[Bug-gnubg] Building the 6x10 gnubg_ts.bd

[Morten Wang]

After a bit of testing and a larger bit of patience I've been able to
build the 6x10 two-sided bearoff database.  My AMD XP2000+ spent about
nine hours and forty minutes, using the default hash setting.

What I've been able to find is:

1: It took nearly ten hours, but it's doable.
2: Memory usage maxed at around 286MB
3: Trying to reuse the 6x9 database resulted in segfaults and core
   dumps, so I had to start from scratch.  As far as I could tell from
   looking at the output from 'top' the memory usage never stops
   increasing, so I expect the segfault comes from lack of memory.
   When reusing the 6x9 database initial memory usage far exceeded the
   given hash setting, does the program read the whole database into
   memory or something?
4: md5 checksum:

   12dc70c86f356d06bc96ee38dee40c62  gnubg_ts.bd

Output to terminal, for those interested:

address@hidden temp]$ nice /home/gnubg/makebearoff -t 6x10 -f gnubg_ts.bd
Two-sided database:
Number of points             :            6
Number of chequers           :           10
Calculate equities           : cubeless and cubeful
Write header                 : yes
Number of one-sided positions:         8008
Total number of positions    :     64128064
Size of resulting file       : 513024512 bytes (489.3 MB)
Size of xhash                 :    100000000 bytes
Reuse old bearoff database   : no
Xhash status:
Size:    12500000 elements
Queries: 0 (hits: 0)
Entries: 0 (overwrites: 0)
64128064/64128064
Xhash status:
Size:    12500000 elements
Queries: 3538897343 (hits: -1855761767)
Entries: 12500000 (overwrites: 1151319878)
Number of re-reads while generating: 1099691814

Since the memory usage is stable I'd expect better speed if I increase
the hash size, and perhaps even a bit more if I get a newer compiler.
With a good deal of patience I wouldn't be surprised if it's possible
to do the 6x11 database too, maybe even the 6x12.


Morten!

-- 
"...a liquid which was almost, but not quite, entirely unlike tea."
  My opinions are not necessarily those of my provider,
  not necessarily mine, and probably not necessary.