Laser 1.3 Release

[jeffreyan11]

Laser 1.3 has now been released and is available at
https://github.com/jeffreyan11/uci-ches ... s/tag/v1.3

Self-play regressions:
+131 in 8000 games at 3+0.02
+128 in 4000 games at 15+0.05

Estimated CCRL elo is 2970

Major changes:
Syzygy TB support
Evaluation tuning using a method based on Texel's Tuning Method
LMR at PV nodes
Change LMR formula, create a search stack to provide more information for pruning decisions (ideas from Stockfish)

Overall, Laser should be less materialistic and search 1-2 depths deeper on average.

[Graham Banks]

Nice improvement!

Thanks Jeffrey.

[MikeB]

Laser 1.3 has now been released and is available at
https://github.com/jeffreyan11/uci-ches ... s/tag/v1.3

Self-play regressions:
+131 in 8000 games at 3+0.02
+128 in 4000 games at 15+0.05

Estimated CCRL elo is 2970

Major changes:
Syzygy TB support
Evaluation tuning using a method based on Texel's Tuning Method
LMR at PV nodes
Change LMR formula, create a search stack to provide more information for pruning decisions (ideas from Stockfish)



Overall, Laser should be less materialistic and search 1-2 depths deeper on average.

Thanks Jeffrey

macOS version with source

https://www.dropbox.com/s/w1juxbhxpzjyr ... e-1.3.zip?dl=1

Code: Select all

bench with 1 core:
       Hash hit rate: 37.89% of 31400 probes
 Hash score cut rate: 28.49% of 11899 hash hits
  Hash move cut rate: 93.19% of 4294 hash moves
First fail high rate: 81.32% of 12088 fail highs
            QS Nodes: 31604 (38.22%)
         QS FFH rate: 93.78% of 14308 qs fail highs
 Eval cache hit rate: 42.83% of 69477 probes
bestmove g3g4
Nodes: 2818297
Time: 2220
Nodes/second: 1269503

with 12 cores:
info depth 10 seldepth 17 score cp 105 time 31 nodes 325780 nps 10509032 tbhits 0 hashfull 34 pv g3g4 f6d7 g4g5 h6g5 f3g4 e6d6 g4d7 d6d7 f4g5 d7e6 h3g4
       Hash hit rate: 55.02% of 144152 probes
 Hash score cut rate: 48.86% of 79326 hash hits
  Hash move cut rate: 92.82% of 18306 hash moves
First fail high rate: 82.28% of 40489 fail highs
            QS Nodes: 107980 (33.14%)
         QS FFH rate: 92.24% of 46543 qs fail highs
 Eval cache hit rate: 65.68% of 251538 probes
bestmove g3g4
Nodes: 15401752
Time: 1434
Nodes/second: 10740412

[jeffreyan11]

Thanks Michael! I have included the compile in the release.

Laser 1.3 now has a 32-bit compile included with the release! There is no special optimization done for 32-bit at the moment but this should improve in a future release.

[Colin-G]

Could not compile a linux executable on my old computer running Mint 17 32bit.
I tried different values in "CFLAGS" including -m64 but no success, just different error messages.
It did compile ok (without any changes to the makefile) as a 64 bit executable on my modern computer running Mint 18 64bit, but surprisingly I could not create a 32bit executable on that by using the -m32 flag.

[thunderstruck]

Hi Colin,

It's explained in the README in the last line of 'Notes' section.
For older or 32-bit systems, set the preprocessor flag USE_INLINE_ASM in common.h to false.

Cheers

[Colin-G]

Hi Colin,

It's explained in the README in the last line of 'Notes' section.
For older or 32-bit systems, set the preprocessor flag USE_INLINE_ASM in common.h to false.

Cheers

Thanks, I usually check the readme files, but missed this one.
It now compiled ok on my old 32bit linux machine, and gave its uci info on the command line, with the uci command.
However it immediately crashed on making its first move when out of book when playing in Xboard with polyglot.
This happened every time.
The end of the polyglot log file gave this message

Code: Select all

1485009288.347 Adapter->Engine: position startpos moves e2e4 e7e5 g1f3 b8c6 h2h4
1485009288.347 Adapter->Engine: go wtime 292640 btime 299950 movestogo 38
1485009288.349 Engine->Adapter: terminate called after throwing an instance of 'std::system_error'
1485009288.349 Engine->Adapter:   what():  Enable multithreading to use std::thread: Operation not permitted
1485009288.349 Engine->Adapter: EOF
1485009288.349 POLYGLOT *** EOF from Engine ***

I fixed the problem by removing just the "-flto" flag from CFLAGS in the Makefile.
I seem to remember having to do this with some other engine compile in the past.

[Adam Hair]

Laser 1.3 has now been released and is available at
https://github.com/jeffreyan11/uci-ches ... s/tag/v1.3

Self-play regressions:
+131 in 8000 games at 3+0.02
+128 in 4000 games at 15+0.05

Estimated CCRL elo is 2970

Major changes:
Syzygy TB support
Evaluation tuning using a method based on Texel's Tuning Method
LMR at PV nodes
Change LMR formula, create a search stack to provide more information for pruning decisions (ideas from Stockfish)

Overall, Laser should be less materialistic and search 1-2 depths deeper on average.

Good estimate! Within the error bars:

Code: Select all

39 Laser 1.3 64-bit                   2957  +19  -19  55.1%  -37.9   986

http://www.computerchess.org.uk/ccrl/404/