Crafty 22.0 Windows x64 available

[Werner]

Hi Peter,
I just started under Arenaan eng-eng match with your build

I HIGHLY recommend Arena. It simply works...

Hi Peter,
it was Arena - and later Winboard -
and now the same with my core2duo...

here Naum 2.2 x64 2CPU got 75% and Crafty used 25% while idle.
And this is my only PC with VISTA 64bit.

[Werner]

Hi Peter,
the best is to delete most of my posts here
- as the 2nd carrige return after the word exit in the .rc file helped.

regards
Werner

[Jim Ablett]

I just updated my Windows x64 Crafty 22.0 build. I reverted back to an earlier Intel compiler version
and was able to use full optimizations again. (I couldn't do this before without the compiled exe crashing.)
Much faster now.

http://www.zshare.net/download/800719515c3eda/

Jim.

[Ómar]

Hi Werner,

Can you please post here the .rc file that did help you at last.
(I don´t understand this: - as the 2nd carrige return after the word exit in the .rc file helped.)

Best regards.

[Werner]

Hi Werner,

Can you please post here the .rc file that did help you at last.
(I don´t understand this: - as the 2nd carrige return after the word exit in the .rc file helped.)

Best regards.

Ok, once more then crafty.rc:

mt=2
egtb
tbpath=J:\tbs
ponder=off
hash=384M
hashp=128M
cache=32M
log=off
exit

as you see here is a free line after the word exit - this is necessary!

and the wrong version was:

mt=2
egtb
tbpath=J:\tbs
ponder=off
hash=384M
hashp=128M
cache=32M
log=off
exit
this was my version before: no free line after the word exit.

Does this help you?

[Ómar]

Thanks Werner but sorry it´s still pondering on its opponents time
Yes I´m using notepad2 and the PS builds and my comp is like yours,
AMD with dual core.
But thanks again for your help.

[rhollay]

Hi Werner,

Can you please post here the .rc file that did help you at last.
(I don´t understand this: - as the 2nd carrige return after the word exit in the .rc file helped.)

Best regards.

Ok, once more then crafty.rc:

mt=2
egtb
tbpath=J:\tbs
ponder=off
hash=384M
hashp=128M
cache=32M
log=off
exit

as you see here is a free line after the word exit - this is necessary!

and the wrong version was:

mt=2
egtb
tbpath=J:\tbs
ponder=off
hash=384M
hashp=128M
cache=32M
log=off
exit
this was my version before: no free line after the word exit.

Does this help you?

On my WindowsXP (32bit) all Crafty 32 bit SMP builds I have regularly use to ponder during opponent's time (Crafty 19.19-smp, 19.20-smp,
v21.6 ....).
I tried all your suggestions, there is CR/LF after every line.
When started from command prompt it says "pondering disabled", still it uses 25% processor time while it's poor Scorpio's time to think.

Probably this behavior is not tied to 64bit builds of new versions.
I cannot find a solution.

(Oh, it's an Intel Core2Duo, Arena 1.99 beta5)

[bob]

Comments:

you have two egtb commands. Only one is needed.

Mode=tournament is a special-case piece of code that is only used if pondering. It basically says "when I get ready to ponder, find the set of valid book moves for my opponent, then generate the set of legal moves for my opponent and remove the book moves. From that set, do a short ponder search to find the best non-book move, and then ponder that."

Since you are not pondering, I would remove that for simplicity.

learn 7 is also the default and should be removed for the same reason, as should "book random 1" as it is also the default.

book on is the default and should be removed

"computer" no longer does anything since crafty's eval is fully symmetric and there is no special-case computer-only code in it.

smpmt=2 and mt=2 are redundant. They do exactly the same thing so just one is needed.

The "timebook 80 8" is speculative and you can use it if you want. Makes crafty use a lot more time on the first 8 moves out of book, which may or may not be good. the "80" says use 80% more time, which might be extreme or not...

I am not following your "not crashing" comment. If it is hanging or crashing when it prints out "terminating SMP processes" then you have the old 22.0 that was replaced with the new smpnice fix because the older version was broken...

I just checked to verify and the current version on the ftp machine is the current version and I've had no reports of it crashing with ponder=off using the smpnice=1 (which is now the default).[/quote]

Very special thanks again! With Peter Skinners download and his minimal .rc everything works fine, adding smpnice=1 makes Crafty use one core at thinking time of opponent up to 25%, as for myself, I can live with this. mt=2 I still have to add to Wb2UCI.eng beside engine name, otherwise Crafty uses only one core even with mt=2 in .rc
best regards
Peter.[/quote]

I don't follow your last sentence. With smpnice=1, which is now the default, crafty will not use _any_ cpu time once it makes a move, as the second process is terminated instantly upon making a move (the message "terminating SMP processes" is produced when this happens).

So I don't understand what you mean by using up to 25%???

[bob]

Hi Peter,
I just started under Arena an eng-eng match with your build
Crafty 22.0 2CPU
Naum 3 2CPU
on my dual AMD 64x2
here is the .rc file
mt=2
egtb
tbpath=J:\tbs
ponder=off
hash=384M
hashp=128M
cache=32M
log=off
exit

if smpnice=1 causes it to crash, that means you are running the original release of 22.0, not the fixed version. I probably should have just released a 22.1 to avoid this confusion, but it is too late now...

I just played 5,000 games with mt=2, po=off, and the default smpnice=1 on my cluster and all 5,000 games finished normally and at no point in time did any node show a load average > 2.0. this was crafty vs crafty, which with ponder=on produces a load average of 4.0 when using mt=2... so it is properly terminating the processes as far as I can determine here...


When I watch Windows task manager Naum 3 gets only 50 % so I stopped the match. The same happens under Winboard.
When I include the command smpnice=1 in the .rc file Crafty is unloaded before it moves...
So I wonder how do you manage to make eng-eng matches on a dual with even condions
What´s my fault now?

[bob]

Hi Werner,

Can you please post here the .rc file that did help you at last.
(I don´t understand this: - as the 2nd carrige return after the word exit in the .rc file helped.)

Best regards.

Here is an explanation of the problem:

If you use the standard windows text editor, the last line in the file does not get a CR/LF appended to it. When Crafty reads the crafty.rc file, the last command goes into the input buffer, but without a CR/LF, it reverts to reading from stdin, and whatever comes in next is jammed against the command without the CR/LF, and it is likely crafty will ignore both since they are unparsable...

if you always put a blank line at the end of your crafty.rc, then the line before that blank line is guaranteed to have the necessary CR/LF, and reading a blank line without the CR/LF won't cause a problem since appending the next command to a blank line makes it parsable since I ignore whitespace...

It is a bug in the windows editor that has been around for years, and a simple blank line on the end solves it completely... The other solution is to use a different editor that doesn't make this particular mistake...