Texel 1.03

[Dr.Wael Deeb]

- Implemented pawn storm bonus for opposite castling positions

I am extremely eager to test this new parameter

Thanks for the release....
Dr.D

[petero2]

- Implemented pawn storm bonus for opposite castling positions

I am extremely eager to test this new parameter

In the executables I shipped it is not a UCI parameter. If you can compile the source code yourself it only takes a one-line change in src/parameters.hpp to make it a UCI parameter. If not, I can make the compile for you if you let me know which executable version you want.

There are two reasons why I don't expose all evaluation function parameters as UCI parameters in the release executables.

1. There are 362 parameters in the evaluation function and I think most GUIs would not be happy with that. At least xboard 4.5.2a from fedora 17 fails for this case.

2. UCI parameters are slightly slower than compile time constants. I have not measured how big the slowdown is though.

[lucasart]

Impressive work Peter!

[tpetzke]

Hi Peter,

if I understood it right you have the same piece material values for midgame and endgame.

Have you tried different mid- and endgame values also but found no benefit?

Thomas...

[Marek Soszynski]

There are 7 compiled versions included:

texel-arm : For the armv7-a architecture. Should work on most modern android devices.
texel32.exe : For 32-bit windows systems with SSE42 and POPCOUNT.
texel32old.exe : For 32-bit windows systems without SSE42 and POPCOUNT.
texel64 : For 64-bit linux intel systems with SSE42 and POPCOUNT.
texel64.exe : For 64-bit windows intel systems with SSE42 and POPCOUNT.
texel64amd.exe : For 64-bit windows systems with SSE42 and POPCOUNT.
texel64old.exe : For 64-bit windows systems without SSE42 and POPCOUNT.

Any possibility of a texel64old for Linux systems without SSE42 and POPCOUNT?

[Marek Soszynski]

There are 7 compiled versions included:

texel-arm : For the armv7-a architecture. Should work on most modern android devices.
texel32.exe : For 32-bit windows systems with SSE42 and POPCOUNT.
texel32old.exe : For 32-bit windows systems without SSE42 and POPCOUNT.
texel64 : For 64-bit linux intel systems with SSE42 and POPCOUNT.
texel64.exe : For 64-bit windows intel systems with SSE42 and POPCOUNT.
texel64amd.exe : For 64-bit windows systems with SSE42 and POPCOUNT.
texel64old.exe : For 64-bit windows systems without SSE42 and POPCOUNT.

Any possibility of a texel64old for Linux systems without SSE42 and POPCOUNT?

It's okay. I've hacked the MAKE file and compiled one that works.

[petero2]

There are 7 compiled versions included:

texel-arm : For the armv7-a architecture. Should work on most modern android devices.
texel32.exe : For 32-bit windows systems with SSE42 and POPCOUNT.
texel32old.exe : For 32-bit windows systems without SSE42 and POPCOUNT.
texel64 : For 64-bit linux intel systems with SSE42 and POPCOUNT.
texel64.exe : For 64-bit windows intel systems with SSE42 and POPCOUNT.
texel64amd.exe : For 64-bit windows systems with SSE42 and POPCOUNT.
texel64old.exe : For 64-bit windows systems without SSE42 and POPCOUNT.

Any possibility of a texel64old for Linux systems without SSE42 and POPCOUNT?

It's okay. I've hacked the MAKE file and compiled one that works.

Good. My thought was that most linux users could compile their own version, so I only included one linux executable.

If you have a 64-bit system, just running "make" without parameters should build an executable called "texel" that does not require any special CPU features.

I intended this to work also for 32-bit systems, but due to a Makefile bug it does not. You have to remove the "-m64" from the rule for the $(OBJS_DEF) and "texel" targets to make it work.

[petero2]

if I understood it right you have the same piece material values for midgame and endgame.

Have you tried different mid- and endgame values also but found no benefit?

The base piece values are the same in MG and EG, but most piece types have different piece square tables in MG and EG, so this information is already encoded in the piece square tables.

The exception is the queen which has the same PST in MG and EG. I plan to test adding an EG PST for the queen and to make the MG PST contain more parameters. Currently it only contains 10 because it has horizontal, vertical and diagonal symmetry.

[Dr.Wael Deeb]

- Implemented pawn storm bonus for opposite castling positions

I am extremely eager to test this new parameter

In the executables I shipped it is not a UCI parameter. If you can compile the source code yourself it only takes a one-line change in src/parameters.hpp to make it a UCI parameter. If not, I can make the compile for you if you let me know which executable version you want.

There are two reasons why I don't expose all evaluation function parameters as UCI parameters in the release executables.

1. There are 362 parameters in the evaluation function and I think most GUIs would not be happy with that. At least xboard 4.5.2a from fedora 17 fails for this case.

2. UCI parameters are slightly slower than compile time constants. I have not measured how big the slowdown is though.

Thanks Peter....

I'll let you know when I fire up the engine regards,
Dr.D

[Jamal Bubker]

Thanks Peter