Author: Vincent Diepeveen
Date: 13:55:14 02/15/06
Go up one level in this thread
On February 15, 2006 at 15:20:16, Sune Larsson wrote:
>On February 15, 2006 at 15:11:30, Orlando Mouchel wrote:
>
>>>So, Vasik, please tell me: why would you would want to hide the node count of
>>>Rybka?
>>
>>Hi Anthony,(sorry for my bad english)
>>
>>You know, you are not the only one to have doubts about the honesty of Mr.
>>Rajlich. To be more precise, these doubts multiply more and more among the
>>programmers all around the world. More especially as I have just noticed that it
>>did not answer your question! Admittedly, this right belongs to him.
>>
>>While waiting, why not make free the very good "Zappa 2.0"? Using more tests,
>>one could then better appreciate his real power !
>>
>>Best regards
>>
>>Orlando
>
>
> You have to back this up.
>
> 1) Please specify your doubts about the *honesty* of Mr Rajlich.
>
> 2) Which are the "programmers all around the world" - that share
> your doubts about the *honesty* of Mr Rajlich.
> If you can't back it up - I will consider your post as a troll and
> it should be removed.
Actually it's not a troll.
I talked to 30 programmers. All 30 except for 2 found it bad taste. So that's
28-2. Look a lawyer undoubtfully can defend anything, but there is a difference
between making marketing for your program and doing statements which only a
laywer can defend.
Note i didn't talk with hyatt yet.
But there is a weird difference between the average engine, Junior, Tiger, Rybka
and Hydra.
Junior is creative in counting nodes. Futile evaluated nodes seem to get counted
as nodes. More nps = good.
A big handicap of diep always was that despite running on 512 cpu's my nps was
always low. So sponsors then ask: "how many positions a second do you get?"
Then i tell them the nps of diep, and you see their face change in a kind of
expression like: "did i give you such good hardware to get such a low nps?"
Tigers nps works different again. As a professional programmer explained who is
using his debugger as an extension of his crucial physical part: "Tiger just
counts full evaluated nodes". Tiger in terms of nps is the fastest pc program.
Vasik will get wet dreams ever reaching that nps, as in bitboards it's simply
not possible to get such amounts of nps.
Single cpu, Tiger will be the first to hit 10+ mln.
Obviously it's too fast to count therefore all nodes.
Now Rybka. As shown by a post of Enrico, Rybka seems to have some very weird
function inside its program that deliberately manipulates its nps. At a 233Mhz
MMX processor it first shows a nps like 200k nps or so in order to at 6 ply and
above manual adjust the nps printed to the screen to some low nps.
This is kind of real bad taste. There is a difference by a different way of
counting and manipulating the nps you *print* to the screen.
Oh talking about Hydra's nps. That's by far most transparant way to count nodes
to make his sponsor happy:
That's a printf("nps = %i\n",220000000+(random()%1000000));
Vincent
>
> /S
This page took 0.01 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.