Elo gains of Jim Albett Compiles against generic compiles

[mhalstern]

I think that it is awesome that our community has Jim Albett to make fast compiles.

I'm curious as to how much elo gain can be had from his compiles.

What if we take every engine that he has compiled and have it play a long match against a generic default compile from the same source code. What would the results be?

Has anybody done this? How much stronger, in terms of ELO are his compiles?

[swami]

Hello Marc,

I think that it is awesome that our community has Jim Albett to make fast compiles.

I'm curious as to how much elo gain can be had from his compiles.

What if we take every engine that he has compiled and have it play a long match against a generic default compile from the same source code. What would the results be?

Has anybody done this? How much stronger, in terms of ELO are his compiles?

Jim's compiles are certainly much faster.

I've tested Daydreamer normal compile and Daydreamer JA compiles and noticed much improvement in latter in STS scores:

How much elo improvement with JA over normal compile is open to discussion, but I believe that in case of Daydreamer, it's atleast 15-20 elo.

Daydreamer 1.6: (491/800)
http://www.talkchess.com/forum/viewtopi ... 02&t=31200

Daydreamer 1.6 JA: (505/800)
http://www.talkchess.com/forum/viewtopi ... 61&t=31200

[maxchgr]

Does Jim do something special when compiling them?

[rvida]

Does Jim do something special when compiling them?

I think he has a good understanding of optimization related compiler switches and he does PGO.

[Sedat Canbaz]

Yes...Jim Ablett is doing a great job,but we should not forget Dann Corbit'compiles too,he is also making a big work

For example,currently Stockfish 1.6.2s,Stocco.bin (Stockfish 1.6.2s x64 4CPU compiled by Dann Corbit) has the highest Stockfish ranking:
http://www.talkchess.com/forum/viewtopi ... 61&start=0



Best Regards,
Sedat Canbaz