Author: Robert Hyatt
Date: 19:59:31 04/25/00
Go up one level in this thread
On April 25, 2000 at 10:30:48, Brian Richardson wrote: >I did quite a bit of testing rotated (Crafty-style) vs non-rotated (nand path >[f][t] with occupied). Overall movegen speed was under 10% faster non-rotated. >However, I used much slower gen_push (see TSCP) function for saving moves vs >much faster Crafty *move++ with shifts method (overall would save about 5-10% of >movegen speed). > >I also combined bitmaps for bishops/queens and rooks/queens to save all the or's >during generation. The downside is in eval I have to and rq and bq to get just >queens (and minor overhead of updating both in make/unmake). Another price is >it may be harder to do a SEE (need to know if bishop or queen to take with LVA), >but I have not done one yet. Since movegen is under 20%, I am comfortable >taking the lazy non-rotated way out. Note that I do _both_ in Crafty. I have a bitmap for white bishops, white queens, black bishops, black queens, and _all_ bishops/queens together. That is most useful for Swap() calculations.
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.