6 Sep
2003
6 Sep
'03
10:17 p.m.
One particular issue that was mentioned on those pages
was that timing
is not very accurate (dependent on HZ?) in 2.4 kernels, with great
improvments in 2.5/2.6 revs.
timing and scheduling are quite separate and independent. that is,
asking "what time is it now?" is extremely accurate, with resolution
determined by your CPU clock speed. HZ plays no role in this.
scheduling resolution (i.e. getting something to happen at time T) is
limited by HZ and/or some other interrupt source. many kernels (even
2.4) now have HZ=1000, and in 2.5/2.6, the dependence on HZ is removed
completely - the kernel has microsecond-level scheduling
capabilities. i have yet to test any of this myself, but everything
i've read looks very very good so far.
--p