Javascript question

[Gabor Szots]

You have a javascript engine, 'engine.js' in directory 'JSengine' on a HDD together with node.exe version 'A' in the same directory. You also have node version 'B' installed on an SSD, the path environment variable containing its path.

You install engine.js under Arena in two ways: one as '<path to JSengine>\node.exe engine.js', the other as 'node <path to JSengine>\engine.js. In other words, the first one uses node A started from the engine's directory, the second one uses node B started from node B's directory.

And now the question: Is it possible that one of the engines plays better than the other?

[mar]

I don't know, in theory a newer version of node may have a better JIT compiler so if the engine runs faster with newer node then it might in theory score better.
JS typically runs about 4 times slower than C/C++, but it really depends on the program. In the case of chess engines which typically contain a lot of jumps, the gap between JS and C++ might be actually lower than a factor of 4.

[towforce]

I don't know, in theory a newer version of node may have a better JIT compiler so if the engine runs faster with newer node then it might in theory score better.
JS typically runs about 4 times slower than C/C++, but it really depends on the program. In the case of chess engines which typically contain a lot of jumps, the gap between JS and C++ might be actually lower than a factor of 4.

This brings back memories of trying to work out which version of Java was running code when more than one version of Java was installed on a PC!

My answer to the OP would be that I'd expect the two different ways of starting JS programs to not impact speed, but computers and OSs are complex and you never know what implementation detail you're not aware of might be having an effect.

For a complex program, I would expect JS to be more than 4x slower than a compiled program: it was over 20 years ago, but I remember issues like multi-dimensional arrays running like a three-legged dog under JS.

[mar]

For a complex program, I would expect JS to be more than 4x slower than a compiled program: it was over 20 years ago, but I remember issues like multi-dimensional arrays running like a three-legged dog under JS.

AFAIK 20 years ago JS used to be interpreted, these days it's JIT-compiled into machine code

[Guenther]

You have a javascript engine, 'engine.js' in directory 'JSengine' on a HDD together with node.exe version 'A' in the same directory. You also have node version 'B' installed on an SSD, the path environment variable containing its path.

You install engine.js under Arena in two ways: one as '<path to JSengine>\node.exe engine.js', the other as 'node <path to JSengine>\engine.js. In other words, the first one uses node A started from the engine's directory, the second one uses node B started from node B's directory.

And now the question: Is it possible that one of the engines plays better than the other?

Did you try to just measure the average speed on some positions?
Various node versions really can make a difference (but I doubt by more than 20%).

I was forced to test various node versions for Wukong after my old node did not run with it and my last JS tests were from 4-5 years ago.
For me it was even more complicated, because being on Win7 on old hardware newest nodes didn't work either ;-)

http://talkchess.com/forum3/viewtopic.p ... 18#p878815

[Gabor Szots]

Did you try to just measure the average speed on some positions?
Various node versions really can make a difference (but I doubt by more than 20%).

That was a good idea. I measured almost no difference.

This was with Lozza. In a tournament it started rather 'slow', then I changed the installation method and it improved. Odd was that better results were achieved using the older node. Must have been a statistical fluke, number of games was not particularly high.

[Guenther]

Did you try to just measure the average speed on some positions?
Various node versions really can make a difference (but I doubt by more than 20%).

That was a good idea. I measured almost no difference.

This was with Lozza. In a tournament it started rather 'slow', then I changed the installation method and it improved. Odd was that better results were achieved using the older node. Must have been a statistical fluke, number of games was not particularly high.

If you want I could look into the games.

[op12no2]

I follow the V8 engine twitter account and they are constantly improving it. Very old nodes from ~2015 had some bugs: if I used const instead of var Lozza got slower, ditto if I used switch. I always try and keep up to date with the latest node assuming it's better than previous ones. Performance is a big thing for them so I think it's a safe assumption. As you have discovered modern nodes need Windows 8.1 at least. Also Lozza 2.0 needs a fairly modern node and will stall on very old ones.

PS: To use Lozza 2.0 on old nodes on Windows 7, comment out the line below like this:-

//process.stdin.resume();

Ditto for Tomitank.

[Gabor Szots]

Thanks all for the replies.