Winboard Forum

Xpdnt 080209 by David B. Weller
http://www.debog.com/ai/index.html

Windows 32 bit Intel compiler 10 p.g.o build

http://www.zshare.net/download/738016460671df/

Jim.

I have two questions for you. Do you have any idea about the latest changes in Typhoon? Is there any chance for a PGO compile of Phalanx?

Hi Tony,

I have two questions for you. Do you have any idea about the latest changes in Typhoon? Is there any chance for a PGO compile of Phalanx?

Scott Gasch is currently converting Typhoon's code to make it 64-bit compatible
so I can compile it, but it's gonna take a bit of time. Recent updates of
Typhoon have been in improving the eval. Scott has been running Typhoon on a
cluster at work using a specially written program to gather evaluation data
(that's the 'dna' file you may have noticed).

Last time I tried a fast compile of Phalanx, there was no real improvement
over the msvc++/cygwin compiles. I'll try again and see if I can squeeze a bit
more speed out.

regards,
Jim.

Jim, thanks for the quick reply..I had no idea what the DNA file was, I was thinking more along the lines of computer based bio-intelligence that evolves. I asked about Phalanx because I am sad to see it going down through my divisions. If you get a 100% speed increase it might stay in third division for another year Phalanx was one of the first engines along with crafty, yace and Crazybishop that I got to work under winboard.

Jim Ablett wrote:Hi Tony,

I have two questions for you. Do you have any idea about the latest changes in Typhoon? Is there any chance for a PGO compile of Phalanx?

Scott Gasch is currently converting Typhoon's code to make it 64-bit compatible
so I can compile it, but it's gonna take a bit of time. Recent updates of
Typhoon have been in improving the eval. Scott has been running Typhoon on a
cluster at work using a specially written program to gather evaluation data
(that's the 'dna' file you may have noticed).


Last time I tried a fast compile of Phalanx, there was no real improvement
over the msvc++/cygwin compiles. I'll try again and see if I can squeeze a bit
more speed out.

regards,
Jim.

It's nice to hear news that Scott Gaksch (another nice Gaksch )is finally making it 64-bit compatible code. I'll be waiting for this.

Since Thomas mentioned Phalanx, I too haven't haven't had luck building a stable compile on ICC. It compiles but it doesn't make a move when I first tested it - src from SF.bet.. So I just used your compile instead. I'm curious as well, since when I built it in MSVC, it worked. I didn't paid much attention then as I was still a beginner. Maybe I'll try it again.

But the good news was your new Romichess P3k with personality settings. Very nice Master Jim. It's very similar to your Toga-per. I'm more inclined to like Romichess because of another added feature (aside from its good learning algorithm). I'm still training my Obender with learning and see how it pairs with Romichess. But I guess Romichess is indeed a formidable foe.

Thanks again and still waiting for your next builds.

Denis

Hi Denis,

I too haven't haven't had luck building a stable compile on ICC. It compiles but it doesn't make a move

I suspect that there is nothing wrong with your Intel Phalanx builds, it's
just that you need to turn off 'polling input' with the msvc/intel compiles.

You can can either use the switch below (command line) or disable it entirely in the src before compiling >

Code: Select all

phalanx -c+ -o- -b- -f10 < test.fin | tee result
    ( Where: -c+ .... use cpu time
             -o- .... don't use polling input              < add this
             -b- .... no opening book
             -f10 ... fixed time 10 seconds per move )


regards,
Jim.

Thanks Jim. I'll try using this tip.