New Engine: Avalanche

[Yinuo Huang]

Hello CCRL community!
I've been working on a new chess engine for around 2-3 months and I would like to test it on CCRL, if possible
Yesterday I released the first version where it can play decent matches and defeat me

Its name is Avalanche and it is the first chess engine written in the Zig Programming Language (still a very young language, but it works).
It uses the UCI protocol. It currently does not have any UCI options, but the Transposition Table is hardcoded to 32MB, FYI.
I am expecting it to be rated around [2350, 2450] on CCRL when testing against some other 2400-rated engines on CCRL locally.

It is open-source on Github: https://github.com/SnowballSH/Avalanche.
Some binaries are available in the release page (https://github.com/SnowballSH/Avalanche ... s/tag/v0.2), but the Windows and Linux binaries are cross-compiled on my Mac so it might be slow. But if you don't want to bother downloading Zig (as few people use it) you can just use my binaries.

Please let me know if there is anything I can do.
Thank you!

[Yinuo Huang]

I hope Guenther can see this,
The website says I do not have permission to send PMs so I will send it here.
Can you try this binary? I think it is more general now.
https://github.com/SnowballSH/Avalanche ... indows.exe

Thanks!

[Gabor Szots]

I have started a tournament with Avalanche 0.2. The results will be up by next week's update.

[Yinuo Huang]

Thank you very much!

[Graham Banks]

http://tlcv.net/16074

Gauntlet for CCRL 40/15.

[Yinuo Huang]

Hi,
It seems like the live board is stuck with 0 time on Avalanche’s side. Is this an issue of Avalanche or the broadcast?

[Graham Banks]

Hi,
It seems like the live board is stuck with 0 time on Avalanche’s side. Is this an issue of Avalanche or the broadcast?

Have you tried refreshing?

[Yinuo Huang]

Hi,
It seems like the live board is stuck with 0 time on Avalanche’s side. Is this an issue of Avalanche or the broadcast?

Have you tried refreshing?

It’s working now, thanks.

[Yinuo Huang]

I'm probably updating way too fast here, but here is a bug-fix improvement version 0.2.1 that is ~70 ELO higher than v0.2
https://github.com/SnowballSH/Avalanche ... tag/v0.2.1
As always, binaries in the binaries folder.

You don't have to test it if you feel too busy, I just want to put this out here.
Thanks!

[Graham Banks]

Any chance of a Windows exe for modern computers?

[Yinuo Huang]

Hi,
The only SIMD instruction I’m using right now is POPCNT, and there doesn’t seem to be a big speed difference…
I think until I use more SIMD instructions, a basic binary is fine.
But if you really need one, can you tell me the name of the CPU (I.e. skylake etc.) so I can put it as an argument (zig has a list of CPU names it can optimize for) to the compiler?
Thanks!

[Graham Banks]

Hi,
The only SIMD instruction I’m using right now is POPCNT, and there doesn’t seem to be a big speed difference…
I think until I use more SIMD instructions, a basic binary is fine.
But if you really need one, can you tell me the name of the CPU (I.e. skylake etc.) so I can put it as an argument (zig has a list of CPU names it can optimize for) to the compiler?
Thanks!

One is a Haswell and the other is a Ryzen 5950x.
I use avx2 on both.

[Yinuo Huang]

Hi,
I compiled a windows binary for the Haswell CPU for x86_64. I could not find the Ryzen target, probably due to Zig's current lack of support for cross-compiling to AMD-type CPUs

Can you try if this works for you? https://github.com/SnowballSH/Avalanche ... aswell.exe

Thanks.

[Yinuo Huang]

Also by the way, I added a (hardcoded) implementation for UCI option for "Hash", set default to 32 MB but can be set to anywhere within [1, 4096].
The engine performs slightly differently based on hash size, so my request is that please set it to a power of 2.

Thanks!

[Graham Banks]

Also by the way, I added a (hardcoded) implementation for UCI option for "Hash", set default to 32 MB but can be set to anywhere within [1, 4096].
The engine performs slightly differently based on hash size, so my request is that please set it to a power of 2.

Thanks!

I use 256mb hash for my 1CPU 40/15 testing, and 1024mb hash for 4CPU.

[Graham Banks]

Also by the way, I added a (hardcoded) implementation for UCI option for "Hash", set default to 32 MB but can be set to anywhere within [1, 4096].
The engine performs slightly differently based on hash size, so my request is that please set it to a power of 2.

Thanks!

