Date: Sun, 15 Jun 1997 12:02:08 +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
Subject: Re: size of stdio buffer
In-Reply-To: <199706122259.SAA21535@delorie.com>
Message-ID: <Pine.SUN.3.91.970615120131.14561K-100000@is>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Precedence: bulk


On Thu, 12 Jun 1997, DJ Delorie wrote:

> Would it make sense to have stdio's buffer fill dynamic - start at 512
> bytes, and double with each read() (up to BUFSIZ) but reset on each
> lseek() ?

Can somebody please time the current stdio functions on different setups
(local vs networked drives, compressed vs uncompressed drives, disk cache
vs vanilla DOS, etc.) and see if such a change will indeed make it
significantly faster?  I'm worried about making a change that will
complicate the code without knowing what we'd get in return.  The
explanation that Alexander gave makes a lot of sense to me, but I have
seen too many cases where an obvious explanation has proven to be wrong.