Peter Skinner wrote:Houdini wrote:
Yawn.
How many times do I need to repeat that Houdini is original at the source code level, and doesn't contain any licensed code other than the Gaviota and Nalimov EGTB code?
If you search the forum (and/or the Rybka forum) you'll find several occurrences of my saying so.
My current reply will certainly not stop you from asking the same kind of question in 6 months time, pretend I never answered it, and then make big statements about "speaking volumes"...
Robert
Ok so it includes "public domain" code, and was not 100% originally written by yourself. Thank you for your answer.
Peter
Peter,
You did not read his reply carefully. He did not say that Houdini is original. He said that "Houdini is original at the source code level." You can take a program written in Pascal, make a direct translation to C and be able to make this claim even if you have a functionally identical program. Here is an example using pure C:
// add all the numbers from 0 to 99
int x = 0;
int y = 0;
for (x = 0; x < 100; x++) y += x;
Now here is a program that is functionally identical but is "completely different" at the source code level:
sum ^= sum; // clear sum
int i = 99;
while (i > 0) { sum = sum + i ; i-- }
Roberts statement is more or less completely true. He made massive changes at the "source code level" - for example he completely rewrote Ivanhoe's move generator to gain a small speedup. Essentially he applied a lot of optimization's to the EXISTING algorithms. You cannot pin him down on this because he also made actual improvements to Houdini other than mundane optimization's, enough to convince (at least himself) that he is a real programming and just a cleanup boy. It's the difference between being an actual engineer who designs engines for high performance vehicles versus being a mechanic at a garage who specializes in modifying these engines to perform better with bolt on parts and machine work.
There are people who say the first Houdini is not even a translation, just the original Robolito code with a few changes of his own. There have been dumps that supposedly prove that but I have not seen them with my own eyes. But what ANYONE can see with their own eyes is the experiment on an early version of Houdnii where every move of a 1 ply search matches one of the Robbo versions. Not just most but EVERY single move selected from random positions. It's undeniable proof of how this program was engineered and even MORE so when combined with the scores generated from these one ply searches - which are identical after scaling. He probably believed that multiplying the scores by some constant would fool people or else he did it for performance reasons - when you do a deeper search programs behave different even when scaled identically because it automatically changes the meaning of the margins used.
So it doesn't matter what Robert says and getting him to make a statement just so that we can marvel at his ingenuity in evading the truth is not going to get you anywhere.
But I don't want people to be misled by his statements so I wanted to make sure it got straightened out. He did not lie here, he just evaded.
"It depends on what the meaning of the words 'is' is." –Bill Clinton, during his 1998 grand jury testimony on the Monica Lewinsky affair
Capital punishment would be more effective as a preventive measure if it were administered prior to the crime.