More Participants or More Number of Games?

[bob]

This is a bit of a tangent, but do you think it's possible to OVER-test?

I was just thinking most of the current values in crafty are tested to work near optimally with each other, so any tweak of the old settings makes it worse. When testing additions, it could upset this careful balance and make them appear worse, even if they could potentially be better if all the other settings were tuned with respect to this new addition.

I wonder if a better test-bed for new ideas might contain settings that are more generic, less tuned to work perfectly together. I suppose on the other hand, it could lead one down a lot of roads that are dead ends. I don't know, just rambling...

There are lots of places for interaction, for sure. But most of what we are doing is not taking existing code and tweaking values, it is adding new code, or replacing old code, and tuning the changed code's values...

[Ralf]

Finding decent opponents is not easy. (1) not all run under linux. Most don't. (2) of those that do, many have trouble with fast time controls and lose too many games on time, which doesn't help me at all; (3) of those that are left, many are too weak. You don't learn anything playing programs that are more than 200 Elo weaker.

If these are your requirements for an opponent, you could also try the linux-version of Spike 1.2 Turin (http://spike.lazypics.de/downloads/Spik ... nux.tar.gz). At least we never heard of any problems regarding the time-management or stability. Its not open-source, though.

best wishes
Ralf

[Daniel Shawul]

Don't get on scorpio's way please Honestly, I don't know of stabiltiy issues you are refering to. With the current version I had some issues with the dll which is resolved now but other than that it should work I guess.

DAniel

[Edsel Apostol]

This is a bit of a tangent, but do you think it's possible to OVER-test?

I was just thinking most of the current values in crafty are tested to work near optimally with each other, so any tweak of the old settings makes it worse. When testing additions, it could upset this careful balance and make them appear worse, even if they could potentially be better if all the other settings were tuned with respect to this new addition.

I wonder if a better test-bed for new ideas might contain settings that are more generic, less tuned to work perfectly together. I suppose on the other hand, it could lead one down a lot of roads that are dead ends. I don't know, just rambling...

You have a good point here Matt. Adding new eval feature to an engine might disrupt the balance of the current features and you need to adjust them for the new feature to work.

There should be orthogonality of the features to make sure that they doesn't overlap so as it would possible to test a new feature without worrying that it will disrupt the balance of the old features.

I am talking about eval here but it may also be true when it comes to search.

[bob]

Finding decent opponents is not easy. (1) not all run under linux. Most don't. (2) of those that do, many have trouble with fast time controls and lose too many games on time, which doesn't help me at all; (3) of those that are left, many are too weak. You don't learn anything playing programs that are more than 200 Elo weaker.

If these are your requirements for an opponent, you could also try the linux-version of Spike 1.2 Turin (http://spike.lazypics.de/downloads/Spik ... nux.tar.gz). At least we never heard of any problems regarding the time-management or stability. Its not open-source, though.

best wishes
Ralf

That's yet another problem. If a static lib executable is available, I can certainly give it a try, although not all work since we are always running the very latest of everything here, which can cause an issue with a static lib now and then.

[krazyken]

Don't get on scorpio's way please Honestly, I don't know of stabiltiy issues you are refering to. With the current version I had some issues with the dll which is resolved now but other than that it should work I guess.

DAniel

2.5 I can't get to run at all, always segfaults immediately on starting, 2.0x series crashes about 1 in 3 games at slow time controls, I haven't really tried it for blitz. But I'm trying to run it with 8 threads on Mac OS X not a scenario you've tested I'm sure.

[Daniel Shawul]

I recently tried linux 64 bit and found a bug with lock implmentation. Spinlocks are broken there so you have to use mutexes. Just comment
#define USESPINLOCK in my_types.h and it should be work (i.e for the 2.1 version). I don't know if there are problems on Mac since i did not test on it as you pointed out. If there is a segfault there, it is most probably due to the locks.

DAniel

[frankp]

I recently tried linux 64 bit and found a bug with lock implmentation. Spinlocks are broken there so you have to use mutexes. Just comment
#define USESPINLOCK in my_types.h and it should be work (i.e for the 2.1 version). I don't know if there are problems on Mac since i did not test on it as you pointed out. If there is a segfault there, it is most probably due to the locks.

DAniel

Which distro. How does it manifest itself. (I am using Debian Lenny.).

Frank

[bob]

I recently tried linux 64 bit and found a bug with lock implmentation. Spinlocks are broken there so you have to use mutexes. Just comment
#define USESPINLOCK in my_types.h and it should be work (i.e for the 2.1 version). I don't know if there are problems on Mac since i did not test on it as you pointed out. If there is a segfault there, it is most probably due to the locks.

DAniel

How can a spinlock work on 32 bit linux and not work on 64 bit linux???

[Daniel Shawul]

Red Hat linux 5. But the problem is the posix spinlock implementation is wrong so it won't work without the suggestion i made. Or if you still want to use the spinlocks just change the :"q" (x) to :"q" (&x)

Daniel