Your exe works on both computers. Thanks.

[Yinuo Huang]

Cool!
I'll start working on v0.3 now, still a lot of improvements to do!
I feel like this is the fun part of writing an engine. When writing standard CLI apps or servers, you are always fixing new bugs. Here, you are always adding new features and improve.
Thanks for everything

[Gabor Szots]

Hi,
I compiled a windows binary for the Haswell CPU for x86_64. I could not find the Ryzen target, probably due to Zig's current lack of support for cross-compiling to AMD-type CPUs

Can you try if this works for you? https://github.com/SnowballSH/Avalanche ... aswell.exe

Thanks.

Could you please include the version number in the file name as well?

[Yinuo Huang]

I will try to modify my bash script to do that starting the next commit, thanks for the suggestion.

[Yinuo Huang]

Hi,
I see an Amteur chess engine tournament (Avalanche seems to be in div. 8) is coming up, so I want to save my progress so far here.
Can you use v0.2.2 https://github.com/SnowballSH/Avalanche ... tag/v0.2.2 for the division 8 tournament? It is stronger and more stable.
Thanks!

Binaries: https://github.com/SnowballSH/Avalanche ... b/binaries

[Graham Banks]

Hi,
I see an Amteur chess engine tournament (Avalanche seems to be in div. is coming up, so I want to save my progress so far here.
Can you use v0.2.2 https://github.com/SnowballSH/Avalanche ... tag/v0.2.2 for the division 8 tournament? It is stronger and more stable.
Thanks!

Binaries: https://github.com/SnowballSH/Avalanche ... b/binaries

Okay. You might be in Division 7 depending on how the current gauntlet goes.

[Yinuo Huang]

Hi,
I see an Amteur chess engine tournament (Avalanche seems to be in div. is coming up, so I want to save my progress so far here.
Can you use v0.2.2 https://github.com/SnowballSH/Avalanche ... tag/v0.2.2 for the division 8 tournament? It is stronger and more stable.
Thanks!

Binaries: https://github.com/SnowballSH/Avalanche ... b/binaries

Okay. You might be in Division 7 depending on how the current gauntlet goes.

Cool! Do you mind giving me the broadcast link (if possible) for the gauntlet?
It’s fun to watch it play during my spare time.
Thank so very much ))

[Graham Banks]

Hi,
I see an Amteur chess engine tournament (Avalanche seems to be in div. is coming up, so I want to save my progress so far here.
Can you use v0.2.2 https://github.com/SnowballSH/Avalanche ... tag/v0.2.2 for the division 8 tournament? It is stronger and more stable.
Thanks!

Binaries: https://github.com/SnowballSH/Avalanche ... b/binaries

Okay. You might be in Division 7 depending on how the current gauntlet goes.

Cool! Do you mind giving me the broadcast link (if possible) for the gauntlet?
It’s fun to watch it play during my spare time.
Thank so very much ))

http://tlcv.net/16072

[Yinuo Huang]

Hi, I'm back after around 1.5 months of coding!
Since v0.2.2, I felt the code is too messy and impractical, so I erased most files and started a rewrite from scratch.
It is now released as v1.0.0!

The release and information is here: https://github.com/SnowballSH/Avalanche ... tag/v1.0.0
The binaries are here: https://github.com/SnowballSH/Avalanche ... e/binaries

I am expecting a ~130 ELO gain and v1.0.0 be around 2760 ELO.
It would be awesome if you can test this new version. And please let me know if you find any issue running the engine, as I have not tested it on a windows machine.
Thank you!

[Guenther]

Hi, I'm back after around 1.5 months of coding!
Since v0.2.2, I felt the code is too messy and impractical, so I erased most files and started a rewrite from scratch.
It is now released as v1.0.0!

The release and information is here: https://github.com/SnowballSH/Avalanche ... tag/v1.0.0
The binaries are here: https://github.com/SnowballSH/Avalanche ... e/binaries

I am expecting a ~130 ELO gain and v1.0.0 be around 2760 ELO.
It would be awesome if you can test this new version. And please let me know if you find any issue running the engine, as I have not tested it on a windows machine.
Thank you!

The new binary for win-64 does not work anymore on my old hardware with Win7-64 Ultimate.
(core2 = yorkfield)

Code: Select all

---------------------------
Avalanche_x86_64_windows_1.0.0.exe - Entry Point Not Found
---------------------------
The procedure entry point RtlWaitOnAddress could not be located in the dynamic link library ntdll.dll. 
---------------------------