New J.A Compiles

[pohl4711]

I've updated the Proton link with new Windows & Linux builds. The previous ones had avx512 instructions in all the builds.

Jim.

There seems something wrong here: What makes Velvet special is its 2nd risky-playing-style net. For using it, there is an UCI-command "Style". In offical Velvet releases, you can set it to "Risky" or "Normal". But here, only Normal is allowed, Risky is gone ???

[Jim Ablett]

I've updated the Proton link with new Windows & Linux builds. The previous ones had avx512 instructions in all the builds.

Jim.

There seems something wrong here: What makes Velvet special is its 2nd risky-playing-style net. For using it, there is an UCI-command "Style". In offical Velvet releases, you can set it to "Risky" or "Normal". But here, only Normal is allowed, Risky is gone ???

https://github.com/mhonert/velvet-chess ... d6ef695f42

## Changes

- New default network with increased hidden layer size (2x1024 => 2x1536) trained using new generated training data
- Temporarily removed the "Risky" style net, as a new version has not yet been trained for the increased hidden layer size
- Increased maximum thread count
- Increased maximum hash size
- Some refactorings
- Bugfix for movelists out of bound access, which caused segfault in TCEC test game

Jim.

[Jim Ablett]

Good morning Jim,

now its hard to create a wish for you.
Thanks again for your last compiles.

This one is definitiv very intersting!!

Illumina 3 NN
https://github.com/Mergener/illumina
by Thomas Mergener, Brazil

Best
Frank

I compiled the last dev version a few months back > https://drive.proton.me/urls/7D0ZKKFMS0#WhLWAAIowNT2

Jim.

[Frank Quisinsky]

Hi Jim,

sorry, not saw your compile for Illumina3, great!

Have a nice day!

Best
Frank

[Jim Ablett]

Tomcat 1.0 by Gunnar Harms

Tomcat is 'Bobcat' by Gunnar Harms with enhancements by Norman Schmidt.

chess engine based on Bobcat 8.0 by Gunnar Harms

https://github.com/Bobcat/bobcat

- code optimizations and eval tuning

development tests (ultra-fast/self-play) indicate about a 45 elo improvement

Norman Schmidt firefather@telenet.be 6/26/2017

Tomcat 1.0 64 JA

Code ported to Linux by me.

Windows/Linux & Android profiled builds

Proton link:
https://drive.proton.me/urls/N9HHM9605M#KOi4tD8ZkDPa

Smash link:
https://fromsmash.com/Tomcat10JA

Jim.

[Frank Quisinsky]

Good morning Jim,

I overwork my overview for most of TOP-100 engines (urls, information and direct links about your compiles).
I do that after I not saw that Illumina is compiled by yourself and found a lot of mistakes.
https://www.amateurschach.de/download/_overview.zip ... *.htm file (~ 0,317 Mb)

Two wishes:

1. Cinder (your last compile is version 0.1.3
https://github.com/brunocodutra/cinder

2 Zahak 10.0 (your last compile is version 7.1)
https://github.com/amanjpro/zahak/

If you have time?!
The current dev from Cinder seems to be very interesting, many changes since 0.2.0.

Best
Frank

3. Maybe a new Sapling?
Current version is v1.2.7, your last compile is from 240910 dev.
https://github.com/Timmoth/Sapling

[Frank Quisinsky]

Good morning Jim,

I checked older compiles by yourself.

The following compiles are clearly slower or light slower as the original compiles by the programmers:

1. Avalanche 2.1.0 NN JA
2. BlackCore 6.0 NN JA
3. Carp 3.01 NN JA
4. Odonata 1.0.0 NN JA
5. Rice 8.0.0 NN JA (as dev in your directory, but same version as 8.0.0)
6. Smallbrain 7.0.0 NN JA (as dev in your directory, but the same version as 7.0)
7. Peacekepper 3.01 NN JA (if I compare the original with your compiles ... eval is allways 0,00 in search ... something seems to be wrong).

Best
Frank

[Frank Quisinsky]

Hi Jim,

now that you're cursing me ... hope not!

From the following engine versions are newer compiles online.

1. Bagatur 5.3 (your last compile is 5.2c)
https://github.com/bagaturchess/Bagatur

2. pawnocchio 1.7.2 (your last compile is 1.7.1)
https://github.com/JonathanHallstrom/pawnocchio

3. Sapling 1.2.7 (your last compile is 240910 dev)
https://github.com/Timmoth/Sapling

4. Cinder 0.2.0 (your last compile is 0.1.3) ... but a newer dev could be interesting, see one of my last messages
https://github.com/brunocodutra/cinder

5. And interesting could be Saturn 1.3 (never compiled by yourself)
https://github.com/egormoroz/saturn

Enough is enough ...
After these three messages, I no longer dare to enter the forum.

Best
Frank

But perhaps somebody know the country of the programmer from Iris?

[Jim Ablett]

https://github.com/amanjpro/zahak
Zahak 10.0 by Amanj Sherwany

Zahak 10.0 64 JA

Windows bmi2 build

Proton link:
https://drive.proton.me/urls/434S21BPZC#dx6e8i3uXIOR

Smash link:
https://fromsmash.com/Zahak-100JA

Jim.

[Frank Quisinsky]

Hi Jim,

short information!
Zahak 10.0 NN JA is clearly faster as the compile by the programmer!
Perhaps a bit animation for the programmer of Zahak, gave us so many years a fantastic "endgame" engine.

Great!

Best
Frank