Author: Peter Fendrich
Date: 11:57:46 10/26/98
Go up one level in this thread
On October 26, 1998 at 09:23:19, Robert Hyatt wrote: >On October 25, 1998 at 19:58:08, Peter Fendrich wrote: > >>On October 25, 1998 at 19:28:54, Robert Hyatt wrote: >> >>>On October 25, 1998 at 19:16:05, Roberto Waldteufel wrote: >>> >>>>Hi all, >>>> >>>>Has anyone tried this, or thought of trying this before? >>>> >>It certainly doesn't break up my rotated bitboards. >>When rotating 90 degrees, I pack the diagonals together in 8-bit chunks like >>this: >>h1 + b1-h7 >>f1-h3 + d1-h5 >>a2-g8 + a8 >>a4-e8 + a6-c8 >>g1-h2 + c1-h6 >>and so on >> >>all the adjecent squares are together and it works fine for me. >> >>I don't try to access these 8-bit chunks - it's merely a way of visualising the >>board for myself... >> >>//Peter > > >I agree there. But what about the _files_ and _ranks_??? they won't be >adjacent in this scheme... which is the part that will take some fiddling to >make it work... This a completely different story and I didn't even try... I haven't tried to utilise the grouping in white and black squares either, it's just he way I stuff the bits for 90 degrees of rotating :) > >BTW, the current scheme is *not* bad on 32 bit architectures, because most of >what is done is AND/OR/XOR, and it simply takes two instructions no matter >what. And since the current processors (pentium and on) do at least two >instructions/cycle, there really isn't any overhead, except for those cases >where a shift is needed... and I bet that while that is going on there are >other instructions that the super-scalar units can grab to execute anyway... True, I don't think it matters much but haven't messured it. As all "bitbonkers" I'm aiming for 64-bit architecture. But doesn't possess any - yet! //Peter
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.