Gaviota 0.74 released

[Spock]

I think all is OK with the x64 exe !

I deleted all tbs and started again, and I now have 98 files generated so far, which is further than before, and task manager showed memory usage at 1,698 at it's peak. Now down to about 900

So, as has been pointed out:
- the generation process must not be interrupted, must go from start to finish
- greater chance of success with the x64 exe

[Dr.Wael Deeb]

Dann, that was my theory also. Gabor is currently generating them, I'm expecting him to have the same issue.

Here is a {preliminary} 64 bit Gaviota binary:
http://cap.connx.com/chess-engines/new- ... ta.exe.bz2

Now, it's just a bare binary, so you will need to download the whole package from Miguel's site first (I do not include any of the auxiliary files here). I guess that the binary will not function without the other things in the Gaviota package.

Also, that binary will only work on 64 bit windows machines with a 64 bit operating system. It will not function on 32 bit Windows versions.

Also, here is a copy of the pre-built 4 man files:
http://cap.connx.com/chess-engines/new-approach/gtb.7z

Hi Dann,
The x64 Gaviota binary archive is not extracting the .exe for me....
It shows ?Kb when opening the archive....
I use Power Archiver 11....
Any idea why this is happening
Dr.D

P.S.The 4 men archive works fine though....

[michiguel]

I think all is OK with the x64 exe !

I deleted all tbs and started again, and I now have 98 files generated so far, which is further than before, and task manager showed memory usage at 1,698 at it's peak. Now down to about 900

That is exactly the behavior I would expect. If you passed that peak, I think you should be able to go all the way.

when you finish, type tbtest 5
It will internally check the longest mates from tablebases.
if you see this:

CHECKING LONGEST MATE POSITIONS FROM TABLEBASES
221 positions checked, 221 correct, 0 wrong
tablebases test --> PASSED

you are done!

Miguel

So, as has been pointed out:
- the generation process must not be interrupted, must go from start to finish
- greater chance of success with the x64 exe

[Spock]

Well I'm close to giving up now. I had about 35GB generated, and when I next came back to the machine, there was lots of disk activity and I just had a blank screen, and the box had vanished from my network. I suspect that the memory usage had jumped dramatically again, making the machine totally non-responsive. I had no option but to hit the reset button. Windows is just running checkdisk on the hard drives now, I'll be able to see soon what may have happened (typing this from another machine)

I'll just have to give up on this I think. I do have a Linux box, but I would have no idea how to do it on that, and to get the files back into the Windows environment

[Dr.Wael Deeb]

Can someone provide me with a link to the x64 version as Dann is ignoring me obviously
Thanks in advance,
Dr.D

[Spock]

My machine is back up, and I have 145 files using 38.2GB. I'll run the tablebase verification and see if they are all OK

[Spock]

OK, I ran tbtest and got the message "218 correct, 3 wrong."
KPPPK, KPPKP, KPKPP are "not found". So we know at what point the tb generation failed. I won't be trying again, will make do with 4-men.

[AdminX]

Can someone provide me with a link to the x64 version as Dann is ignoring me obviously
Thanks in advance,
Dr.D

http://www.tedsummers.com/gaviotax64.zip

[Dr.Wael Deeb]

Can someone provide me with a link to the x64 version as Dann is ignoring me obviously
Thanks in advance,
Dr.D

http://www.tedsummers.com/gaviotax64.zip

Thanks buddy,much appreciated
Dr.D

[michiguel]

OK, I ran tbtest and got the message "218 correct, 3 wrong."
KPPPK, KPPKP, KPKPP are "not found". So we know at what point the tb generation failed. I won't be trying again, will make do with 4-men.

Ray,

I appreciate a lot your tests. I am working now to make sure that the generation goes smoothly in the next version so other people won't deal with similar problems. I fixed already the problem that the generation has to go uninterrupted. This is important because you can stop at any time, and later type tbgen 5 to continue. I am working now to make sure that you need less RAM memory for pawn endgames (saving partial files to disk). In that way, even 32 bit systems would do it well.

I am sorry that it was not perfect the first time, but i did not have a suitable windows system to test (the two windows pc I have do not have a lot of memory). In Linux everything went fine and I assume it would be the same.

Thanks again,
Miguel