home bbs files messages ]

Forums before death by AOL, social media and spammers... "We can't have nice things"

   comp.lang.pascal.borland      Borland Pascal was actually pretty neat      2,978 messages   

[   << oldest   |   < older   |   list   |   newer >   |   newest >>   ]

   Message 1,013 of 2,978   
   Marco van de Voort to Glenn Someone   
   Re: FPC Questions   
   30 Sep 04 11:12:54   
   
   From: marcov@stack.nl   
      
   On 2004-09-30, Glenn Someone    
   wrote:   
   > number of records to read in specified as a constant, so we have a   
   > perfect testbed (change constant, recompile, run), especially since   
   > this program reads in 2000000 4 byte records in its processing time   
   > (and I don't have to go to any work to make the testbed so).   
   >   
   > These times in Delphi 3:   
   >   
   > records read in at a time in the program / time program ran. (run in   
   > "D" mode)   
   >   
   > 1 record at a time = 22.30 s   
   > 2048 records at a time = 16.64 s   
   > 16384 records at a time = 12.92 s   
      
   Plotting this gives you an optimum, which is typically 64kb or in extreme   
   cases 256kb. No need for hundreds of MBs, unless it really is totally random.   
      
      
      
   > I'm not surprised since I ran this test long ago when I posted that   
   > program, in fact I'm not making contentions out of thin air here.  But   
   > if I was surprised at anything, I'm surprised at how slow reading one   
   > record at a time really is.  Trust Windows disk buffering at the cost   
   > of 10 seconds processing time?  No way.   
   >   
   > Am I making much of nothing at all.  Try doing a rewrite of that   
   > program you mentioned and see whether you get a similar change,   
   > assuming the function of the program suits it.   
   >   
   > And let's just say the program I'm working on as a target is much more   
   > complex than the simple sequential file I/O that this decimal testbed   
   > does.   
   >   
   > Of course, I'm willing to hear a good explanation of what I'm seeing.   
   >   
   > On Mon, 27 Sep 2004 16:19:06 -0400, L D Blake  wrote:   
   >   
   >>You mentioned that you can make your program work in about 76k of dynamic   
   >>variable space... try this...  Trust the Windows disk buffering.  Go record   
   by   
   >>record, and see what happens.  I'm thinking you'll be pleasantly surprised.   
   >   
   >>Methinks, yer making rather much of nothing at all.   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)   

[   << oldest   |   < older   |   list   |   newer >   |   newest >>   ]


(c) 1994,  bbs@darkrealms.ca