Program Suggestions

[Marek Soszynski]

Dear Programmers,

We have plenty of analysis engines already. Here are a couple of things we are genuinely short of:

1) A dedicated mate-solver that is UCI, 64-bit and multi-processor. In other words, a successor to Chest.

2) A cross-platform, graphical chess benchmark. In other words, a successor to FritzMark.

[MattieShoes]

Graphical benchmark? Teats on a bull.

[Dann Corbit]

Dear Programmers,

We have plenty of analysis engines already. Here are a couple of things we are genuinely short of:

1) A dedicated mate-solver that is UCI, 64-bit and multi-processor. In other words, a successor to Chest.

There is nothing else close to Chest in this regard, and I think that both the engine author and the UCI interface author are tired of working on it.
Now, Chest does not use bitboards, so the only real gain of 64 bit will be bigger hash tables. So if you want to see a real huge boost, you will have to talk someone into writing a 64 bit SMP chess mate solver.

Here's the rub:
It's at quarter of a million dollar effort to write it and the only reward is that people will give you nothing for it and in five years they are going to whine because it is not 128 bit and cluster ready.

2) A cross-platform, graphical chess benchmark. In other words, a successor to FritzMark.

STS run under your favorite GUI. Or have crafty throw up a logo after running ''bench''

[Marek Soszynski]

Dear Programmers,

We have plenty of analysis engines already. Here are a couple of things we are genuinely short of:

1) A dedicated mate-solver that is UCI, 64-bit and multi-processor. In other words, a successor to Chest.

There is nothing else close to Chest in this regard, and I think that both the engine author and the UCI interface author are tired of working on it.
Now, Chest does not use bitboards, so the only real gain of 64 bit will be bigger hash tables. So if you want to see a real huge boost, you will have to talk someone into writing a 64 bit SMP chess mate solver.

Here's the rub:
It's at quarter of a million dollar effort to write it and the only reward is that people will give you nothing for it and in five years they are going to whine because it is not 128 bit and cluster ready.
;-)

2) A cross-platform, graphical chess benchmark. In other words, a successor to FritzMark.

STS run under your favorite GUI. Or have crafty throw up a logo after running ''bench''
;-)

Programmers write analysis engines that people give nothing for; but it's mate-solvers we're short of.

As for a graphical benchmark, I see I'm getting the typical technician's response.

Maybe, just maybe, someone here will quietly take up the challenge of either of my suggestions.

[Dann Corbit]

Dear Programmers,

We have plenty of analysis engines already. Here are a couple of things we are genuinely short of:

1) A dedicated mate-solver that is UCI, 64-bit and multi-processor. In other words, a successor to Chest.

There is nothing else close to Chest in this regard, and I think that both the engine author and the UCI interface author are tired of working on it.
Now, Chest does not use bitboards, so the only real gain of 64 bit will be bigger hash tables. So if you want to see a real huge boost, you will have to talk someone into writing a 64 bit SMP chess mate solver.

Here's the rub:
It's at quarter of a million dollar effort to write it and the only reward is that people will give you nothing for it and in five years they are going to whine because it is not 128 bit and cluster ready.

2) A cross-platform, graphical chess benchmark. In other words, a successor to FritzMark.

STS run under your favorite GUI. Or have crafty throw up a logo after running ''bench''

Programmers write analysis engines that people give nothing for; but it's mate-solvers we're short of.

As for a graphical benchmark, I see I'm getting the typical technician's response.

Maybe, just maybe, someone here will quietly take up the challenge of either of my suggestions.

What is it you want to see in a graphical benchmark?
I think that the UCI engine design is clear from the start.
It is not at all clear to me what you would like the graphical benchmark to do.

[Marek Soszynski]

>What is it you want to see in a graphical benchmark?
>It is not at all clear to me what you would like the graphical benchmark to do.

A simple-looking program that you can stick on most PCs (32-bit, 64-bit, multiprocessor or not) running a graphical OS (many if not most versions of Windows, Macintosh, Linux); then you click one button, and you get 1 or 2 numbers related to chess performance on the particular PC. In other words, like I said, a successor to FritzMark. Here are a couple of links...

http://www.chessbase.com/Support/support.asp?pid=150
http://www.jens.tauchclub-krems.at/dive ... marks.html

[Dann Corbit]

>What is it you want to see in a graphical benchmark?
>It is not at all clear to me what you would like the graphical benchmark to do.

A simple-looking program that you can stick on most PCs (32-bit, 64-bit, multiprocessor or not) running a graphical OS (many if not most versions of Windows, Macintosh, Linux); then you click one button, and you get 1 or 2 numbers related to chess performance on the particular PC. In other words, like I said, a successor to FritzMark. Here are a couple of links...

http://www.chessbase.com/Support/support.asp?pid=150
http://www.jens.tauchclub-krems.at/dive ... marks.html

A fritzmark shows how fast Fritz runs under certain conditions.
For environments where Fritz does not run, it lacks meaning.
For instance, it won't tell you how fast Stockfish will run or ChestUCI or some other program. Also, I suspect it does not really benchmark things that are actually important for chess performance like disk seek times and I/O rate for tablebase files, how much RAM is available for hashing, etc.

The closest thing to the Fritzmark is the crafty bench command. It spits out some numbers and runs on everything. However, it will only measure crafty performance and only for a single set of 6 chess positions.

Perhaps someone could modify the crafty output to insert html tags and write it to disk. Then it could launch a browser with the file name as a command line argument.

[Marek Soszynski]

A fritzmark shows how fast Fritz runs under certain conditions.
For environments where Fritz does not run, it lacks meaning.
For instance, it won't tell you how fast Stockfish will run or ChestUCI or some other program.

That's why I asked for a successor to FritzMark. If a new graphical benchmark could be based on a specific Crafty version, for example, that's fine. The point is to have a relatively cross-platform, hardware-unspecific chess benchmark that will give an indication of which system will be better than another for chess - and equally important - that will achieve some popularity so that the results could be standardised and shared.

[David Dahlem]

STS run under your favorite GUI.

Which graphical interface can run STS properly? I know Arena has a problem with the STS epd format -

http://www.talkchess.com/forum/viewtopic.php?t=32022

[swami]

STS run under your favorite GUI.

Which graphical interface can run STS properly? I know Arena has a problem with the STS epd format -

http://www.talkchess.com/forum/viewtopic.php?t=32022

I believe Arena functions well only if the EPD set don't have partial credit moves.

With the analysis log, You can use STS stat 2.5 to calculate the partial credit scores provided you have full EPD's (with partial credit scoring) in the STS Stat subfolder.