TalkChess.com

Texel!

Nobody until at least a year ?

At chess960 it will probably be a very long time before another engine overtakes Houdini. I guess Fire is most likely, but Norman did not have time to add chess960 support to Fire 4, so I don't know if he next version will support it or not.

Modern Times wrote:At chess960 it will probably be a very long time before another engine overtakes Houdini. I guess Fire is most likely, but Norman did not have time to add chess960 support to Fire 4, so I don't know if he next version will support it or not.

For the last few years I've been doing 1 release per year in December.
As you know Ray, I do like Fischer Random very much, and really wish I had had time to include it.

Much of my time with Fire 4 was spent implementing and debugging Thread Local Storage SMP...
which minimizes the amount of 'shared' resources among threads...i.e. each thread maintains data 'local' to that thread.
(It can increase MP speed by reducing the need for synchronization...thread locking/waiting/ etc.)

https://software.intel.com/en-us/articl ... ronization

so, from a memory management perspective, Fire 4 is quite a unique implementation...
(I'm not aware of any other chess engine utilizing this technology, but I did see a post where John Dart indicated he was intending to implement it in Arasan.)
It still has room for improvement as the current implementation is not optimal.

Anyway, I do intend to include FRC support in the next release, probably at end of year again.

Norm

Modern Times wrote:At chess960 it will probably be a very long time before another engine overtakes Houdini.

Yes, Houdini 4 is especially strong at FRC...catching it would be a real challenge.
This leads me to believe it has separate PST and other eval values, used solely in Chess960 'mode'.

If so, this is quite unique. All open-source FRC engines I've seen utilize the same eval for both FRC and normal chess.
(although SF does have an eval term TrappedBishopA1H1, for use only in FRC)

As you know, for development of Houdini 4, RH had access to a large IBM database center in Dubai, where he conducted massive parallel tuning/testing.
It's my theory that he also highly tuned H4 using separate eval terms for use in Chess960 play.

Dr.Wael Deeb wrote:It's hard to guess....

Gull would be an excellent pretender but it's development is postponed for now....

Fire is a great piece of software but even as such it has a long way to go and knock off Houdini to claim the third place....

Equinox would be an option but in my private rating list it's around 75 Elo weaker than Houdini 4....

So who knows

Excuse my English,I am arabian regards,
Dr.D

I thought you were British or similar as you were once very close with Ray.

S.Taylor wrote:

Dr.Wael Deeb wrote:It's hard to guess....

Gull would be an excellent pretender but it's development is postponed for now....

Fire is a great piece of software but even as such it has a long way to go and knock off Houdini to claim the third place....

Equinox would be an option but in my private rating list it's around 75 Elo weaker than Houdini 4....

So who knows

Excuse my English,I am arabian regards,
Dr.D

I thought you were British or similar as you were once very close with Ray.

Nope,wrong number Shimon
Best regards,
Dr.D

kranium wrote: Anyway, I do intend to include FRC support in the next release, probably at end of year again.

Norm

Yes I understand you had other priorities for Fire 4. Good news that it should be back for Fire 5

If it wasn't private there would be a clear number 3: Cryptic.

Werewolf wrote:If it wasn't private there would be a clear number 3: Cryptic.

If they don't want to at least let fellows like Graham include it in their ratings lists which involves testing Cryptic then it doesn't deserve any sort of top-tier recognition in my opinion. There needs to be unequivocal proof that it is in fact the clear number three. Moreover, there is just too much of this derivative/clone stuff that has happened as of late that doesn't serve well for Cryptic, especially since it is private.

I really don't get the whole private thing. Rating lists help to show the strength of chess engines. It would help Cryptic be effectively recognized - and outstandingly respected without any question.