''[Sarnold] 2005-09-02'' - After playing some time with [Mancala], I improved the AI, but did encounter some perf issues. So I decided to write the most used function in C with [Critcl]. Here is the C procedure for the move : ===c critcl::cproc cmove {char* vboard int player int pit} string { int go_again=0; int side; int board[2][6]; int store[2]; int i; int orig_side,orig_pit,their_pit,opp; int stones=0; char temp; int walk=0; /* to walk through the result */ char error[10]="error"; char result[45+1]; /* 12 for the holes, 2 for the stores, 1 for go_again and 1 null-termination */ /* initialization of the board */ orig_side=side=player; orig_pit=pit; i=walk=0; while (vboard[i]!='\0') { temp=0; while (vboard[i]!=' '&& vboard[i]!='\0') { if (vboard[i]<'0'||vboard[i]>'9') { return "error"; } temp=temp*10+vboard[i]-'0'; i++; if (temp>36) { return "error"; } } if (walk<12) { board[walk/6][walk%6]=(int) temp; } else { store[13-walk]=(int) temp; } walk++; i++; if (walk>13) { break; } } for (i=0;i 0) { /* decrement the origin from one stone */ stones--; board[orig_side][orig_pit]--; if (pit >= 6) { /* in that case we put a stone into one of the 2 stores. */ store[side]++; /* when we put the last stone into the store of the opponent, ** it should be the turn of the opponent */ go_again= (side==player || stones!=0); pit=0; side=!side; } else { go_again =0; i = ++board[side][pit]; /* See if we captured any opponent stones */ if (stones==0 && player==side && i == 1) { opp=!side; their_pit=5-pit; /* capture_opposite */ if (board[opp][their_pit]!=0) { store[side]+=1+board[opp][their_pit]; board[side][pit]=0; board[opp][their_pit]=0; } break; } pit++; } } /* converting board, go_again and store for returning a string list-like */ temp=(char)go_again; if (temp>10) { return "error"; } temp='0'+temp; result[0]=temp; walk=1; for (i=0;i<14;i++) { if (i<12) { temp=(char)board[i/6][i%6]; } else { temp=(char)store[13-i]; } result[walk]=' '; walk++; if (temp>99) { return "error"; } if (temp>=10) { result[walk]='0'+(temp/10); walk++; temp=temp%10; } result[walk]='0'+temp; walk++; } result[walk]='\0'; return result; } === It takes as arguments : the board, which is a list of numbers converted to a string (char* type), and two integers : side and go_again, both booleans. The result returned is a string containing : * go_again at index 0 * the modified board from index 1 to end as if it was a list. (numbers are separated by a space) Here is the tcl wrapping : * into the move proc : # cmove is our C proc set res [cmove $board $player $pit] if {[string equal $res error]} { error "error in move" } return [list [lindex $res 0] [lrange $res 1 end]] Of course I created a similar cproc which simulates the other game mode. ---- ---- '''Compiling issues''' Having MinGW/MSys on my WindowsME machine, what was needed for compiling under the DOS console is : set PATH=%PATH%;d:\mingw\bin;d:\tcl\bin tclsh critcl.kit -lib cmancala cmancala.tcl , where ''cmancala.tcl'' is our cproc definition. Then to access the cmove proc : tclsh % load cmancala.dll % cmove [split 33333333333300 ""] 0 0 0 0 4 4 4 3 3 3 3 3 3 3 3 0 0 That's it ! <> Critcl