On Mon, 14 Jun 1999, Robert M. Hyatt wrote:
> Linux sees "processes" and nothing else. And the scheduler says "I
> want something to run, so it looks for processes that are not blocked,
> and grabs the one it thinks is most reasonable (ie highest priority,
> recently ran on this processor [to improve cache hits], etc.) But
> threads bounce around the processors pretty wildly if you have more
> threads than processors (or even if you have the same number but you
> are getting lots of interrupts).
the more correct term to use: "Linux only sees threads". The basic
scheduling entity for Linux is a thread. Threads have resources attached
to them. If the 'VM' resource happens to be shared between two Linux
threads then we get what is often referred to as 'kernel mode threads'. If
a Linux thread has only private resources then it's what is usually called
a 'process'. But this is all irrelevant to the scheduler - it sees
threads. (and uses certain heuristics, eg. there is a small plus for
scheduling two threads who happen to share the same VM) Also, the speed
difference between scheduling 'kernel mode threads' and 'processes' is
almost invisible.
-- mingo
-
Linux SMP list: FIRST see FAQ at http://www.irisa.fr/prive/mentre/smp-faq/
To Unsubscribe: send "unsubscribe linux-smp" to [EMAIL PROTECTED]
