Tucano 3.00 Release

[sedicla]

Hello,

There is a new Tucano engine available. Should be about 30 elo points stronger than 2.00.

Changes are mostly code review, improved pruning at search, some evaluation tuning and a time management bug fix.

Next step is to keep working on improving Tucano, as always

Hope you have fun !

Download available from https://sites.google.com/site/tucanochess/

[JuLieN]

Hello,

There is a new Tucano engine available. Should be about 30 elo points stronger than 2.00.

Changes are mostly code review, improved pruning at search, some evaluation tuning and a time management bug fix.

Next step is to keep working on improving Tucano, as always

Hope you have fun !

Download available from https://sites.google.com/site/tucanochess/

Congrats Alcide! Your release pace is hard to cope with! Tucano 2 is doing very well in my Mac engines tournament. It's already very strong!

[Gerd Isenberg]

Hello,

There is a new Tucano engine available. Should be about 30 elo points stronger than 2.00.

Changes are mostly code review, improved pruning at search, some evaluation tuning and a time management bug fix.

Next step is to keep working on improving Tucano, as always

Hope you have fun !

Download available from https://sites.google.com/site/tucanochess/

Hi Alcides,

I recognized your now reduced magic lookup table for bishops for max 9 bit occupancies:

Code: Select all

U64     bishop_attack_table[64][512];

But why do you still use variable shift for that memory layout?

You may initialize the tables that way, to shift always by const 64-9 (bishop) or 64-12 (rook) even for lines with less populated occupancies, which safes not only the lookup for the variable shift, but "frees" register cl (ex, ecx, rcx).

Gerd

[sedicla]

Hi Alcides,

I recognized your now reduced magic lookup table for bishops for max 9 bit occupancies:

Code: Select all

U64     bishop_attack_table[64][512];

But why do you still use variable shift for that memory layout?

You may initialize the tables that way, to shift always by const 64-9 (bishop) or 64-12 (rook) even for lines with less populated occupancies, which safes not only the lookup for the variable shift, but "frees" register cl (ex, ecx, rcx).

Gerd

Hi Gerd, thanks for updating the wiki page already. I was surprised

I will take a close look. I didn't paid attention to this detail. First make it work and then improve.

Once again I really appreciate you took the time to check it and also the wiki page update.

Thanks.

[Graham Banks]

Hello,

There is a new Tucano engine available. Should be about 30 elo points stronger than 2.00.

Changes are mostly code review, improved pruning at search, some evaluation tuning and a time management bug fix.

Next step is to keep working on improving Tucano, as always

Hope you have fun !

Download available from https://sites.google.com/site/tucanochess/

Thanks Alcides.