Message-Id:	<199803171616.SAA29175@ankara.duzen.com.tr>
From:	"S. M. Halloran" <mitch AT duzen DOT com DOT tr>
Organization:	User RFC 822- and 1123-compliant
To:	djgpp AT delorie DOT com
Date:	Tue, 17 Mar 1998 18:20:31 +0200
MIME-Version:	1.0
Subject:	Re: Q: Compile speed increase from 4MB to 8MB RAM
In-reply-to:	<350E6104.D67B4D8C@cornell.edu>

On 17 Mar 98 A. Sinan Unur was found to have commented thusly:

> David Vrabel wrote:
> > 
> > I currently have a 386sx20Mhz (Don't laugh I can't afford anything
> > better) with 4MB of RAM.  I have configured the computer according to
> > the instructions in the FAQ.  I am using DJGPP v2.01 and GCC 2.8.0.
> > 
> > The problem is the incredibly long comple times (about 20 minutes)
> > particularly when using any part of the STL (ie. lots of templates).
> > 
> > How much difference would an extra 4MB of RAM make to the compile
> > speed? (Actual figures would be nice if possible)
> 
> i can't give you exact figures, but i can only tell you that it would be
> worth it. i would give smartdrv 1 MB cache, as opposed to the default 2
> MB. (you don't mention which version of dos you are using and there are
> performance differences between the various versions of smartdrv.) i
> would also either not use emm386 at all, or at the very least, use it
> with the noems option. totally off the top of my head, i think you
> should see a two or three-fold improvement.
> 
> also, how fragmented is your hard disk and is there enough room on it?
> if most of that time is spent with the heads moving all over the place,
> you may not get too much of an improvement.
> -- 

None of this really helps with the really big speed improvements which I 
believe he wants.  I have a machine at home 486DX (33 MHz) w/ 4 MB RAM, 
with virtually no drivers (other than xms--I can print out the config.sys 
and autoexec.bat if you want to take a look), and I typically defrag it to 
get as much VM as possible (I am also tight for disk space on this museum 
piece, with a 250 MB, half of it compressed and containing the djgpp 
distribution).  All compiling, linking, and temporary work is done from 
the uncompressed part of the disk, although clearly the executables are 
read in from the compressed part, and there is at least 30 MB of free, 
unfragmented disk space.

When developing I have the debug option '-g' set but no optimizing!  With 
source modules of only one function and maybe a few lines, it takes about 
15-20 seconds from the time the program starts thinking about how to pass 
the compile command line to finishing the compile.  As the number of lines 
and complexity of the code seems to increase, there appears to be an 
exponential rise (I probably exaggerate just a little).

It is my own opinion, and stressed in the djgpp introduction/faq, that 
the more RAM (physical memory) you can give--never mind processor 
clocking, although that certainly helps too--the more you see your compile 
times go down.  And you probably could not set the options for optimizing, 
since your machine might burp or just hang (although I haven't tried).

I think if you want to use this development system, you might really want 
to find the money for a system upgrade that gives you at least 16 MB 
RAM --more than that would be for your programs requiring huge arrays, not 
for development on djgpp.

Mitch Halloran
Research (Bio)chemist
Duzen Laboratories Group
Ankara   TURKEY
mitch AT duzen DOT com DOT tr

other job title:  Sequoia's (dob 12-20-95) daddy

- Raw text -