Message-Id: <199706051442.QAA28972@zeus.hu.bonus.com>
From: "Gyorgy Abonyi" <loop AT hu DOT bonus DOT com>
To: "Eli Zaretskii" <eliz AT is DOT elta DOT co DOT il>
Cc: <djgpp AT delorie DOT com>
Subject: Re: Speed of DJGPP?
Date: Thu, 5 Jun 1997 16:41:56 +0200
MIME-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Precedence: bulk

 On Thu, 5 Jun 1997, I wrote it:
 
 > This's becouse the DOS4GW extender what watcom uses, replaces the most
real
 > mode DOS functions, with it's own 32-bit versions, so with Watcom the
file
 > I/O operations don't need 32/16 bit mode switching, what is really
slow... 

Eli's answer:This is all incorrect...

You are right, most of my words was incorrect, ooops:(

I just checked it, and DOS4GW calls real mode DOS... Sorry for missinform
everybody..

But i still say that DJGPP uses a lot of file I/O  (cpp->cc1->as) and speed
up I/O (Memory, Cache & Co.) helps a lot. On my system (p133, 16Mbyte ram,
Win95), DJGPP usually spends equal time in CPP and AS than in CC1, so I
don't think that the optimizing stage is the most time consuming part of
DJGPP. Of course with larger modules (my C modules are usually sized btw
4-15 Kbytes ), gcc spent more and more time int CC1, but it's  significant
only if You got fast file I/O (fast harddrive and large cache).

BCC and Watcom usually don't use external C pre-processor, and assembler
module, so they don't need to pass a lot of an information via file I/O (I
know thay creates temp files, but not as much as gcc...).

And if You can access a U..ix version of gcc, than You can check that the
-pipe option speeds up really the compilation.

Gyorgy Abonyi Jr.
loop AT hu DOT bonus DOT com