CCRL Discussion Board

Hi Martin, The good news is that the iterations on kqpkrn should speed up over time, but the bad news is that there are going to be 110 iterations. 69 is just the minimum necessary to resolve all promotions and captures. I'm attaching my results for kqpkrn. Your program is very slow. The Nalimov alg...

I have created the "missing" Nalimov endgame kqpkrb by transforming the output from Johan de Koning's FEG program. This takes several hours and lots of disk space, but is still quite a bit faster than directly creating it with the Nalimov program. Also, the conversion has not been extensiv...

#include <stdlib.h> #include <stdio.h> #define CHUNK_SIZE (128ul*1024ul*1024ul) #define FILE_SIZE (65536ul*32768ul) int main (int argc, char *argv[]) { FILE *fpIn, *fpOut; char *pbBuf, *pchSrc, *pchDst; char szFile[1024]; int iFile, fDot; long cbRead, cbWritten; unsigned i; if (2 != argc && ...

There was a tbsplit utility available in one of the Nalimov code releases.

The function is extremely simple. Start with the uncompressed .nb? files, and sequentially split them into files of size 2^31, with extensions .0.nb?, .1.nb?, etc.

-Marc

There seem to be two alternatives. Win-Lose-Draw (WLD) tables could become the standard; they are much smaller and perhaps some hybrid system of WLD and Nalimov would work. That seems to be the approach most in favor now, I hear. Alternately, several years from now the gigantic Nalimov files for 7-...