Author: Peter Kappler
Date: 14:27:27 01/29/02
Go up one level in this thread
On January 29, 2002 at 16:09:32, Dann Corbit wrote: >On January 29, 2002 at 12:09:13, Peter Kappler wrote: > >>On January 28, 2002 at 23:39:40, Dann Corbit wrote: >> >>>On January 28, 2002 at 23:30:30, James Swafford wrote: >>>[snip] >>>>Are you aware there's a reasonably strong program on ICC written in Java? >>>> >>>>Information about Grok(C) (Last disconnected Mon Jan 28 2002 01:28): >>>> >>>> rating [need] win loss draw total best >>>>Bullet 2532 [8] 266 49 29 344 2532 (05-Jan-2002) >>>>Blitz 2572 865 304 130 1299 2668 (04-Apr-2001) >>>>Standard 2288 27 59 18 104 2346 (08-Sep-2001) >>>> >>>> 1: Grok v2.10 >>>> 2: A Java chess program! >>>> 3: 1.4GHz Athlon >>>> 4: Written and operated by Pyro. >>>> >>>>Pyro is Pete Kappler. Pretty good, huh? :) >>> >>>Mind-blowing amazement. I would like to hear what techniques he has used in the >>>Java engine. When you see a cow that can run in the Kentucky Derby, it makes >>>you wonder what they have been feeding her. >> >> >>Hi Dann, >> >>No special techniques. Java simply isn't as slow as most folks think. I use >>IBM's VM which is excellent (50% faster than Sun's). I've toyed with the idea >>of a C++ rewrite, but I don't think I would get more than a 1.5-2x speedup. >> >>Actually, this reminds me that I've been meaning to look at some of the native >>compilers that are now available for Java. Maybe I can find out if my 1.5-2x >>estimate is correct. > >My early experiences with speed (when Java first came out) were pretty dismal. >Pretty much just made me abandon learning more about it. From here (this from a >Java proponent), we have this: >http://www.javacoffeebreak.com/articles/thinkinginjava/comparingc++andjava.html > >"The biggest potential stumbling block is speed: interpreted Java runs in the >range of 20 times slower than C. Nothing prevents the Java language from being >compiled and there are just-in-time compilers appearing at this writing that >offer significant speed-ups. It is not inconceivable that full native compilers >will appear for the more popular platforms, but without those there are classes >of problems that will be insoluble with Java because of the speed issue." > >See also: >http://www.cs.colostate.edu/~cs154/PerfComp/ > >I would be interested to see how your experiments in timing turn out. I suspect >a lot has improved since my early experiments. I also found that the java sites >that benchmarked against C in those early tests were definitely sandbagging the >C compilers in a serious way (easily cutting the performance in half). >Probably, they just did not know how to use the C compilers. I think if both >sides (Java and C) are trying hard it will make a much better test. > >If Java can be speed competitive, I would like to know how to do it. I have >some friends who are Java developers, and I am sure they would be keen to know >about it too. Dann, I took a very quick glance at those articles and they are badly outdated (1996-1998 timeframe is my guess). Grok averages around 200knps in the middlegame on my Athlon 1400, and speed has never been a big concern of mine (obviously). I guess what I'm saying is that if I ported Grok to C++, I'd still expect it to be a slower-than-average engine. I might fiddle with a native Java compiler this weekend. If so, I'll report what I find. -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.