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

Re: IDL and 'nice' question

In article <onsngkaabp.fsf@cow.physics.wisc.edu>, Craig Markwardt <craigmnet@cow.physics.wisc.edu> writes:
> Randall Skelton <rhskelto@atm.ox.ac.uk> writes:
> > Hi all,
> > 
> > I have a question regarding setting the priority of IDL on a *nix
> > operating system.  There are certain instances when it is desirable to
> > set the priority of idl to a lower priority with the nice command.  Of
> > course, typing 'idl' at the command-line is actually a front-end to a shell
> > script and not an actual binary.  Are there any foreseeable problems in
> > starting the idl binary directly with 'nice -19 $IDL_DIR/bin/idl' as
> > opposed to staring the shell script?
> Before you go get yourself all twisted in a knot of DLMs, I think
> things are alot easier than you thing.
> First thing, I think you are confusing low and high priority.  For the
> non-unix among us, the "nice" command allows a user to set the process
> priority, which is essentially how much attention the CPU will give a
> program.  Running programs with low priority are readily bumped in
> favor of higher priority programs.  A *positive* nice number indicates
> a lower priority -- it is more "nice" to others; a negative nice
> number is a higher priority.  Thus your use of "-19" and "low
> priority" don't seem to be the right mix.

If you are a windows user I would recommend that you skip press the "next"
button, or whatever goes onto the next message in your newsreader.

Actually the value -19 is the lowest priority you can normally set with
nice. It's one of the quirks of the nice command that the argument looks
like a negative number but actually means increase the nice value by that 
amount. This *is* UNIX, you have to expect these idiosyncrasies.

Most versions of nice now accept a more reasonable argument "-n nice_value" 
which is unsigned for an increase in "niceness" and negative for a decrease.

> Second, I believe that a process's "nice" level is inherited by any
> subprocesses.  [ That has to be the case, otherwise a program could
> escape it's priority constraints by spawning a new copy of itself. ]
> So it shouldn't matter that the "idl" command is a script.

Indeed. In fact, the script exec's the IDL binary so it is actually the
same process with the same nice value.

Nigel Wade, System Administrator, Space Plasma Physics Group,
            University of Leicester, Leicester, LE1 7RH, UK 
E-mail :    nmw@ion.le.ac.uk 
Phone :     +44 (0)116 2523568, Fax : +44 (0)116 2523555