Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Rybka Discussion / Unusual hash signatures
- - By user923005 (****) Date 2013-09-20 23:22
This is from KnightCap (an MTD(f) engine):

struct hash_entry {
  uint32 hash1;
  uint32 hash2;
  Eval low;
  Eval high;
  unsigned char depth_low;
  unsigned char depth_high;
  unsigned from:6;
  unsigned to:6;
  unsigned tag:4;
  unsigned char move_num;
};
Parent - - By Rebel (****) Date 2013-09-20 23:27
KnightCap is from ?

Open source?

Year?

Interesting...
Parent - By user923005 (****) Date 2013-09-20 23:42
Andrew Tridgell wrote an interesting, learning, MTD(f) engine called KnightCap a long time ago (around 1996-1997, IIRC).  Here are some versions stored for posterity:
ftp://samba.org/pub/KnightCap

I did a port of KnightCap to Windows a long time ago for Frank Q.
Parent - By user923005 (****) Date 2013-09-20 23:49
Parent - - By bob (Gold) Date 2013-09-20 23:59
That looks pretty similar to what I had done for Crafty's mtd(f) conversion.  Only differences were I did a 64 bit hash sig, not two 32 bit values, and I had the move in the usual crafty compressed format, not using bitfields which are horribly slow.

I notice only two depth entries, which makes sense.  Fruit stores 4.

That is just what one would expect in an mtd(f) program, of course...
Parent - - By user923005 (****) Date 2013-09-21 00:13 Upvotes 1
I also saw that the idea of multiple depths in Fruit's hash table was in Fruit for quite a while (for instance, it was in the transposition table in 2004, which predates the CCC conversation with Vas about storing multiple bounds).
So I think it is safe to say that Fabian did not get the idea from Vas originally.
Parent - By bob (Gold) Date 2013-09-21 16:33
I never thought he did.  But there were earlier mtd(f) discussions.  I have an OLD version of my first experiments with mtd(f) which were based on version 12.0.  I don't have any good timestamps any longer, but that is certainly WAY old.  Version 10.x was the 1996 WMCCC version, I'd guess this version would be somewhere in the 1997-1998 time period.  Don might remember when he triggered all the mtd(f) discussions on CCC.  It was in that time-frame...

The basic idea, however, was to add an additional bound (since you only store bounds) and a draft for that bound.  Nothing more...
Up Topic Rybka Support & Discussion / Rybka Discussion / Unusual hash signatures

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill