New engine releases 2020

[Werner]

-Popcount (no bmi2)
-Bmi2/AVX

Also there are those who still like to use non-popcount.

Sorry if I'm not understanding. Perhaps my question wasn't clear. Your system has BMI2 and AVX (not AVX2)? If you provide me with the name of your CPU + your OS (Linux or Windows?), I should hopefully be able to provide you with a compile.

same here: Windows 10 64bit; intel Core i7-7700 Kaby Lake (Bmi2,AVX2) and Intel Core I7 Ivy Bridge (popcount, no Bmi2)

[Gabor Szots]

I could generate an exe via the CMake GUI. The makefile was a huge 37k and it also created about 20 other exes which I don't need. Compiling took several minutes. But the engie seems to work.

[connor_mcmonigle]

I could generate an exe via the CMake GUI. The makefile was a huge 37k and it also created about 20 other exes which I don't need. Compiling took several minutes. But the engie seems to work.

Yay! It would seem you built the entire project instead of just the engine. From the command line, entering "make seer" instead of "make" in the build directory will just build the engine. I'm not familiar with cmake-gui so sorry about all the extra executables. Are you able to load the weights provided with the release? Additionally, what kind of speed are you observing?

Thanks,
Connor

[Rebel]

Weiss 1.2 released with 120 self-play elo, I expect 80 or so in lists

https://github.com/TerjeKir/weiss/releases/tag/v1.2

Static exchange evaluation has sadly made Weiss a bit slower but on the plus side compiling is now easier - just 'make' will likely be optimal for any machine. Also no longer requires 'isready' (or anything else) before 'go'

SRL 1000ms +89

Code: Select all

                           The SRL rating list of : Sun Oct 18 10:23:08 2020
    EPD  : epd\lc1.epd
    Time : 1000ms
                                                                     Max            Time   Hash          
    Engine                 Points  Used Time   Found   Pos    Elo   Score   Score    ms     Mb  Cpu  Errors
35  Weiss 1.2              293517  11:27:33.1  20047  40000  2935  400000  73.38%   1000   128    1     0
54  Weiss 1.0              284693  11:23:14.3  19241  40000  2846  400000  71.17%   1000   128    1     0

Cool....

Thanks for testing, a bit smaller of a gap to 1.0 than I'd have expected - any chance you could add in 1.1?

Code: Select all

                           The SRL rating list of : Sun Oct 18 13:11:21 2020
    EPD  : epd\lc1.epd
    Time : 1000ms
                                                                     Max            Time   Hash          
    Engine                 Points  Used Time   Found   Pos    Elo   Score   Score    ms     Mb  Cpu  Errors
35  Weiss 1.2              293517  11:27:33.1  20047  40000  2935  400000  73.38%   1000   128    1     0
52  Weiss 1.1              286949  11:27:27.3  19401  40000  2869  400000  71.74%   1000   128    1     0
55  Weiss 1.0              284693  11:23:14.3  19241  40000  2846  400000  71.17%   1000   128    1     0

[Terje]

Weiss 1.2 released with 120 self-play elo, I expect 80 or so in lists

https://github.com/TerjeKir/weiss/releases/tag/v1.2

Static exchange evaluation has sadly made Weiss a bit slower but on the plus side compiling is now easier - just 'make' will likely be optimal for any machine. Also no longer requires 'isready' (or anything else) before 'go'

SRL 1000ms +89

Code: Select all

                           The SRL rating list of : Sun Oct 18 10:23:08 2020
    EPD  : epd\lc1.epd
    Time : 1000ms
                                                                     Max            Time   Hash          
    Engine                 Points  Used Time   Found   Pos    Elo   Score   Score    ms     Mb  Cpu  Errors
35  Weiss 1.2              293517  11:27:33.1  20047  40000  2935  400000  73.38%   1000   128    1     0
54  Weiss 1.0              284693  11:23:14.3  19241  40000  2846  400000  71.17%   1000   128    1     0

Cool....

Thanks for testing, a bit smaller of a gap to 1.0 than I'd have expected - any chance you could add in 1.1?

Code: Select all

                           The SRL rating list of : Sun Oct 18 13:11:21 2020
    EPD  : epd\lc1.epd
    Time : 1000ms
                                                                     Max            Time   Hash          
    Engine                 Points  Used Time   Found   Pos    Elo   Score   Score    ms     Mb  Cpu  Errors
35  Weiss 1.2              293517  11:27:33.1  20047  40000  2935  400000  73.38%   1000   128    1     0
52  Weiss 1.1              286949  11:27:27.3  19401  40000  2869  400000  71.74%   1000   128    1     0
55  Weiss 1.0              284693  11:23:14.3  19241  40000  2846  400000  71.17%   1000   128    1     0

I guess elo is just a bit more compressed in SRL, relative to eachother the gains seem reasonable. Thanks again!

[CMCanavessi]

-Popcount (no bmi2)
-Bmi2/AVX

Also there are those who still like to use non-popcount.

