Author: Robert Hyatt
Date: 14:34:41 01/25/00
Go up one level in this thread
On January 25, 2000 at 15:51:54, Ernst A. Heinz wrote: >On January 25, 2000 at 15:35:58, Robert Hyatt wrote: >> >> [...] >> >>Hsu mentioned "we disabled the futility pruning in the q-search which slowed >>the >>chip by a factor of 10..." Most noticed the factor of 10. _I_ noticed the >>"we disabled the futility pruning". Because Hsu and I had had a _long_ >>discussion about this in r.g.c a few years back. I ran a bunch of tests using >>SEE, MVV/LVA and SEE+futility, and posted the results. Apparently he liked the >>data and implemented this quietly. > >As far as I know, Hsu was always convinced that MVV/LVA with >futility pruning in the quiescence search was the right way >to go -- he already wrote about it in his Ph.D. thesis. Even >"ChipTest" did it exactly like this I suppose. > So far as I know, he didn't do futility pruning in the q-search. When we had the original discussion in r.g.c, I mentioned that SEE was making my code in Crafty (and Cray Blitz since I did it the same way there) over 50% faster than pure MVV/LVA ordering. He asked for details. At that point, I realized that there were two distinct issues: (1) ordering moves with SEE vs MVV/LVA. (2) I was doing a type of futility pruning (tossing out captures that were hopeless.) I then reworded my note and tried tests. I first found that SEE was about 10% better than MVV/LVA, looking at the tree size. And since SEE was pretty cheap in bitboards, overall it was faster as well. I then found that tossing bum captures was a 50% gain. He thought that would cause problems. We had a long discussion, with several test positions. I can certainly be wrong about whether he used it or not, but he certainly said that he was looking at _all_ captures at the time of the discussion which was somewhere around 1993-1994... So his doing some sort of futility tossing was a surprise to me... I didn't notice this in his thesis as I was more interested in the parallel search stuff. >So, no surprises here. He did not implement anything quietly. > >>We learned something about DB (or at least >>I did) in this simple discussion... > >Not really anything new ... unfortunately (see above). > >=Ernst= Not even the branching factor?
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.