[Top][All Lists]

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: CPU affinity patch for 'nice'

From: Paul Eggert
Subject: Re: CPU affinity patch for 'nice'
Date: Tue, 21 Sep 2004 23:29:28 -0700
User-agent: Gnus/5.1006 (Gnus v5.10.6) Emacs/21.3 (gnu/linux)

A few thoughts about that nice fix.  First, the implementation appears
to be limited to 32 processors, and there should not be arbitrary
limits in GNU software.  Second, I'm not sure the user interface is
intuitive (though perhaps you could give some usage examples to
explain what's going on).  Third, it'd need user documentation (e.g.,
doc/coreutils.texi, "nice --help").  Fourth, I'd like to see the code
written generically, so that it compiles on older Linuxes without any
problem (e.g., no magic _syscall3 functions; just call
sched_setaffinity if it exists when configuring, and don't call it if
it doesn't.)  Fifth, we should port the same UI to Solaris and/or
POSIX, to make sure we're not committing to a not-portable-enough UI

Last but not least, is there prior art for this sort of thing?
Anybody else got a shell interface for simple processor-scheduling?
Would rather not reinvent the wheel.....

Whew!  Enough comments for now.  The bottom line is that I like the
basic idea, but it needs some polish.

reply via email to

[Prev in Thread] Current Thread [Next in Thread]