Rybka 4 just around the corner it seems

[Graham Banks]

only my grandma was allowed to call me Mikey. Grandma!

No - my secret's out!
Now listen here sonny! This walking stick can still land a good whack.

[Graham Banks]

He looked at it and saw that this little engineoutputs shake his happy little world. Instead he starts to telling me stories of old womans

If you were such a reliable expert, there would be no argument from anybody. The only top programmer (as in top ten engines) who has anything to say disagrees with you for a start.

[bob]

The problem is, you are leaving out a _lot_ of "middle-work". Where does that large set of independent positions come from? If you feed them to the program (say take 300 WAC positions and say "gimme the results for these" then I agree, it will be faster to search all to the same depth, independently. But if you are analyzing a _game_, how do you create those positions without first doing analysis to discover the interesting positions, and then analyzing those to discover more interesting positions, etc. That is not going to scale perfectly. It is going to scale poorly.

Ok. we first agree to the obvious. If the positions are different then we agree.

Now lets come to the situation of the same game.

Lets say we have a position, and take the first 4-best moves (and also suppose they are not highly transposive).
The task is the following.

"for every one of the 4 best moves, autoplay for 20 plies (depth 15)"
(autoplay with fixed depth ==15)

Now you could do it in 2 way.
a) One 4-core engine will get through its variation
b) 4x 1-core engines will get extend one move each.

Which is faster? The 1-core engines _DO_ get benefit of the hash along the prolongation.

And the 1-core engines do not get the benefit of the information seen by the other engines. If you have 4 people working on analyzing a game, would you rather have them close together so they can exchange important information, or would you rather have them in separate rooms so that they share nothing, and waste lots of time re-computing the same stuff?

What leads you to believe that the 4 engines will finish at the same time? It is _highly_ unlikely. That's a classic load-balancing issue. If it were that easy to balance loads among processors much of our present difficulty would go away. Different branches of a game take different twists and turns, and take wildly varying amounts of time to search to the same depth. This is a fatal flaw.

There are others as well. Parallel search is not quite as simple as you are trying to make it.

On the other hand, when the MP-engine finishes the first task (autoplay first one of the 4-candidate moves) and comes to extend the second, the hash is worthless (i have never seen hash working more than 4-5 plies, )

I see hash information being used search-to-search all the time. It often greatly reduces the effort for succeeding searches.

[Graham Banks]

He looked at it and saw that this little engineoutputs shake his happy little world. Instead he starts to telling me stories of old womans

If you were such a reliable expert, there would be no argument from anybody. The only top programmer (as in top ten engines) who has anything to say disagrees with you for a start.

I don't think there's any point continuing to go round in circles.
I get the impression that most are sick and tired of the whole affair, and people have pretty much made their minds up about what they believe.
I'm going to leave it there.
And despite my harsh words in my last post, I do still have a lot of respect for your abilities. I have no intention of making enemies from the debate.

Cheers,
Graham.

[oreopoulos]

I see hash information being used search-to-search all the time. It often greatly reduces the effort for succeeding searches.

I think you got it wrong on from the start.

It is obvious (to me) that if the tree creation was handled by an engine , with a central shared hash, the result would be tons faster. But there is no arguing about that. (i would really love to see engines designed with analysis in mind and not playing, but noone does that. Everyone focuses on playing)

If i can make an analogy it would be like this.
In an engine you have evaluation functions. You use them to sort moves and help with your search. If you have 200 cores that can "evaluate" at the same time, you are speeding up 200 times the evaluation part. Nothing more nothing less.

For idea, the evaluator is the engine. There you have the speed up. nowhere else.

i really dont see any point to disagree.

[Dr.Wael Deeb]

only my grandma was allowed to call me Mikey. Grandma!

No - my secret's out!
Now listen here sonny! This walking stick can still land a good whack.

[Dr.Wael Deeb]

He looked at it and saw that this little engineoutputs shake his happy little world. Instead he starts to telling me stories of old womans

If you were such a reliable expert, there would be no argument from anybody. The only top programmer (as in top ten engines) who has anything to say disagrees with you for a start.

I don't think there's any point continuing to go round in circles.
I get the impression that most are sick and tired of the whole affair, and people have pretty much made their minds up about what they believe.
I'm going to leave it there.
And despite my harsh words in my last post, I do still have a lot of respect for your abilities. I have no intention of making enemies from the debate.

Cheers,
Graham.

Exactly my reasons why I chilled out from the Rybka's affair....
Dr.D

[Graham Banks]

I have no intention of making enemies from the debate.

Cheers,
Graham.

Exactly my reasons why I chilled out from the Rybka's affair....
Dr.D

Yeah - we should be able to disagree with others without lasting problems.

[Dr.Wael Deeb]

I have no intention of making enemies from the debate.

Cheers,
Graham.

Exactly my reasons why I chilled out from the Rybka's affair....
Dr.D

Yeah - we should be able to disagree with others without lasting problems.

Fully agreed
Dr.D

[Alexander Schmidt]

If you were such a reliable expert, there would be no argument from anybody. The only top programmer (as in top ten engines) who has anything to say disagrees with you for a start.

My abilities where OK when I posted about Kaissa, Patriot, Deep<9>, Fafis, Strelka, Rashess and Demon just to name a few.

But I am not able to proove that Ippolit is no clone. I just can show you the diferences.

If you stop claiming it is a clone, there is no need to argue with you