Search found 16821 matches

by bob
Sat May 26, 2007 5:08 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: question about symmertic evaluation
Replies: 37
Views: 21936

Re: question about symmertic evaluation

I don't think so. I am searching about 1.7M nps on my core-2 duo at 2.0ghz. I guess I could use the hardware counters to count the number of instructions, but it has to be way more than 1000 instructions per node. Last time I did use hardware it was around 2700 instructions per node but that was pr...
by bob
Sat May 26, 2007 1:05 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: question about symmertic evaluation
Replies: 37
Views: 21936

Re: question about symmertic evaluation

However, when a program spends > 50% of the total search time inside the evaluation, 2-3 divides don't even show up in the timing and don't affect my NPS at all. Verified by lots of testing of course... I thought Crafty was in a ~1000 cycles range per node. Thus 2-3 times 42 cycles don't looks negl...
by bob
Thu May 24, 2007 8:20 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: question about symmertic evaluation
Replies: 37
Views: 21936

Re: question about symmertic evaluation

I'd be willing to bet that a plain divide will not noticably affect his overall search speed. And it will solve this correctly. Agreed. Furthermore, a decent optimizing compiler will code a power of two division as a right shift where it is possible (and correct) to do so. except mine are not power...
by bob
Thu May 24, 2007 8:18 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: question about symmertic evaluation
Replies: 37
Views: 21936

Re: question about symmertic evaluation

The latency of divide instructions is such that they almost always cause a hefty pipeline stall. The re-order buffer is simply not large enough to hide the latency, even if it is not on a critical path. The divide cannot be retired before it has finished, and this means that it will stall the retir...
by bob
Thu May 24, 2007 8:13 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: question about symmertic evaluation
Replies: 37
Views: 21936

Re: question about symmertic evaluation

The thing is it interlaces nicely with other instructions around it... I have been using several divides in my code to scale parts of the evaluation, and it made absolutely no difference in my NPS... K8 32-bit idiv is 42 cycles vector path. I can't imagine that this is noise in Crafty. As mentioned...
by bob
Thu May 24, 2007 12:01 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: question about symmertic evaluation
Replies: 37
Views: 21936

Re: question about symmertic evaluation

I'd be willing to bet that a plain divide will not noticably affect his overall search speed. And it will solve this correctly. Idiv is still quite expensive (idiv r32/imm ~18-42 cycles on core2 duo) - but if seldomly used it does not hurt. Anyway, vc2005 seems aware of that trick: int idiv8(in...
by bob
Wed May 23, 2007 9:51 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: question about symmertic evaluation
Replies: 37
Views: 21936

Re: question about symmertic evaluation

One idea is to avoid the asymmetrical arithmetical shift right 3 at all and to mul all other eval aspects by 8. You may add 7 if the value to shift arithmetical right 3 is negative, but this is of course more expensive. symmetrical arithmetical shift right 3 ::= (x + ((x>>31)&7)) >>...
by bob
Wed May 23, 2007 9:50 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: question about symmertic evaluation
Replies: 37
Views: 21936

Re: question about symmertic evaluation

just don't use shift as it is not correct for positive and negative numbers.
by bob
Mon May 21, 2007 9:47 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Speedup with bitboards on 64-bit CPUs
Replies: 49
Views: 36406

Re: Speedup with bitboards on 64-bit CPUs

[This is far faster at a K7 than crafty at a 64 bits chip. Vincent Ofcourse, you can say anything you want because we have no way of testing Crafty against your engine. May be you and Hyatt should do a 100 game online match at ICC at a decent time control. The statement makes no sense. That implies...
by bob
Mon May 21, 2007 5:34 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Speedup with bitboards on 64-bit CPUs
Replies: 49
Views: 36406

Re: Speedup with bitboards on 64-bit CPUs

<sigh>

comments from someone who knows little about bitboards doesn't provide much useful information...