New J.A Compiles

[Jim Ablett]

Dear Jim,
I tried to download some of your compiles (Stash and Vault), but my Windows defender stopped both downloads
Virus detected Trojan:Win32/Wacatac.B!ml
I tried again with the same result.
Sincerely Yours
Viktor

Hi Viktor,

Other people have reported this issue as well. I don't know what is causing it. Downloads are completely safe.
You can add an exception in Windows defender to allow the download.

Here are the virustotal reports >

https://www.virustotal.com/gui/file/2d1 ... ?nocache=1

https://www.virustotal.com/gui/file/b81 ... ?nocache=1

Jim.

[Sylwy]

Vault_041_64_ja on my system:

-bmi2 clang version:



-bmi2 gcc version:



Clang version is incomparable faster.

[Jim Ablett]

Simbelmyne 1.7.0 c612130 by Sam Roelants
https://github.com/sroelants/simbelmyne

Simbelmyne 1.7.0 c612130 64 ja

Windows/Linux & Android builds.

https://drive.proton.me/urls/8MX6XDKWSR#YoQGiRA5gNi3

The Android build works for me ok in Droidfish, but works incorrectly
in Chess For Android.

Jim.

[Guenther]

Dear Jim,
I tried to download some of your compiles (Stash and Vault), but my Windows defender stopped both downloads
Virus detected Trojan:Win32/Wacatac.B!ml
I tried again with the same result.
Sincerely Yours
Viktor

At least Vault is additionally compressed with UPX 4.23 this makes it suspicious for quite some virus scanners.
I guess Jim wants to save space for his cloud storage.

I have hundreds of unreleased versions/unknown programs compiled here, not sure if I should make them available too ;)

[Guenther]

...

Hi Guenther,

That's strange. What about the generic clang version ?

Jim.

Surprisingly (or not really, because I believe you pgo'ed vs. bench?) with bench it's a bit different.
The clang bench is a bit faster than the gcc bench 4.86x sec. vs. 5.55x sec. (several runs), but as the diff
in start pos is so overwhelming I won't trust the bench over real games here.
Probably a test of both versions over lots of games (10000s') would reveal more truth.

Guenther

P.S.
Well, the original 0.4b which has just two non-game playing relevant commits less, even has a better bench
for both avx2 and bmi2 of around 4.2x sec -4.3xsec for both. (but 5-6% lower nps for startpos, not conclusive)

[Guenther]

...

Hi Guenther,

That's strange. What about the generic clang version ?

Jim.

Surprisingly (or not really, because I believe you pgo'ed vs. bench?) with bench it's a bit different.
The clang bench is a bit faster than the gcc bench 4.86x sec. vs. 5.55x sec. (several runs), but as the diff
in start pos is so overwhelming I won't trust the bench over real games here.
Probably a test of both versions over lots of games (10000s') would reveal more truth.

Guenther

P.S.
Well, the original 0.4b which has just two non-game playing relevant commits less, even has a better bench
for both avx2 and bmi2 of around 4.2x sec -4.3xsec for both. (but 5-6% lower nps for startpos, not conclusive)

Out of curiosity I did a test between both versions and the result was even more significant than expected
for a 22-23% nps increase for the gcc version in searching from startpos.

Not sure though I have ever seen that bench and search from startpos contradict for the same binary.
Usually I only decide via search from startpos with go infinite (or go depth XX) runs, which of my compilations (or from other people)
are faster.

