Re: Remote fork() and Parallel programming

Mathieu Bouchard (boum01@UQAH.UQuebec.CA)
Tue, 16 Jun 1998 19:03:03 -0400 (EDT)


> > If you make a program waste 500k, and 100,000 users do use it, it's 50
> > gigs of harddisk wasted, which is worth several thousand bucks. From a
> > holistic point of view, if you are worth 50$/hour and it takes you less
> > than {several thousand bucks}/50 hours to do the optimization, well, why
> > not?
> Of course. But what if the optimization saves only 10K? Now you're saving 1G of
> disk space, which is worth maybe $30.

$30 ? or 100?... oh well. it depends on whether you buy a small or a
large HD...

> $500-$1000. And disk space has not declined in cost as quickly as CPU
> cycles and proably RAM.

Hey, a good reason to use compressed filesystems.

> To ignore the extent to which declining hardward costs have made many
> optimizations uneconomical (and many inefficient implementation techniques
> practical) is silly.

I am resigned to understanding that. Everything that is an added constant
in a number we try to minimize loses importance with time. But if
something makes your program twice as slow, this might be more annoying,
though not as annoying as exponential computation time (O(n*n)) :-)

> However, I do agree that optimization can be fun, and I also recognize that
> programming is to a certain extent an art, where compactness and efficiency is a
> form of beauty which has value beyond its economic savings.

In how many years Future Crew's obsolete asm code will be sold around at
prices rivalling the paintings of Van Gogh? :-)

just to keep the bandwidth low, no need to reply to this :-)

I'll keep my e-saliva for something more important... I have a bug report
to submit today or tomorrow.

matju

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu