From: janburse@fastmail.fm   
      
   We recently presented a fast constraint solver   
   termed Railgun CLP(FD) that modelled attributed   
   variables simply via ‘$ATTR’/2 compounds. We   
   already went through an iteration which allowed   
   (#\=)/2 constraints. In this instalment we   
   present some further progress. In particular we   
   intoduce a discount(C) = 1/k heuristic.   
      
   We demonstrated an additional scoring of variables   
   among the same degree to improve the static variable   
   ordering (SVO). It can give a 100x times speed up   
   in a magic square example. Putting Dogelog Player   
   on equal foot with SWI and ECLiPSe, while recent   
   versions of Scryer and Trealla were still 10x-30x   
   times slower.   
      
   See also:   
      
   Supersonic FF for Railgun CLP(FD)   
   https://qiita.com/j4n_bur53/items/0d1334f1f4c2234f3f47   
      
   Mild Shock schrieb:   
   >   
   > We recently presented a fast constraint solver   
   > termed Railgun CLP(FD) that modelled attributed   
   > variables simply via ‘$ATTR’/2 compounds and could   
   > deal with integer dif/2 constraints. In this   
   > instalment we allow (#\=)2 constraints and demonstrate   
   > that it can be ported to SWI-Prolog.   
   >   
   > Using polyfill for ‘$SEQ’/2 from Dogelog Player,   
   > we observed that running Railgun CLP(FD) inside   
   > SWI-Prolog gives a 2–3x speed-up for the Queens   
   > example, on both 32-bit and 64-bit. On the other   
   > hand the price tag for big integer flexibility   
   > seems to be a factor 35x slow down.   
   >   
   > See also:   
   >   
   > Porting Railgun CLP(FD) to SWI-Prolog   
   > https://medium.com/2989/e9f2ef4e6878   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)