Sorry if I'm not understanding. Perhaps my question wasn't clear. Your system has BMI2 and AVX (not AVX2)? If you provide me with the name of your CPU + your OS (Linux or Windows?), I should hopefully be able to provide you with a compile.

Ryzen 7 1700

[Gabor Szots]

I could generate an exe via the CMake GUI. The makefile was a huge 37k and it also created about 20 other exes which I don't need. Compiling took several minutes. But the engie seems to work.

Yay! It would seem you built the entire project instead of just the engine. From the command line, entering "make seer" instead of "make" in the build directory will just build the engine. I'm not familiar with cmake-gui so sorry about all the extra executables. Are you able to load the weights provided with the release? Additionally, what kind of speed are you observing?

Thanks,
Connor

I don't have cmake, I have the CMake GUI. It seems that's different from what you use. Typing cmake has no effect here.

I think I could load the bin file. Speed is about 241 knps in the start position. However, thinking is not displayed while analysing. Stopping analysis produces this display:

rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq -

Engine: Seer 1.0 64-bit [new] (256 MB)
by C. McMonigle

5/05 0:00 -256.00 1.h3 h5 2.g3 h4 3.f3 (3.198) 213

6/06 0:00 -256.00 1.h3 h5 2.g3 h4 3.f3 hxg3 (13.423) 244

7/07 0:00 -256.00 1.h3 h5 2.g3 h4 3.f3 hxg3 4.e3 (17.815) 237

8/08 0:00 -256.00 1.h3 h5 2.g3 h4 3.f3 hxg3 4.e3 g2 (56.661) 241

best move: h2-h3 time: 0:19.953 min n/s: 241.110 nodes: 56.661

Weird.

Also, I started a game: 1. e4 g6 (from external book) 2. e5 g5 3. e6 g4 were the first moves. Under Shredder 13.

So, I must say, something is very wrong.
I suggest you provide a generic 64-bit compile so that everyone can try it. Maybe there is some serious mistake on my part.
A 32-bit compile would also be welcomed. Believe it or not.

BTW, maybe it would be a good idea if the engine displayed its version number.

[CMCanavessi]

BTW, maybe it would be a good idea if the engine displayed its version number.

Oh yes PLEASE!

[connor_mcmonigle]

I could generate an exe via the CMake GUI. The makefile was a huge 37k and it also created about 20 other exes which I don't need. Compiling took several minutes. But the engie seems to work.

Yay! It would seem you built the entire project instead of just the engine. From the command line, entering "make seer" instead of "make" in the build directory will just build the engine. I'm not familiar with cmake-gui so sorry about all the extra executables. Are you able to load the weights provided with the release? Additionally, what kind of speed are you observing?

Thanks,
Connor

I don't have cmake, I have the CMake GUI. It seems that's different from what you use. Typing cmake has no effect here.

I think I could load the bin file. Speed is about 241 knps in the start position. However, thinking is not displayed while analysing. Stopping analysis produces this display:

rnbqkbnr/pppppppp/8/8/8/8/PPPPPPPP/RNBQKBNR w KQkq -

Engine: Seer 1.0 64-bit [new] (256 MB)
by C. McMonigle

5/05 0:00 -256.00 1.h3 h5 2.g3 h4 3.f3 (3.198) 213

6/06 0:00 -256.00 1.h3 h5 2.g3 h4 3.f3 hxg3 (13.423) 244

7/07 0:00 -256.00 1.h3 h5 2.g3 h4 3.f3 hxg3 4.e3 (17.815) 237

8/08 0:00 -256.00 1.h3 h5 2.g3 h4 3.f3 hxg3 4.e3 g2 (56.661) 241

best move: h2-h3 time: 0:19.953 min n/s: 241.110 nodes: 56.661

Weird.

Also, I started a game: 1. e4 g6 (from external book) 2. e5 g5 3. e6 g4 were the first moves. Under Shredder 13.

So, I must say, something is very wrong.
I suggest you provide a generic 64-bit compile so that everyone can try it. Maybe there is some serious mistake on my part.
A 32-bit compile would also be welcomed. Believe it or not.

BTW, maybe it would be a good idea if the engine displayed its version number.

Weird that thinking is not displayed in Shredder. I've only tested with Cute Chess as well as Polyglot+Xboard on Linux and not observed such issues. Without loading a network, the network weights are uninitialized so the behavior you observe is expected (There's effectively no evaluation function!). I fear there's no way a generic compile would be competitive, but 241 knps is good, especially w/o evaluation function (lots of king moves in the search tree slows down the search due to the incremental updates).

I'll be sure to add a version number.

[Gabor Szots]

I fear there's no way a generic compile would be competitive, but 241 knps is good, especially w/o evaluation function (lots of king moves in the search tree slows down the search due to the incremental updates).

I'll be sure to add a version number.

Many people here love chess engines for the fun, not for their strength. And a surely working compile which you tested could reveal the problems on other PC's.

Is there a way to know if the weights are loaded and used? BTW, why is that file so incredibly large?