JasonLion wrote:dobrichev's code is in the same general range as JSolve, slightly better with some compilers/puzzles, slightly worse with others. Preliminary tests show that zhouyundong_2012's code is now faster than either of them by roughly 10%. However, I haven't run zhouyundong_2012's code against the complete range of puzzle collections, so there is a small chance of better or worse times on some kinds of puzzles.
thank you very much for these informations.
I think one need a wider gap to change a code which does not weight too much in most of the tasks;
but this can be a new base for future improvements.
champagne