2002 at 10:56:06AM -0800, Mike Kravetz wrote: > Below is preliminary patch to implement some form of NUMA scheduling > on top of Ingo's K3 scheduler patch for 2.4.17. This is VERY early > code and brings up some issues that need to be discussed/explored in > more detail. This patch was created to form a basis for discussion, > rather than as a solution. The patch was created for the i386 based > NUMA system I have access to. It will not work on other architectures. > However, the only architecture specific code is a call to initialize > some of the NUMA specific scheduling data structures. Therefore, it > should be trivial to port. Ah, you beat me to it; I was working on code very similar to this when I got your e-mail. I think this sort of thing will address the problem we've been seeing on machines with more than 16 cpus or so (IDLE_REBALANCE_TICK is too small, flooding CPUs with load_balance requests), as well as making numa scheduling a little more efficient. I'll see if I can make it work on our platform and let you know how it goes. Jesse |