One problem with this is that most vendors of kernels ship large complex
OS's that will reliably have blasted the last task out of the caches by
the time they lumber back around to running it again. In Linux thats not
always the case (and with big per cpu l2 caches I submit thats not the
case for most OS's now). We put a small damping on tasks bouncing across cpus,
just enough to stop syscall patterns causing them to skip around the processors.
It definitely helped on PPro machines, it only showed up for some use patterns
(2 tasks lots of syscalls) on pentiums (shared L2, private L1)
Remember in 1992 128K of cache was a lot. Now we have 1Mb routinely, 2Mb
Xeons and I believe even more L2/L3 than that on the next generation SMP
Alphas.
Alan
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/