I ran a gauntlet between Crafty 22.1 64-bit compiled by Jim Ablett and BugChess2, then between Crafty and Thinker.
In both cases Crafty's memory usage was creeping upwards, it started at 160 MB and after 12 games it already used 1 GB and disk swapping started to occur (I have 2GB RAM).
This on Vista x64. Machine is Q6600.
What can be the reason?
My Crafty.rc:
log off
egtb
#adaptive 750K 24M 192M 24M 48M
tbpath=C:\Sakk\TB
cache=32M
ponder off
# mode=tournament # (for use use with ponder on)
swindle on # (Allows Crafty to try to win drawn games (according to Endgame Tables))
learn 0 # (default)
book off #(default = book on)
mt=2 # (Increases Crafty's MaxThreads to 2 for a dual CPU computer )
smpnice=1 # (Make Crafty not use cpu on opponents time)
timebook 80 8 # (Makes crafty use a lot more time on the first 8 moves out of book)
#adaptive NPS a
hash=96M
hashp=40M
exit
Hi Gabor,
Smpnice is really Smpbad as far as Windows is concerned. You can't use it. Running processes are not terminated and just multiply.
As I understand it this has been fixed in latest unreleased Crafty by converting to threads. We'll have to wait for official release for the fix.
I ran a gauntlet between Crafty 22.1 64-bit compiled by Jim Ablett and BugChess2, then between Crafty and Thinker.
In both cases Crafty's memory usage was creeping upwards, it started at 160 MB and after 12 games it already used 1 GB and disk swapping started to occur (I have 2GB RAM).
This on Vista x64. Machine is Q6600.
What can be the reason?
My Crafty.rc:
log off
egtb
#adaptive 750K 24M 192M 24M 48M
tbpath=C:\Sakk\TB
cache=32M
ponder off
# mode=tournament # (for use use with ponder on)
swindle on # (Allows Crafty to try to win drawn games (according to Endgame Tables))
learn 0 # (default)
book off #(default = book on)
mt=2 # (Increases Crafty's MaxThreads to 2 for a dual CPU computer )
smpnice=1 # (Make Crafty not use cpu on opponents time)
timebook 80 8 # (Makes crafty use a lot more time on the first 8 moves out of book)
#adaptive NPS a
hash=96M
hashp=40M
exit
I suspect this has to do with the smpnice=1 feature. You appear to be using two threads. Once we get 22.2 cleaned up and out, I believe this will all go away as I have gone back to threads on linux as well, since NPTL is now functional and working quite well.
