Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Persistent Hash size bug?
- - By Martin Hander (*) Date 2008-08-10 12:51
I tested the following with 32-bit and 64-bit single cpu Rybka 3:

I created a persistent hash file with a size > 1GB.

The uci commands are:

setoption name Persistent Hash Enabled value true
setoption name Persistent Hash Size value 2048
setoption name Persistent Hash File value PHF

Up to 1GB (1024) everything is ok, the file will be created with the correct size.

With 2GB (2048) the following happens:

The file is created with a size of zero Bytes,
and then steadily grows during calculations (after the go-command) up to the correct size of 2GB.

If Rybka is stopped before the size of the hash file has reached 2GB, it remains on the disk with a smaller size.

I don't know if this can cause an error, but how should Rybka know the correct size of the hash-file after a restart?
(The hash-file seems not to have some header bytes with the correct size)
Parent - - By Fulcrum2000 (****) Date 2008-08-10 13:18
Might be an OS limitation, FAT-32 is for example limited to 2GB filesize. Creating a file over 2GB might cause some strange behavior.
Parent - - By Martin Hander (*) Date 2008-08-10 13:48
This is correct, but I tested Windows XP with NTFS (the default File system).
NTFS can handle files with 100GB without any problems.

But the software api to access files >=2GB on 32bit is a little different,
so it is easy to make an error here.
Parent - By Vasik Rajlich (Silver) Date 2008-08-11 20:09
Hmmm - Rybka is just using ::fopen, ::fwrite, etc. I'll have a look.

Vas
Up Topic Rybka Support & Discussion / Rybka Discussion / Persistent Hash size bug?

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill