search

thomas-maeder / popeye

Popeye is a chess problem solving and testing software with strong support for fairy chess and heterodox genres. For more information cf. topic "Popeye (chess)" on http://en.wikipedia.org/
30 stars 13 forks source link

PSer, PhSer processed as series #337

Open BjornEnemark opened 2 years ago

BjornEnemark commented 2 years ago

Ser-H#, Ser-# and Ser-S# are solved very fast. I think that Popeye keeps a table of positions previously obtained, so that sequences like 1. Se4 2. Sf6 3. Se4 is not examined further. The same principle should be used for PSer-/PhSer- problems

JoshuaGreen commented 2 years ago

I doubt that dramatic improvement here will be so simple, but just to pin down the request, are you thinking of ordinary solving or intelligent mode solving?

BjornEnemark commented 2 years ago

I am thinking of ordinary solving for selfmate seriesmovers (Ser-S# or Pser-S#), no intelligent mode possible as far as I know. If a position is reached that has been encountered before (like 1. Sf6

  1. Sg4 3. Sf6) then this branch should be cut. As I understand it, solving of seriesmovers for helpmates will do just that: check each move to see if the current position has been seen before in the solution. All intermediate positions are stored in memory, controlled by maxmem. I think that Pser-H# are not treated like this in ordinary solving.

Bjørn

Den 28-05-2022 kl. 02:47 skrev Joshua Green:

I doubt that dramatic improvement here will be so simple, but just to pin down the request, are you thinking of ordinary solving or intelligent mode solving?

— Reply to this email directly, view it on GitHub https://github.com/thomas-maeder/popeye/issues/337#issuecomment-1140121826, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADYN5CUM4ANN2K62PS2G7G3VMFUKRANCNFSM5V4U4YLA. You are receiving this because you authored the thread.Message ID: @.***>

JoshuaGreen commented 2 years ago

Yeah, intelligent mode doesn't work for selfmates (though I feel that with some work it could be effective against reflexmates, Issue #315). At any rate, though my understanding of the infrastructure is rather limited, I'm sure that positions are cached in pser problems. The issue is that any piece can potentially move, hence a lot more positions may in principle be stored, and that can require a lot more memory (both in total, and possibly per position).