Protector 1.3.2 released

[Frank Quisinsky]

Hello Raimund,

I never test the for me "newer" egbbs. I gave the engine egbbs use it only the directory to find it without looking. Now with your statement I have the information I search since a while ... thanks for it and your time for the answer in detail.

If you like I can send me results by mail.
But I think you have enough results.

My mail address can be find ny profile.

Best
Frank

[noctiferus]

"3) In contrast to Nalimov tablebases the path to the egbbs is a special param which the user has to configure for every engine separately. Also

This is exactly the problem we found with some Toga versions, that had to be modified to allow an option for changing default location.

[BBauer]

Someone said that Eugene doesn't answer any questions about permission to use the table bases. So you may simply ask him if you can use them. If you don't get an answer ...
"qui taket assentiri videtur"

kind regards
Bernhard

[Norm Pollock]

Any chance for a non-SMP win32 version?

[RaimundHeid]

Thanks for the release.
Some questions:
Can protector make use of several threads? How?
Can protector use tablebases?
The makefile doesn't work for me (ubuntu9.10, 64-bit system).
Neither icc nor gcc work. Hundreds of warnings.
regards
Bernhard

I've just upgraded to Ubuntu 9.10 (and kept the libstdc++.so.5). ICC 11.1 compiles without warnings using the Makefile from the Protector release.

Please note that GCC won't compile Protector since it doesn't support inline functions in headers. I have removed support for GCC (=inline functions in separate files) a long time ago because the binaries have been significantly slower than the ones produced by ICC.

Some enhancement to the current situation would be using the ICC profiler. But this has low priority for me since I develop Protector on Windows.

[Volker Pittlik]

...I've just upgraded to Ubuntu 9.10 (and kept the libstdc++.so.5). ICC 11.1 compiles without warnings using the Makefile from the Protector release.

It compiles fine on Ubuntu 8.04 LTS using ICC 10.2 as well.

...Some enhancement to the current situation would be using the ICC profiler. But this has low priority for me since I develop Protector on Windows.

Yes. Although it seems the difference to gcc isn't as big as as it was in the past (at least for 64-bit compilations see: http://www.open-aurec.com/wbforum/viewt ... =4&t=50580) to use the -ipo makes a HUGE difference here:

using the binary coming with the package generates 6619138 moves in 10 seconds.
Using the Makefile as it is resulting in 6356993 moves in the same time.
Adding -ipo to the CFLAGS is resulting in 11206656 moves in the same time!

vp

BTW: search doesn't stop when using "go depth 14" or "go nodes 10000000" thus I used "go movetime 10000" to get comparable results.