A new version of my EGTB generator

[guido]

I wanted to inform who could be interested that I put in my site (http://users.iol.it/rigel_g/) a new version of my generator gafs, where I eliminated some little errors and improved the program.
Even if gafs is in executable format only for 32 or 64 bits Unix/Linux OS, I put also an object file (gafslib.obj) that should permit the use of my TBs, once generated, from a chess program written under Windows. If someone will try to use this file, I would be interested for me to know if he finds problems to link and to use.
In my site there is also a new version of gafsoc, a program for evaluating the dimension of a single TB or of a group of TBs until 9 men.
Another program (gafsdim) gives the number of positions until 32 man, so it is possible to have an idea of the occupations of all the future TBs and of the total number of positions in chess.
This program is also present in source format (but comments are only in Italian). I tried to write it in a style easy to be deciphered in order to permit to all interested people to understand the code and eventually improve the
program. It requests the use of the gmp (GNU multiple precision) library because the program utilizes very long integers to perform the calculation.

Guido

[Wulff]

Hi Guido!

I have a couple of questions:

1) Does your TB generator already support generating 9-men TB's ??
2) Are you planning to release a compiled version of the generator for Windows (either 32 or 64 bit) ?
3) Is it possible to download some already generated TB's anywhere ?

Keep up the good work, I want to see that 32-men TB

[guido]

Hi Guido!

I have a couple of questions:

1) Does your TB generator already support generating 9-men TB's ??
2) Are you planning to release a compiled version of the generator for Windows (either 32 or 64 bit) ?
3) Is it possible to download some already generated TB's anywhere ?

Keep up the good work, I want to see that 32-men TB

Hi Wulff!

I try to answer to your questions:

1) In theory the algorithms used in my program are completely general, so it could generate TBs with any number of men, but in the reality this is not true. In the last version I put a limit to 6 men and inside the program there is a
statically allocated vector of structures whose dimension is given by the total number of endgames with 6 or less men. Modifying this point is not a problem and in future I'll substitute this vector with a dinamically allocated vector
on the basis of the length of the endgame requested in the command line.
But the real problems are the following (in order of importance):
- Excessive use of the I/O with the method 3 (retrograde algorithm) which is the fastest but requests more I/O and sometimes produces an error (I'm just now looking for it). So the method 2 (semiretrograde algorithm) is at
present globally better. I'm trying to reduce the use of I/O in both these methods.
- Cpu speed and optimization of the code: I should rewrite many parts of the program, write some functions in assembler and try to use a compiler faster than gcc.
- RAM and disk dimension
- Possible still existent errors.

2) I had always problems with Windows, so for various reasons I decided to use only Linux, where problems with 64 bits are more easy solved by gcc, in particular if the PC is a 64 bits machine. With some difficulties I produced and put on my site also a Windows version of gafslib, hoping that it can be linked to a program written under Windows. I have now the djgpp compiler (a C compiler under Windows similar to gcc) but I still have problems with the compilation and linkage of the generator.

3) No. At present nobody put these files at public's disposal. If someone wants ...

Ciao
Guid