1) Can't You put 5 piece TBs to RAM to get very fast access?
2) When You have TBs can't You remove all redundant code from engine to speed up search? You don't need any code about mating with rook, queen and so on.
Jouni
On my cluster tests, which used 12gb nodes, before starting a match, my script did "cat TB/* > /dev/null" to ensure they were all loaded into memory. However, you still have the significant access time for that much memory which works on the TLB, and the decompression time which is also significant... This seemed to make no significant difference over just doing normal I/O. All tests were within 2-3 Elo of each other...
Bitbases might have a chance for improving things since they are far smaller, when I have time I will add them to Crafty and run the test again.
bob wrote:Say 1 game? That is not going to be an Elo changer.
One game in a thousand, or thirty in thirty thousand? I think that will show up. Of course, it might well be less. How confident can you be without testing? But that's just a starting point. If you look two plies deep I would think it would actually be faster using the EGTBs, so the help would be magnified.
bob wrote:Say 1 game? That is not going to be an Elo changer.
One game in a thousand, or thirty in thirty thousand? I think that will show up. Of course, it might well be less. How confident can you be without testing? But that's just a starting point. If you look two plies deep I would think it would actually be faster using the EGTBs, so the help would be magnified.
Just for fun, I will run the test, where it probes only at ply 1, then at ply 2, ... and see what happens...
the "-n" in R03 is the cutoff point for egtb probes. I only probe the egtbs if current ply < n. -1 doesn't probe at all. -2 probes only at ply=1, -3 probes at ply=1 or 2, etc...
Again, 23.4-1 and 23.4-2 are the normal 23.4 versions. R03-n versions all use EGTBs, but only probe when ply < n. As you can see, no real significant trend other than that in general, deeper probing drags the Elo down just a bit, but not significantly...
There just doesn't seem to be any "setting" where EGTBs help... The very shallow limits don't hurt much at all, but also don't help significantly either...
bob wrote:Again, 23.4-1 and 23.4-2 are the normal 23.4 versions. R03-n versions all use EGTBs, but only probe when ply < n. As you can see, no real significant trend other than that in general, deeper probing drags the Elo down just a bit, but not significantly...
There just doesn't seem to be any "setting" where EGTBs help... The very shallow limits don't hurt much at all, but also don't help significantly either...
Just a silly idea, how about doing EGBTs only on extensions?
bob wrote:Again, 23.4-1 and 23.4-2 are the normal 23.4 versions. R03-n versions all use EGTBs, but only probe when ply < n. As you can see, no real significant trend other than that in general, deeper probing drags the Elo down just a bit, but not significantly...
There just doesn't seem to be any "setting" where EGTBs help... The very shallow limits don't hurt much at all, but also don't help significantly either...
Just a silly idea, how about doing EGBTs only on extensions?
Not sure what you mean. I only extend on checks and nothing else...
bob wrote:Again, 23.4-1 and 23.4-2 are the normal 23.4 versions. R03-n versions all use EGTBs, but only probe when ply < n. As you can see, no real significant trend other than that in general, deeper probing drags the Elo down just a bit, but not significantly...
There just doesn't seem to be any "setting" where EGTBs help... The very shallow limits don't hurt much at all, but also don't help significantly either...
Just a silly idea, how about doing EGBTs only on extensions?
Not sure what you mean. I only extend on checks and nothing else...
Ah ok, there is not much point doing it on check extensions, but on pushed pawn or singular extensions it might have some sense.
bob wrote:Again, 23.4-1 and 23.4-2 are the normal 23.4 versions. R03-n versions all use EGTBs, but only probe when ply < n. As you can see, no real significant trend other than that in general, deeper probing drags the Elo down just a bit, but not significantly...
There just doesn't seem to be any "setting" where EGTBs help... The very shallow limits don't hurt much at all, but also don't help significantly either...
Just a silly idea, how about doing EGBTs only on extensions?
Not sure what you mean. I only extend on checks and nothing else...
Ah ok, there is not much point doing it on check extensions, but on pushed pawn or singular extensions it might have some sense.
You only want to probe after a capture, nowhere else. As soon as you reach an EGTB position, you don't want to search further since you have perfect information. And if you are not currently in an EGTB position, the only way to get there is to capture something and take the position down to 3-4-5-6 pieces total, depending on which EGTBs you have.