magic bitboard perft

Discussion of chess software programming and technical issues.

Moderators: hgm, Rebel, chrisw

User avatar
Ajedrecista
Posts: 1974
Joined: Wed Jul 13, 2011 9:04 pm
Location: Madrid, Spain.

Re: Magic bitboard perft.

Post by Ajedrecista »

Hello:
NaltaP312 wrote: Sun Apr 12, 2020 10:38 am Yes i have done comparaison just to compare, sorry if this is take for other things.
i don't found gperft for linux on google, if you have a link i will take it
Thanks
The thread at TalkChess dedicated to gperft is here:

gperft

There are versions 1.0, 1.0.1, 1.0.1, 1.0.2 and 1.1 for both Windows and Linux. Disgracefully, all Dropbox links are dead.

There is other Dropbox dead link for version 1.2d here:

Re: Perft using nullmove

I downloaded 1.0 (Windows), 1.0.1 (Windows and Linux), 1.0.2 (Windows and Linux),1.0.3 (Windows and Linux) and 1.1 (Windows and Linux) when those links were active years ago. I tried to attach gperft_1.1_linux.7z to this post but an error message appears:
ERROR
Sorry, the board attachment quota has been reached.
I think this error message is present for everybody for months.

Regards from Spain.

Ajedrecista.
User avatar
hgm
Posts: 27844
Joined: Fri Mar 10, 2006 10:06 am
Location: Amsterdam
Full name: H G Muller

Re: magic bitboard perft

Post by hgm »

That error message has always been present, because the quota is set at 0. TalkChess does not allow attachments.
User avatar
Ajedrecista
Posts: 1974
Joined: Wed Jul 13, 2011 9:04 pm
Location: Madrid, Spain.

Re: Magic bitboard perft.

Post by Ajedrecista »

Hello:
hgm wrote: Sun Apr 12, 2020 9:51 pm That error message has always been present, because the quota is set at 0. TalkChess does not allow attachments.
It has not been always present. Please look at one post by myself from November, 2018, where my upload is still working:

Re: Is Deep Blue source code open?

The size of the attachments were small, otherwise the upload was not allowed. I do not remember when was the change of TalkChess to the present form (i.e. not allowing threading view) but it was for sure before November, 2018 because the other forum software did not allow uploads.

--------------

Regarding gperft uploads, I have just uploaded all gperft I have: 1.0 (Windows), 1.0.1 (Windows and Linux), 1.0.2 (Windows and Linux), 1.0.3 (Windows and Linux) and 1.1 (Windows and Linux).

gperft.7z (1.03 MB)

Mediafire says that it will be online for 14 days counting right now. I hope people can run gperft!

Regards from Spain.

Ajedrecista.
NaltaP312
Posts: 56
Joined: Wed Oct 29, 2008 1:06 pm
Full name: Marc Paule

Re: magic bitboard perft

Post by NaltaP312 »

Hello,
thanks for the link
i have this result
1 thread and depth 7 14.854 seconds
4 thread and depth 7 4 seconds

with this specification
CPU Clock speed – Quad Core 1.1GHz up to 2.2GHz CPU clock
Chipset – Intel Celeron N3450, 64-bit Processor
and DDR3 for RAM

Regards
ibid
Posts: 89
Joined: Mon Jun 13, 2011 12:09 pm

Re: magic bitboard perft

Post by ibid »

Apologies for the dead gperft link(s).

I'm not sure when the files got deleted. With the exception of gperft, I just use dropbox for backups. I'm guessing I deleted everything without thinking at some point when I updated my backups.

In any case, I seem to have free time at the moment so I've been trying to sort out the mess of folders and files that is gperft. It wasn't exactly left in a good state when I last worked on it 5 years ago.

The fastest executable I've found is the 1.2d version someone mentioned. Run times for perft 7 and 8 are 3.722 and 102.852 seconds (windows version, single thread, no hash tables) on my increasingly ancient 3.6 GHz AMD Phenom machine (amazing to me that a 10 year old system is still completely usable!) From memory, the linux version could manage perft 8 in just under 100 seconds, but my linux hard drive failed some years ago so I can't test that right now.

Contrast those times with version 1.1: 6.591 and 178.896 seconds. The 1.2 series is a significant improvement, it just was never quite finished.

Enough rambling. My intent is to distill an up to date version from the mess over the next few days and release it. If I can get the source into a less embarrassing state (there is far more experimental, dead and commented out code than real code right now), I'll include that also.

Sorry again for accidentally breaking the old dropbox links,
-paul