Search found 756 matches

by smatovic
Tue Dec 31, 2019 6:20 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: RMO - Randomized Move Order - yet another Lazy SMP derivate
Replies: 114
Views: 14856

Re: RMO - Randomized Move Order - yet another Lazy SMP derivate

If this gets better game results despite worse time to depth, then it must be searching moves that were incorrectly pruned normally. And it seems that this incorrect pruning has been happening often enough to have a measurable impact. So it seems like it would be worth researching what moves get se...
by smatovic
Tue Dec 31, 2019 3:00 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: RMO - Randomized Move Order - yet another Lazy SMP derivate
Replies: 114
Views: 14856

Re: RMO - Randomized Move Order - yet another Lazy SMP derivate

This basically means that your whole world of pencil and paper is an illusion. Perhaps a big part of reason is the chaotic nature of modern search (with all the path dependt effects like history ordering and its effect on lmr etc.). That and the chaotic nature of thread scheduling, becoming chaos^2...
by smatovic
Mon Dec 30, 2019 6:09 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: RMO - Randomized Move Order - yet another Lazy SMP derivate
Replies: 114
Views: 14856

Re: RMO - Randomized Move Order - yet another Lazy SMP derivate

Ups, I forgot to mention the most important part of all parallel searches, search the oldest son first :oops: So... RMO - basics: - communication of TT move and TT scores between threads via Shared Hash Table - all threads follow the same Iterative Deepening loop with same depth - thread 0 performs ...
by smatovic
Mon Dec 30, 2019 5:06 pm
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: RMO - Randomized Move Order - yet another Lazy SMP derivate
Replies: 114
Views: 14856

RMO - Randomized Move Order - yet another Lazy SMP derivate

I implemented ABDADA in an iterative way in Zeta, and it scales well, but for folks who are too lazy for ABDADA, here is another Lazy SMP derivate - RMO. With <= 64 threads it can not compete with ABDADA, but with >= 128 there is not such a big difference, at least on my time to depth benchmarks. RM...
by smatovic
Wed Dec 25, 2019 7:04 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: Programming for superscalar/pipelined/out-of-order/speculative CPUs?
Replies: 2
Views: 860

Programming for superscalar/pipelined/out-of-order/speculative CPUs?

Either I ignored this topic, or there are not many posts on this. I admit, I am not into assembly, but what do you do with C/C++ concerning modern speculative CPUs with multiple execution units? The only thing I can imagine is to unroll loops and avoid branches. Are there any general rules to follow...
by smatovic
Mon Dec 23, 2019 7:51 pm
Forum: Computer Chess Club: General Topics
Topic: RIP Chuck Peddle (designer of the 6502)
Replies: 4
Views: 745

Re: RIP Chuck Peddle (designer of the 6502)

...when I heard that my Amiga 500 which I got later had an 6502 as keyboard
controller it was some kind of sacrilege for me, how can you downgrade something
that causes so much fun to handle keyboard input only...

--
Srdja
by smatovic
Mon Dec 23, 2019 6:51 am
Forum: Computer Chess Club: General Topics
Topic: RIP Chuck Peddle (designer of the 6502)
Replies: 4
Views: 745

RIP Chuck Peddle (designer of the 6502)

Charles Ingerham Peddle died on December 15, 2019. https://hardware.slashdot.org/story/19/12/23/020254/microchip-pioneer-chuck-peddle-lead-designer-of-the-historic-650x-microprocessors-dies https://en.wikipedia.org/wiki/Chuck_Peddle ...without you, there would have been no Atari in my home, thanks. ...
by smatovic
Mon Dec 16, 2019 6:14 am
Forum: Computer Chess Club: General Topics
Topic: The Mac Pro is back - finally
Replies: 19
Views: 3991

Re: The Mac Pro is back - finally

...obviously the new Mac Pro does not play in the same NPS/Dollar league like systems based on recent AMD Zen CPUs for Stockfish, or Nvidia RTX series for LC0, but for Video/CGI the quad-GPU deal seems a good one, you get 4 AMD Vega II GPUs (with each 32 GB HBM) for ~10K $. -- Srdja ... The Vega II...
by smatovic
Sat Dec 14, 2019 7:36 am
Forum: Computer Chess Club: Programming and Technical Discussions
Topic: China boosts in silicon...
Replies: 1
Views: 1251

China boosts in silicon...

You may have read it, China wants to get independent from Western chips and software within 3 years. So what does China have in hands concerning CPUs? Here a little overview... x86-64 - THATIC, a joint venture with AMD https://en.wikipedia.org/wiki/AMD%E2%80%93Chinese_joint_venture x86-64 - Zhaoxin,...
by smatovic
Thu Dec 12, 2019 4:04 pm
Forum: Computer Chess Club: General Topics
Topic: The Mac Pro is back - finally
Replies: 19
Views: 3991

Re: The Mac Pro is back - finally

...obviously the new Mac Pro does not play in the same NPS/Dollar league like
systems based on recent AMD Zen CPUs for Stockfish, or Nvidia RTX series for LC0,
but for Video/CGI the quad-GPU deal seems a good one, you get 4 AMD Vega II GPUs
(with each 32 GB HBM) for ~10K $.

--
Srdja