Author: Vincent Diepeveen
Date: 15:03:55 08/08/03
Go up one level in this thread
On August 08, 2003 at 13:06:09, Angrim wrote: at k7 and k6 and p3 i am sure 8 bits is a lot faster. but if you maintain 2.5+ MB source code then you love to get rid of all possible casting bugs in advance. Also some so called 'inferior' compilers have big problems with all that casting. Some dirty compilers have less of a problem with it. So i am very sure you can speedup code at k7 using 8 bits stuff. Probably around up to 10% for a running chess program as you avoid huge L2 cache spillings. >Thanks for posting your code, always interesting to see how >someone else does things. I found that all those 1 and 2 letter >variable names made it a bit harder to read, but it will be >worth the trouble if I find any useful ideas in there. > >I wonder how much speed gain you get from storing your >moves as a 32 bit int and doing > >rb->zetend->zet = (SRsq|u|move_captures|move_pbishop); > >compared to storing them as unsigned chars in a struct and assigning >values to each element of the struct, which is the method that I use >currently. Since I store generated moves so very often, doing this >in a slow way may be costing me a detectable amount. > >Angrim
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.