Date: Tue, 9 Sep 1997 14:37:22 +0300 (IDT)
From: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
To: DJ Delorie <dj AT delorie DOT com>
cc: djgpp-workers AT delorie DOT com, Charles Sandmann <sandmann AT clio DOT rice DOT edu>
Subject: Re: fread slowstart
In-Reply-To: <199709090234.WAA00084@delorie.com>
Message-ID: <Pine.SUN.3.91.970909143509.7143M-100000@is>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Precedence: bulk


On Mon, 8 Sep 1997, DJ Delorie wrote:

> Also, I've changed the default transfer buffer size to 32k

Do we have any real reason for this?  It makes less DOS memory
available for subprocesses, which means less nested programs.  I
wouldn't recommend trading 16K per process for marginal speedups.  (If
they aren't marginal, let's hear the numbers.)  I have seen programs 
nested to 6 levels when building some complex packages (such as TeX).

> and stack size to 512k in stub.asm.

I agree to this one.  Maybe even 768k (some C++ programs will still
bomb cc1plus with 512k).

And while at that: I think it's a good idea to have Charles' stackrun
program in the official toolset.  I find myself using it from time to
time, and when I need it, it's invaluable.  (For those who don't know:
stackrun will run a COFF image and report the maximum stack usage when
the program exits.)

> Should we default to a max size (63.5K) transfer buffer now ?

Definitely not.