Re: Tips for code porting to MC

2002-10-29 Thread Dar Scott

On Tuesday, October 29, 2002, at 08:18 AM, [EMAIL PROTECTED] 
wrote:

Im looking for tips regarding porting variable redirection (pointers and
pointers to pointers)
and c-style struct variables.


I wouldn't ignore arrays, but I have been considering substructure in 
strings.  MC stings are byte strings and it might be that same-size char 
chunk replacements are efficient.  If so, you can create a data space 
within a string in a global variable.  I'd run the experiment, anyway.

Or you might try a mix in which some data are packed in strings and 
others are not, in which some pointers are implemented as arrays and 
some are implemented as string offsets.  This might be better if strings 
are always rebuilt.

Dar Scott

___
metacard mailing list
[EMAIL PROTECTED]
http://lists.runrev.com/mailman/listinfo/metacard


Re: Tips for code porting to MC

2002-10-29 Thread Pierre Sahores
[EMAIL PROTECTED] wrote:
 
 Hi Guys,
 
 Im porting a rather large neural network (NN) program written in C++ to
 Metacard to do some stats...
 
 Im looking for tips regarding porting variable redirection (pointers and
 pointers to pointers)
 and c-style struct variables. The key to porting these is speed as NN
 require speed and ease of
 scripting as MC is all about that!
 
  for example:
   how do you port x[a][b] = item b of line a of x
   But this proves extremely slow once you reach a1000...
 
   what about
   p-x-b[i]
 
   so far the best I found (without much searching I admit) is
   p_x_b[i]
 
  but what of
 p-x[a]-b
 
 Im looking for any good schemes you may propose!
 
 Thanks in advance for any help!
 Xavier
 http://monsieurx.com - incoming metacard section... my site is back! ;))
 
 Visit us at http://www.clearstream.com


Hi, 

1.- Use extensively the read-only repeat for each... loop structure.
2.- Use extensively the array vars instead of item-lines tables vars
to store the datas you are searching on, sorting,...
3.- Use B-Tree alike data storage structures if the datas you have to
parse at once exceed 500 ko.
4.- Run your app under Linux (extensively tested as two time faster than
both Windows or MacOSX).

-- 
Cordialement, Pierre Sahores

Inspection académique de Seine-Saint-Denis.
Applications et bases de données WEB et VPN
Qualifier et produire l'avantage compétitif
___
metacard mailing list
[EMAIL PROTECTED]
http://lists.runrev.com/mailman/listinfo/metacard