Date: Wed, 10 Sep 1997 19:22:24 +0300 (IDT)
From: Eli Zaretskii <eliz AT is DOT elta DOT co DOT il>
To: Vik Heyndrickx <Vik DOT Heyndrickx AT rug DOT ac DOT be>
cc: DJ Delorie <dj AT delorie DOT com>, djgpp-workers AT delorie DOT com
Subject: Re: fread slowstart
In-Reply-To: <3416C026.132B@rug.ac.be>
Message-ID: <Pine.SUN.3.91.970910191723.10908E-100000@is>
MIME-Version: 1.0
Content-Type: TEXT/PLAIN; charset=US-ASCII
Precedence: bulk


On Wed, 10 Sep 1997, Vik Heyndrickx wrote:

> DOS internally reads clusters at a time.

Are you sure?  Can you prove it (with a program) or point me to a source
of information that says so?  I always thought DOS reads in 512-byte
chunks. 

> Since the maximum cluster size is 16384, I think,

Partitions larger than 1GB get 32KB clusters.  One of my machines has 
such a partition.  I think the maximum cluster DOS can manage is 64KB.

> and we don't want DOS buffer our data the TB size
> should be a multiple of 16384 (2x or 3x) and any read should concur with
> this limit. e.g. reading 32 times 512 bytes and then switch to 16384
> reads.
> 
> Or reading 16 times 512 bytes = 8192
>             8       1024      = 8192
>            -------------------------
>                                16384
> 
> Or reading 8 times 512, 4 times 1024, 2 times 2048, once 4096 and then
> in chunks of 16384 ?

Seems too complicated, if you ask me.  I still think that we should 
measure the effects of different schemes and decide then what's the best 
compromise.  It doesn't make sense to make code complicated for marginal 
speedups.