X-Authentication-Warning: delorie.com: mail set sender to djgpp-bounces using -f
X-Trace-PostClient-IP: 68.147.177.20
From: Brian Inglis <Brian DOT Inglis AT SystematicSW DOT Invalid>
Newsgroups: comp.os.msdos.djgpp
Subject: Re: DREADED fseek
Organization: Systematic Software
Message-ID: <g1bbt0990aj8g8eqdei7046v8fmpm4a0ts@4ax.com>
References: <8edus09uvkr9or13giuvdflk4ogin5q2l1 AT 4ax DOT com> <01c4ebed$Blat.v2.2.2$c5f598c0 AT zahav DOT net DOT il> <dl52t0tp5o1o6e5m6savh94u3vijmuq43r AT 4ax DOT com> <200412281245 DOT iBSCjQRj001350 AT envy DOT delorie DOT com> <72v3t01u5oeu89j63pfeece2ip780hsb45 AT 4ax DOT com> <01c4ed60$Blat.v2.2.2$fab621a0 AT zahav DOT net DOT il> <lhk4t0lng0g7ca3fc48qt6i7jbv1is4ksh AT 4ax DOT com> <01c4ee2a$Blat.v2.2.2$0f6c2800 AT zahav DOT net DOT il> <qip7t0pp2a6lbcndsjs9m1g4tv2ekfas5m AT 4ax DOT com> <01c4eea1$Blat.v2.2.2$74816940 AT zahav DOT net DOT il>
X-Newsreader: Forte Agent 1.93/32.576 English (American)
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Lines: 50
Date: Fri, 31 Dec 2004 20:35:25 GMT
NNTP-Posting-Host: 24.71.223.147
X-Complaints-To: abuse AT shaw DOT ca
X-Trace: pd7tw3no 1104525325 24.71.223.147 (Fri, 31 Dec 2004 13:35:25 MST)
NNTP-Posting-Date: Fri, 31 Dec 2004 13:35:25 MST
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Reply-To: djgpp AT delorie DOT com

On Thu, 30 Dec 2004 20:56:45 +0200 in comp.os.msdos.djgpp, "Eli Zaretskii" <eliz AT gnu DOT org>
wrote:

>> From: Brian Inglis <Brian DOT Inglis AT SystematicSW DOT Invalid>
>> Date: Thu, 30 Dec 2004 11:32:27 GMT
>> 
>> >> Saw that: looks like it just flushes buffered writes, not all buffered
>> >> data read. 
>> >
>> >On what OS?
>> 
>> In flushdc.{c,txh} and RBIL. 
>> None of them say that the list of all cached blocks is cleared,
>
>Yes, but did you actually try that, or did you only look at the
>sources?  If you did try it and it didn't remove the read data, please
>tell on what OS was that.
>
>Note that the system call used by the function not only flushes the
>cache, but also resets the disk drive, which should do what you want
>(and at least on plain DOS, it actually did).
>
>> or that caching is disabled.
>
>The function doesn't disable caching, but if you invoke it after every
>test, you should see the behavior of uncached data, i.e. raw disk I/O
>performance.  Isn't that what you wanted?

Good call, Eli. 
Calling it before each open, I do see uncached behaviour.
Not what I would call *raw* disk I/O performance: 
10MB/s in a VDM!
But that's exactly what I wanted. 
Typical results (including open and close) are:

                         __ default __  __ setvbuf __
File   Size Buffer Reads Time Transfer  Time Transfer Faster
       (MB)   (B)	 (s)	(MB/s)  (s)    (MB/s)
PD2.DLL  17  64K    278  1.813  9.574	1.762  9.853  setvbuf
PD2.DLL	 17  32K    556  1.835  9.457	1.792  9.684  setvbuf
PD2.DLL	 17  16K   1111  1.802  9.632	1.791  9.692  setvbuf
PD2.DLL	 17   8K   2222  1.836  9.452	1.747  9.935  setvbuf
PD2.DLL	 17   4K   4444  1.809  9.597	1.811  9.585  default
PD2.DLL	 17   2K   8887  1.810  9.587	1.814  9.570  default
PD2.DLL	 17   1K  17774  1.753  9.904	1.802  9.630  default
PD2.DLL	 17  512  35548  1.782  9.740	1.788  9.705  default
PD2.DLL	 17  256  71095  1.795  9.667	1.813  9.573  default

-- 
Thanks. Take care, Brian Inglis