Revisiting Strelka/Rybka

General discussion about computer chess...
jury_osipov
Posts: 24
Joined: Thu Dec 02, 2010 1:41 pm
Real Name: Yury Osipov

Re: Revisiting Strelka/Rybka

Post by jury_osipov » Fri Jan 28, 2011 12:40 pm

And also:

My full name if Yury Ivanovich Osipov.
President of Science Academy is Yury Sergeevich Osipov.

User avatar
Chris Whittington
Posts: 437
Joined: Wed Jun 09, 2010 6:25 pm

Re: Revisiting Strelka/Rybka

Post by Chris Whittington » Fri Jan 28, 2011 12:50 pm

jury_osipov wrote:And also:

My full name if Yury Ivanovich Osipov.
President of Science Academy is Yury Sergeevich Osipov.
Jury, are you descended from these great Russians?

http://film.comparethemeerkat.com/courageousness/

http://film.comparethemeerkat.com/fearlessness/

http://film.comparethemeerkat.com/ambitiousness/

BB+
Posts: 1484
Joined: Thu Jun 10, 2010 4:26 am

Re: Revisiting Strelka/Rybka

Post by BB+ » Fri Jan 28, 2011 11:33 pm

Sergey Markov - russian chess programmer, author of Smarthink.
Sergey Markov - russian political scientist and journalist.
There's also http://en.wikipedia.org/wiki/Sergey_Markov
Russian army general and one of the founders of the Volunteer Army counterrevolutionary force of the White movement in the southern Russia during Russian Civil War.

jury_osipov
Posts: 24
Joined: Thu Dec 02, 2010 1:41 pm
Real Name: Yury Osipov

Re: Revisiting Strelka/Rybka

Post by jury_osipov » Sat Jan 29, 2011 10:57 am

BB+ wrote: For point (a), I don't think 2x is a ridiculous expectation, especially if you compare to 64-bit. I suspect what Osipov says is simply an over-statement (perhaps just rounding up, when it is really something like ~70% faster).
It depends on computer. For example, my latest version of Strelka (based on the code of Rybka 3) show the following speed:

Core Duo E6750: 75% faster than Rybka 3
Core I7 920: only ~50% faster than Rybka 3

(both 32-bit and 1-cpu versions).

Sentinel
Posts: 122
Joined: Thu Jun 10, 2010 12:49 am
Real Name: Milos Stanisavljevic

Re: Revisiting Strelka/Rybka

Post by Sentinel » Mon Jan 31, 2011 5:06 am

jury_osipov wrote:It depends on computer. For example, my latest version of Strelka (based on the code of Rybka 3) show the following speed:

Core Duo E6750: 75% faster than Rybka 3
Core I7 920: only ~50% faster than Rybka 3

(both 32-bit and 1-cpu versions).
Is it possible that popcnt implementation in Rybka 3 is such a crap???

jury_osipov
Posts: 24
Joined: Thu Dec 02, 2010 1:41 pm
Real Name: Yury Osipov

Re: Revisiting Strelka/Rybka

Post by jury_osipov » Mon Jan 31, 2011 9:13 am

Sentinel wrote:
jury_osipov wrote:It depends on computer. For example, my latest version of Strelka (based on the code of Rybka 3) show the following speed:

Core Duo E6750: 75% faster than Rybka 3
Core I7 920: only ~50% faster than Rybka 3

(both 32-bit and 1-cpu versions).
Is it possible that popcnt implementation in Rybka 3 is such a crap???
No, it is without popcnt implementation.

User avatar
kingliveson
Posts: 1388
Joined: Thu Jun 10, 2010 1:22 am
Real Name: Franklin Titus
Location: 28°32'1"N 81°22'33"W

Re: Revisiting Strelka/Rybka

Post by kingliveson » Tue Feb 01, 2011 1:34 pm

jury_osipov wrote:
BB+ wrote: For point (a), I don't think 2x is a ridiculous expectation, especially if you compare to 64-bit. I suspect what Osipov says is simply an over-statement (perhaps just rounding up, when it is really something like ~70% faster).
It depends on computer. For example, my latest version of Strelka (based on the code of Rybka 3) show the following speed:

Core Duo E6750: 75% faster than Rybka 3
Core I7 920: only ~50% faster than Rybka 3

(both 32-bit and 1-cpu versions).
Where can I get the source code?
PAWN : Knight >> Bishop >> Rook >>Queen

jury_osipov
Posts: 24
Joined: Thu Dec 02, 2010 1:41 pm
Real Name: Yury Osipov

Re: Revisiting Strelka/Rybka

Post by jury_osipov » Tue Feb 01, 2011 2:37 pm

kingliveson wrote: Where can I get the source code?
This is private version. I'm not going to release it.
It is called the Strelka 3P, mostly based on code of Rybka 3, but with significant improvements, and exists only in the 1CPU 32 bit version.
I can only report on recent test results:

Code: Select all

Blitz 5"+1"

Strelka 3P - Rybka 3 1-cpu w32:          +190 =238  -72
Strelka 3P - Rybka 4 (32 bit 1cpu):      +107 =246 -147
Strelka 3P - Stockfish-191-32-ja:        +174 =200 -126
Strelka 3P - Ivanhoe 999947c 32x:        +108 =248 -144
Strelka 3P - Fire 1.31 w32 KLO:          +124 =246 -112
Strelka 3P - Houdini-w32-1cpu (1.03a):    +71 =234 -195
Strelka 3P - Houdini 1.5a (32 bit 1cpu):  +68 =185 -247

User avatar
kingliveson
Posts: 1388
Joined: Thu Jun 10, 2010 1:22 am
Real Name: Franklin Titus
Location: 28°32'1"N 81°22'33"W

Re: Revisiting Strelka/Rybka

Post by kingliveson » Tue Feb 01, 2011 4:37 pm

jury_osipov wrote:
kingliveson wrote: Where can I get the source code?
This is private version. I'm not going to release it.
It is called the Strelka 3P, mostly based on code of Rybka 3, but with significant improvements, and exists only in the 1CPU 32 bit version.
I can only report on recent test results:

Code: Select all

Blitz 5"+1"

Strelka 3P - Rybka 3 1-cpu w32:          +190 =238  -72
Strelka 3P - Rybka 4 (32 bit 1cpu):      +107 =246 -147
Strelka 3P - Stockfish-191-32-ja:        +174 =200 -126
Strelka 3P - Ivanhoe 999947c 32x:        +108 =248 -144
Strelka 3P - Fire 1.31 w32 KLO:          +124 =246 -112
Strelka 3P - Houdini-w32-1cpu (1.03a):    +71 =234 -195
Strelka 3P - Houdini 1.5a (32 bit 1cpu):  +68 =185 -247
I promise not to share.

Edit: You can ask Norm. He sent me FireBird/Fire source code for months which I had and not a single soul saw it until he himself decided to publish it.
PAWN : Knight >> Bishop >> Rook >>Queen

Post Reply