Search found 17076 matches

by bob
Sat Aug 25, 2007 9:07 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: re-inventing the SMP wheel
Replies: 81
Views: 38200

Re: re-inventing the SMP wheel

A much more interesting question would be what is the best place for split points. Should they be concentrated at remaining depth =4, or 8, or ~16 ply? I'm glad this kind of discussion is coming up, because I'm at the point where this is pretty much all I have left to worry about (besides a few bug...
by bob
Sat Aug 25, 2007 3:51 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: re-inventing the SMP wheel
Replies: 81
Views: 38200

Re: re-inventing the SMP wheel

Then you are not using 256K pages. You are using 4K pages. The O/S has to keep up with them as 4k pages, not 256K as well. Because when we share stuff between different processes (two different users executing the same executable such as the shell, or VI or whatever) the code parts are shared. This...
by bob
Sat Aug 25, 2007 1:52 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: BitBoard Tests Magic v Non-Rotated 32 Bits v 64 Bits
Replies: 36
Views: 25065

Re: BitBoard Tests Magic v Non-Rotated 32 Bits v 64 Bits

I had reported this earlier. Magic was no faster than rotated. I switched because of two things... 1. magic is simpler, and simpler is better as I get older. :) 2. Magic gives you the opportunity to update the occupied_squares and then generate moves easily. To do this with rotated bitboards first r...
by bob
Sat Aug 25, 2007 1:47 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: re-inventing the SMP wheel
Replies: 81
Views: 38200

Re: re-inventing the SMP wheel

Been giving SMP some thought. It makes sense to use the extra processors for tree splitting when there is nothing else for them to do. I would think that it would be more efficient and friendlier to the cash if say the evaluation routines and move generators were constructed to be done in chunks by...
by bob
Fri Aug 24, 2007 10:06 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: re-inventing the SMP wheel
Replies: 81
Views: 38200

Re: re-inventing the SMP wheel

You can't mix. How on earth is it going to traverse the page tables when each entry could point to a large page or a small page? You get either/or, but not both. I am not talking about mixing sizes. Just setting different access rights on different 4K pages within a 256KB allocated blok. The page-t...
by bob
Fri Aug 24, 2007 7:06 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: re-inventing the SMP wheel
Replies: 81
Views: 38200

Re: re-inventing the SMP wheel

Believe me, this has not been developed in an ad hoc way over the past 30 years, it has been studied, simulated and measured to death, to get us to where we are today. This does not answer the most important question, though: is the optimum page size dependent on total memory size or not? If not, h...
by bob
Fri Aug 24, 2007 4:16 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: re-inventing the SMP wheel
Replies: 81
Views: 38200

Re: re-inventing the SMP wheel

Well, the eval I have in mind doesn't really use much memory at all. 32KB is huge. Adding a few Piece-Square Tables for each piece type (16 x 64 bytes = 1KB) would not kill me. But playing a match without them will just tell you next to nothing. About paging: When you allocate a small object, you a...
by bob
Fri Aug 24, 2007 12:46 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: re-inventing the SMP wheel
Replies: 81
Views: 38200

Re: re-inventing the SMP wheel

First let's try a match between your non-real program and my real-program, to see if there is any advantage to a program that is optimized to fit L1 rather than optimized to play the best possible game of chess. That's my point about "real programs". To play the game of chess well, you need a prett...
by bob
Thu Aug 23, 2007 8:15 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: re-inventing the SMP wheel
Replies: 81
Views: 38200

Re: re-inventing the SMP wheel

I doubt "real programs" are going to fit in L1. No doubt an intentionally small program can, but that won't predict anything about how the real programs are going to perform, which is where we started this discussion... My definition of a "real program" is at the very least, a program that has no c...
by bob
Thu Aug 23, 2007 6:23 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: re-inventing the SMP wheel
Replies: 81
Views: 38200

Re: re-inventing the SMP wheel

To get back on topic: It still bothers me that at split points there would be multiple threads each generatinfg moves for the same position. So I wonder if there would be some gain in having only one thread, the one that 'owns' the node and has opened the hash entry for it, to generate moves, and p...