Search found 1917 matches

by mar
Sat Jun 15, 2019 7:51 am
Forum: Computer Chess Club: General Topics
Topic: Pre-Release McCain X4
Replies: 22
Views: 1026

Re: Pre-Release McCain X4

MikeB wrote:
Sat Jun 15, 2019 1:18 am
I don't mind making mistakes - that's how you learn right? I will have it fix in the release version.
Yes, of course. My last remark wasn't useful or necessary, so I apologize.
by mar
Fri Jun 14, 2019 2:40 pm
Forum: Computer Chess Club: General Topics
Topic: Pre-Release McCain X4
Replies: 22
Views: 1026

Re: Pre-Release McCain X4

I already hinted to a possible cause in one of the former McCain threads: http://talkchess.com/forum3/viewtopic.php?f=2&t=69849 From what I have googled, e. g. uint_fast8_t must NOT necessarily be 8 bits, it can be larger ... See https://stackoverflow.com/questions/35055042/difference-between-uint8...
by mar
Fri Jun 14, 2019 9:31 am
Forum: Computer Chess Club: General Topics
Topic: Pre-Release McCain X4
Replies: 22
Views: 1026

Re: Pre-Release McCain X4

Probably needs an aligned allocation. Objects can be different sizes because of the OS and compiler. For instance, with GCC on a 64 bit Linux, long is 8 bytes but on 64 bit windows it is 4 bytes. I didn't even look at the code yet so this is a guess. Strange that an assert was used since it sounds ...
by mar
Sat Jun 08, 2019 7:22 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Hash collision?
Replies: 42
Views: 1409

Re: Hash collision?

As to Martin's question: the flaw is that in every other hash slot you will have exactly the same problem. So it doesn't help you that I-don't-know-how-many bits in the key first direct you to a bucket where only positions that have the same value for those bits go. Wherever you end up, there will ...
by mar
Sat Jun 08, 2019 3:32 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Hash collision?
Replies: 42
Views: 1409

Re: Hash collision?

I just ran a quick simulation and you were right, the number of probe collisions is independent of TT size, so for 16 bits of hash stored in the TT in 8-bucket scheme, I get a false hit for 1 in 8k probes... so only 13 bits I must really suck at math, but where's the flaw? I thought that increasing ...
by mar
Sat Jun 08, 2019 2:53 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Hash collision?
Replies: 42
Views: 1409

Re: Hash collision?

Collisions in the sense of two positions mapping to the same entry are unavoidable. They cannot cause strang evals and moves as long as the engine detects them as different. So what is important is the ability of the engine to detect that two different positions mapping to the same entry/bucket are...
by mar
Thu Jun 06, 2019 12:48 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Hash collision?
Replies: 42
Views: 1409

Re: Hash collision?

That's a nice idea, so this way it would be possible to store 24 bits of hash in a 8-byte TT entry. Not bad at all. Maybe the move could be encoded in even less bits, I like the idea of always encoding target delta; for queens that would be 14 positions so still fits in 4 bits For pawn promotions, t...
by mar
Wed Jun 05, 2019 10:33 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Hash collision?
Replies: 42
Views: 1409

Re: Hash collision?

When I think about it, you could do 8-byte TT entry while storing more than 16 bits of hash part: 16 bits hash 16 bits value 16 bits move (+1 bit free for say lsbit of bounds: 12 bits to encode from/to, 2 bits to encode promotions and 1 bit for castling; this bit can be saved too if you don't care a...
by mar
Wed Jun 05, 2019 10:04 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Hash collision?
Replies: 42
Views: 1409

Re: Hash collision?

Really? I'm not an expert by any means, but a 16 bit key will give you many collisions each second, so it is very likely that this is the problem. The idea is that you store say n bits as hash in the table and use another m bits to compute the TT index. Unless I miscalculated, a 4MB hash table with...
by mar
Wed Jun 05, 2019 11:01 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: strategies for finding slowdows in lazy smp
Replies: 25
Views: 913

Re: strategies for finding slowdows in lazy smp

The dramatic slow-down is probably because other things were running on it (e.g. the chrome browser). First of all, don't mess with affinity (especially if you don't understand how it works). Let's say your CPU has 2 logical cores per one physical, so if you set affinity mask for one worker to bit ...