Gaviota EGTBs: How long to generate 5-men TBs?

Discussion of chess software programming and technical issues.

Moderators: hgm, Rebel, chrisw

Joerg Oster
Posts: 937
Joined: Fri Mar 10, 2006 4:29 pm
Location: Germany

Gaviota EGTBs: How long to generate 5-men TBs?

Post by Joerg Oster »

Does anybody have an estimation?
Generating the 4-men took 229 sec on my Quad Q6600.

My Download speed is only 384kbit/s, so the complete download should take a while. :(
Generating them is probably faster, isn't it?
Martin Thoresen
Posts: 1833
Joined: Thu Jun 22, 2006 12:07 am

Re: Gaviota EGTBs: How long to generate 5-men TBs?

Post by Martin Thoresen »

First I generated all 3-4-5 men and then compressed it all with compression scheme 3 (cp3) with verification enabled.

The total amount of time was something like 30 hours.

This was on my Core i7 980x @ 4303 MHz.
Joerg Oster
Posts: 937
Joined: Fri Mar 10, 2006 4:29 pm
Location: Germany

Re: Gaviota EGTBs: How long to generate 5-men TBs?

Post by Joerg Oster »

Thank you for your quick reply, Martin.
So generating them should be faster on my machine as well.
Jouni
Posts: 3286
Joined: Wed Mar 08, 2006 8:15 pm

Re: Gaviota EGTBs: How long to generate 5-men TBs?

Post by Jouni »

Can You interrupt calculation and continue later?

Jouni
Martin Thoresen
Posts: 1833
Joined: Thu Jun 22, 2006 12:07 am

Re: Gaviota EGTBs: How long to generate 5-men TBs?

Post by Martin Thoresen »

Jouni wrote:Can You interrupt calculation and continue later?

Jouni
Yeah, that's no problem. When you stop the generation and
restart it, it will first check what pieces you already got.
User avatar
jshriver
Posts: 1342
Joined: Wed Mar 08, 2006 9:41 pm
Location: Morgantown, WV, USA

Re: Gaviota EGTBs: How long to generate 5-men TBs?

Post by jshriver »

I forget how long it took me but I did it on an old P4. Think the compression also took a good bit of time. With modern hardware I would suspect it would be a lot quicker especially with smp support.

You can try downloading them from my site to see if it's any quicker.
-Josh

http://olympuschess.com/egtb/gaviota/
Joerg Oster
Posts: 937
Joined: Fri Mar 10, 2006 4:29 pm
Location: Germany

Re: Gaviota EGTBs: How long to generate 5-men TBs?

Post by Joerg Oster »

For those interested:
It took 24h 21 min to generate all 5men EGTBs on my Quad. Tbcheck gave no errors.
Now I only need to compress them, which might take some hours again.

But still much faster than downloading. :D
Sven
Posts: 4052
Joined: Thu May 15, 2008 9:57 pm
Location: Berlin, Germany
Full name: Sven Schüle

Re: Gaviota EGTBs: How long to generate 5-men TBs?

Post by Sven »

Joerg Oster wrote:For those interested:
It took 24h 21 min to generate all 5men EGTBs on my Quad. Tbcheck gave no errors.
Now I only need to compress them, which might take some hours again.

But still much faster than downloading. :D
Hmmm, depends on your download speed. When I downloaded all 5men Gaviota files I was lucky to use a very fast link, and it took me less than 2 hours.

Sven
Milos
Posts: 4190
Joined: Wed Nov 25, 2009 1:47 am

Re: Gaviota EGTBs: How long to generate 5-men TBs?

Post by Milos »

Sven Schüle wrote:
Joerg Oster wrote:For those interested:
It took 24h 21 min to generate all 5men EGTBs on my Quad. Tbcheck gave no errors.
Now I only need to compress them, which might take some hours again.

But still much faster than downloading. :D
Hmmm, depends on your download speed. When I downloaded all 5men Gaviota files I was lucky to use a very fast link, and it took me less than 2 hours.
He stated in his opening post that his downloading speed is 384kb/s which means it would take him roughly 38 hours to download all in the smallest case.
User avatar
M ANSARI
Posts: 3707
Joined: Thu Mar 16, 2006 7:10 pm

Re: Gaviota EGTBs: How long to generate 5-men TBs?

Post by M ANSARI »

This is getting really interesting and maybe we will reach the stage where calculating EGTB's "on the fly" is getting close. I always thought a hardware solution would be a much better and faster option, albeit more expensive. I wonder if today's programmable GUI cards that can be configured as 2 way or 4 way (maybe even more) SLI or Crossfire is a possible option. Or maybe several Tilera card with 100 processors in series. There are still some very tangible gains for improvement of chess play in engines, most notably 7 or 8 piece EGTB's where rook and pawns are involved.