Stockfish bench

[kgburcham]

Thanks to all that helped me to learn this bench.

[Eelco de Groot]

thanks Eelco
I am going to go back and look at your post and try to find a more simple way to run this bench but maybe there is not.
Now I will try with the 12 thread SR-2

6 threads Intel 980x
HT off

Total time : 4743
Nodes searched: 26154509
Nodes/second : 5514338

less hash more depth
Nodes/second : 8730201

Hi Kim,

that looks great! Your SR2 is almost exactly ten times as fast as 32-bits Stockfish 2.1.1 my old Athlon computer! This was just 6 processors of your twelve do I understand that right, so you could go faster Sorry for bumping in on the thread there but I must confess I think I only run the SF benchtest once or maybe twice before myself If I change the functionality of the program the bench will give different results after each change anyway, so it was not so important to me. I think Dan's advice of making a batchfile is really the easiest way of going about it. That stockfish-bench.bat you should be able to just double-click, and no more frustration with the spaces in filenames that DOS can not handle. And Jim Ablett's link would be another way of quickly going to the right folder as an alternative, in Windows 7. Thanks for all the useful information guys!

The procedure should be easier now because in the newest mastercode that Marco has published, bench can now be given from the command line, and if you like you can give it a FEN or the current position, and also run it multiple times, for instance to see if there is a start-up delay (running the command line inside the folder is better, from G: I had problems the first run). I have not tested Marco's new code myself and since Marco made this about a month ago he then had to fix a couple of bugs, because he forgot between runs the TT should be cleared, but if you try any build after 31-03-2012 it should be able to do bench from within Stockfish, the commit is from april 1st:

Process 'bench' also from SF prompt

It is possible to start with 'stockfish', then from
command prompt type 'bench' and SF will do what you expect.
Old behaviour is anyhow preserved. As a bonus we can now
start from command line any UCI command understood by
Stockfish. The difference is that after execution of a
command from arguments SF quits, while at the end of the
same command from prompt SF stays in UCI loop.

No functional change.

Signed-off-by: Marco Costalba

and the last fix in it is from april 26:

Clear TT before running 'bench'

Now that we can call bench multiple times
from command prompt we need to ensure searched
nodes remain constant across different runs.

Spotted by Bálint Pfliegel.

No functional change.

Signed-off-by: Marco Costalba

but I have not seen much of the latest master-code of april being compiled yet, at least no public compiles. If you wait for an official update at least you know it has been tested, it is legal because it follows the GPL and if Jim Ablett does it it is fast!

Regards, Eelco

[kgburcham]

This was just 6 processors of your twelve do I understand that right, so you could go faster

Dont be sorry, I read most of your posts.
That 6 core test was a single socket 980x.
A different pc.
I am trying on my 12 core, stock clock went ok but 12x4.5 is slow for some reason.
I will try again.
thanks for all the info.

[lucasart]

what a noob... 4 pages in this thread to understand how to open a command line...

[kgburcham]

I knew that was coming.
probably several mad over this thread.
I expected it when I started, the motivation to finally learn this exceeded my worry about being called a noob.

Have a good day Lucasart (Lucas Braesch)