06:31 <razzy> i can execute your Regenaxer parael game function in chess.l . but i cannot debug (!) it in any reasonable way. nor i can (pretty ) print from it. it makes SIG errors. how do you look what it does?

07:26 <Regenaxer> Hi razzy

07:27 <Regenaxer> Hmm, probably cause all processes are on the same tty

07:27 <razzy> Hi :] all

07:28 <razzy> uff, painfull

07:28 <Regenaxer> :)

07:28 <Regenaxer> Somehow '*Dbg' switched on in only one process?

07:29 <Regenaxer> Or config it in such a way that only a single process goes "parallel"?

07:30 <razzy> if i could config it, i would not need to debug it to reason about it :/

07:32 <Regenaxer> true, it simply dispatches processes in the first ply

07:32 <Regenaxer> The SMP flag

07:33 <Regenaxer> I think this parallel version is not useful, as the processes compute rather redundantly the same stuff

07:33 <Regenaxer> The problem is *how* to distribute the search in a clever way to all the processes

07:35 <razzy> Regenaxer: i agree it is mostly not usefull for picochess player. but i hoped it makes subprocess for every move on starting board so i could read out all moves rated with my cost function for analysis

07:36 <Regenaxer> ok

07:38 <razzy> to distribute search on all cores is imho hopeless from top-down aproach as you demostrated. too much overhead. too much redundant stuff

07:40 <razzy> i cannot see simple, powerfull way

07:43 <Regenaxer> yeah

07:44 <Regenaxer> Perhaps paralellize on the bottom level. But there most moves are pruned off

07:44 <razzy> i can see simple, not easy, powerfull, way for bottom-up paraelisation

07:46 <razzy> one would be to build move-tree far ahead and paraelise and store cost evals and sub trees.

07:47 <razzy> which is completely different aproach

07:47 <razzy> than picochess use now

07:47 <Regenaxer> Perhaps, yes, but in this implementation the cost calculation is inexpensive compared to move generation

07:48 <razzy> second would be make picolisp paraelise processes automagically from lisp AST tree:]

07:59 <razzy> Regenaxer: btw, picochess cost function seem oversized and costly.

07:59 <razzy> and i still do not understand half of it :]

08:02 <Regenaxer> It is quite primitive, almost only material counting (with considering possible battles)

08:03 <razzy> and battles are pieces removal from board?

08:04 <razzy> not many moves ahead?

08:06 <Regenaxer> Many moves ahead, but only considering exchages

08:06 <Regenaxer> also "pinned" pieces which may attack another field but cannot move cause otherwise the king would be in check

08:07 <Regenaxer> queen too

08:08 <Regenaxer> I don't remember the details

08:09 <razzy> no problem, thx. i know what to look for.

08:10 <Regenaxer> :)

08:11 <razzy> *now i know