Performance and/or accuracy of C compilers

James Kanze <kanze@us-es.sel.de>, writes:
> > [In the PC tools biz, there seems to be a large market for
> > compilers that produce very fast but wrong code. Don't ask me
> > why. -John]
>
> Actually, the market is for compilers which produce very fast
> code for a number of small benchmarks. I had the occasion to
> evalute a couple of C compilers a few years ago, and found
> that with one of the best selling compilers at the time,
> turning on optimization caused the programs to be bigger and
> run slower, in all compilation models except small. What real
> application will run in small model?


I'll take the bait on this one. Not everyone is writing applications that
use environments that take 4 Meg of memory and 50 meg of disk space to
display "Hello World". :)


MSDOS/PCDOS Device Drivers have to fit into small model unless they've
changed things.


Also, whether the people who are writing heavy and bloated applications
that violate every standard of decency :) want to admit it, if more
applications had been given a little thought into being more careful with
resources, we'd be able to run more programs without a kluge like
Microsloth Windoze or Half an Operating System. :)


The original Turbo Pascal - editor AND compiler - fit into less than 40K
of disk space, 128K of memory and for what it did, was pretty good. Or at
least, (to counter some comments) I didn't have any problems or notice any
errors. The latest monstrosity takes 20+ meg of disk space and a 386.


C Compilers are about as bad, too.


There is something to be said for making programs that will work under
less than optimal conditions. Another thought is the palmtop and handheld
computers have 640K. A small memory model program will fit on one of
these; Windows won't.
--
Paul Robinson - TDARCOS@MCIMAIL.COM
--

Return to the comp.compilers page.
Search the comp.compilers archives again.