Author: Uri Blass
Date: 02:15:47 04/13/03
Go up one level in this thread
On April 13, 2003 at 00:14:52, Omid David Tabibi wrote: >On April 12, 2003 at 22:45:19, Vincent Diepeveen wrote: > >>On April 12, 2003 at 13:20:51, Omid David Tabibi wrote: >> >>>On April 12, 2003 at 10:02:43, Uri Blass wrote: >>> >>>>On April 12, 2003 at 09:17:31, Omid David Tabibi wrote: >>>> >>>>>Quite an interesting read Vincent. >>>>> >>>>>I'm afraid you are investing too much in the parallel speedup though. Any >>>>>hardware speedup will be linear (at best) while algorithmic enhancements are >>>>>exponential. If you manage to search one ply deeper by an algorithmic >>>>>improvement, the gain will be more than any parallel speedup can yield. >>>> >>>>I agree that the hardware speedup from parallel search will be linear at best >>>>but linear improvement is not always less than one ply. >>> >>>Diep is already parallel. I assume that he will get far less than a 4x speedup >>>for his latest work on massive parallelism. Assuming an effective branching >>>factor of 4, that speedup will equal one ply. >> >>b.f. = 2.9 > >Because you are using standard R=3; but is the search reliable? That bf will not >be of much use if it causes Diep to find the correct move two plies later in >comparison to its competitors. When was the last time you compared Diep's >performance to other engines using test suites? I think that in this case the evaluation and the qsearch are important. I believe that the main problem with standard R=3 and no verification search is result of bad evaluation or a qsearch that does not include tactics except captures. I also know that Diep is outsearched(I mean only to plies) by other programs inspite of that branching factor so I guess that the problem of diep is not finding the right move 2 plies later relative to it's competitors. Diep was not first but it did draws with programs like Fritz and Shredder so I do not think that x plies of diep is worse than x plies of most programs. Uri
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.