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 998 of 2,978   
   L D Blake to All   
   Re: FPC Questions   
   27 Sep 04 17:42:14   
   
   From: not@any.adr   
      
   On Mon, 27 Sep 2004 21:34:05 +0000 (UTC), Marco van de Voort    
   wrote:   
      
   >On 2004-09-27, L D Blake  wrote:   
   >>>to work with so I can figure out a happy medium between what would   
   >>>work in Windows and what would work best for the application.   
   >>   
   >> And therein lies your mistake!  You are trying to acquire and keep a fixed   
   >> amount of memory for your program to work in... This simply is *not*   
   necessary   
   >> anymore.  In fact, for large chunks of memory it's a *bad* thing...   
   >>   
   >> Where the crossed wires are coming from is that you seem stuck on this idea   
   >> that you have to somehow magically gain ownership of a big chunk of memory   
   >> when we are telling you this is no longer good practice.   
   >   
   >> From your last set of descriptions I get the impression you are trying to   
   >> buffer some huge number of data records in memory because you believe that   
   >> writing and fetching from disk is painfully slow.   
   >   
   >Hmm, I don't agree. Mem transfer and CPU speed also increased. The ratio is   
   better,   
   >but not much.   
      
   If city traffic moves at 1/2 country traffic speeds... spiking up country   
   traffic speeds by a factor of 6 (about hard disks are up by) is going to   
   result in one damned fast drive through the country... no matter what happens   
   to the city is irrelevant.   
      
      
   >However I miss "cache utilization" in this story.  Often people try to max   
   >the cache unnecessarily, since a small part already gives 95% of the   
   >performance advantage.   
      
   Exactly... 9 of 10 times the data requested is not already in the cache. Oddly   
   enough the one area where caches are of the least benefit (sequential access)   
   is also where they get the best results.  Caches can actually slow you down as   
   the drive is still thrashing about in fragmented files, missed hits,   
   incomplete data blocks and files that aren't even opened yet.   
      
      
   -----   
   Laura   
      
   (http://www.start.ca/users/ldblake)   
      
   --- 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