Code: Select all

CuteChess GUI EXPER 3.5 (alweys fork branch - enhanced & compiled by me)
tc 10+0.2, 6plies start pgn,  16MB hash, only draw and Syzygy5 adjudication, 20ms margin
concurrency 3, Ryzen7 5700G, 32GB Ram, 8 cores (16 threads) @3.8 Ghz
adj. rules: 120/10/0.10 - 200 moves max

Code: Select all

Score of Vault_041clang-64 vs Vault_041gcc-64: 290 - 608 - 1102 [0.420]
...      Vault_041clang-64 playing White: 172 - 244 - 584  [0.464] 1000
...      Vault_041clang-64 playing Black: 118 - 364 - 518  [0.377] 1000
...      White vs Black: 536 - 362 - 1102  [0.543] 2000
Elo difference: -55.7 +/- 10.2, LOS: 0.0 %, DrawRatio: 55.1 %
2000 of 2000 games finished.

Player: Vault_041clang-64
   "Draw by 3-fold repetition": 593
   "Draw by adjudication": 55
   "Draw by adjudication: SyzygyTB": 367
   "Draw by adjudication: maximal game length": 7
   "Draw by fifty moves rule": 72
   "Draw by stalemate": 8
   "Loss: Black mates": 121
   "Loss: Black wins by adjudication: SyzygyTB": 123
   "Loss: White mates": 166
   "Loss: White wins by adjudication: SyzygyTB": 198
   "Win: Black mates": 47
   "Win: Black wins by adjudication: SyzygyTB": 71
   "Win: White mates": 75
   "Win: White wins by adjudication: SyzygyTB": 97
Player: Vault_041gcc-64
   "Draw by 3-fold repetition": 593
   "Draw by adjudication": 55
   "Draw by adjudication: SyzygyTB": 367
   "Draw by adjudication: maximal game length": 7
   "Draw by fifty moves rule": 72
   "Draw by stalemate": 8
   "Loss: Black mates": 47
   "Loss: Black wins by adjudication: SyzygyTB": 71
   "Loss: White mates": 75
   "Loss: White wins by adjudication: SyzygyTB": 97
   "Win: Black mates": 121
   "Win: Black wins by adjudication: SyzygyTB": 123
   "Win: White mates": 166
   "Win: White wins by adjudication: SyzygyTB": 198

[Jim Ablett]

Drosophila 1.6 by Gustaf Ullberg
https://github.com/gustafullberg/drosophila

Logo by by Dusan Stamenkovic

Drosophila 1.6 64 ja

Windows/Linux & Android profiled builds.

https://drive.proton.me/urls/FX7R80QJSG#7t7EHg508rdA

Android builds work in DroidFish but not in Chess For Android.

Jim.

[Elorejano]

Hi Jim!
Could you compile (and improve) GNUchess v6??
And dont forget to add nnue to Protector
Thanks and sorry for my insistences

[Guenther]

Drosophila 1.6 by Gustaf Ullberg
https://github.com/gustafullberg/drosophila

Logo by by Dusan Stamenkovic

Drosophila 1.6 64 ja

Windows/Linux & Android profiled builds.

https://drive.proton.me/urls/FX7R80QJSG#7t7EHg508rdA

Android builds work in DroidFish but not in Chess For Android.

Jim.

Hi Jim,

I think it should not be called plain 1.6, because it is 27 commits (and 3 years later - last commit 2023) above the official release 1.6
and also has uci support now. Also I think this should be mentioned in a readme too.

You might consider either adding the commit number/sha/date of last commit, or increase the version number.

Guenther

[Jim Ablett]

Drosophila 1.6 by Gustaf Ullberg
https://github.com/gustafullberg/drosophila

Logo by by Dusan Stamenkovic

Drosophila 1.6 64 ja

Windows/Linux & Android profiled builds.

https://drive.proton.me/urls/FX7R80QJSG#7t7EHg508rdA

Android builds work in DroidFish but not in Chess For Android.

Jim.

Hi Jim,

I think it should not be called plain 1.6, because it is 27 commits (and 3 years later - last commit 2023) above the official release 1.6
and also has uci support now. Also I think this should be mentioned in a readme too.

You might consider either adding the commit number/sha/date of last commit, or increase the version number.

Guenther

I've updated it now with latest commit number and updated the readme with all changes since last official release.

Jim.