Stockfish 4

[Don]

At least in computer chess, the power of money is now defeated. Is that good or bad? I am not a philosopher.

A few years ago the same exact thing happened with the release of Ippolit and all the derivatives that followed - so this is not a significant event.

The majority of open source software is designed to make money, even if not directly. Often it is through support or value added services or simply add placements on websites.

In computer chess it turns out that labor is cheap. If you are doing it at all it's because you love computer chess and you are not allocating a lot of time to it unless you have nothing better to do - or else you are doing this to support yourself and your family.

But more to the point, there are people who feel it's beneath their dignity to pay for ANY software. They don't have any real appreciation for the amount of effort involved. Many of them have tons of commercial software but they didn't pay for it and they are proud of that. Those people are not going to pay for Komodo (or Houdini) anyway. There will be few that will embrace whatever the best for free is, those are not ones I can reach anyway.

So I don't know what impact it will have on me personally if Stockfish really is the new number 1 but I don't think it will be as much as one might think.

A very likely scenario here is that if Stockfish is really the top program now you will start to see closed source derivatives - I think some already exist. But someone will probably take advantage of the labor the Stockfish team put in and go commercial. They will go to some efforts to hide this - probably with some degree of success.

[Dr.Wael Deeb]

Yes ! Stockfish 4 ROCKS ! Using the latest developmental versions I was already beating Houdini 3 users online regularly ; now with Stockfish 4 I'm CRUSHING Houdini 3 users ; ( of course with one changed parameter which I will not reveal for obvious reasons )

Aggressiveness 120

Cowardice 110

You'll not the only one following closely the computer chess scene
Dr.D

[mcostalba]

Right now I can't test the engine properly and hope that this can be fixed.

Hello Ingo,

thanks for testing SF !

As you probably know this time I had to take care of binaries because of Jim choice to rest from computer chess. So I decided for a drastic simplification of the binaries bundled in the package. Mainly, in the case of Windows we have only 2 binaries, one is our first choice and what we recommend to test SF at its best, the other one is a fallback in _all_ the other cases, so 32 bit but also 64 bit without popcount support (sse4.2 and popcount are the same thing).

Due to hardware limitations you decided to test with an intermediate version (not the best one but not as slow as the 32bit fallback). This is your choice and I thank you for taking care to find a standard x64 compile, of course we will lose some ELO points here due to missing hardware POPCOUNT, but this is not so critical IMHO and we can definitely live with that

In the case of Linux I was even more extreme: we didn't bundled _any_ binary. The rationale for this choice is

- Linux binaries are not universal, they depends on the specific distribution libraries versions, so the Linux binary is not guaranteed to run always in 100% of cases.

- Under Linux self compilation is much easier than under Windows and almost always the produced binary works without issues and almost always is even faster

- Many distributions already package Stockfish, so the user has also the possibility to install from rpm in case he prefers to only install from official repositories

- We have provided both standard C++ sources in src directory and C++11 equivalent version in src_c++11 directory. Under Linux, and in particularly with Clang compiler, has been reported the C++11 version to be the faster one.

So under Linux our suggestion to get the best and fastest SF binary is simply: do it yourself with your preferred compiler and on your platform, chances are that the result you'll get is the fastest for your specific machine.

[IWB]

As you know, windows users do not compile by themselves and 90% (at least) of the Stockfish user are windows users.

I have two problems with your decision:

1. Why did you go for a full SSE 4.2 support insted of a POPCOUNT only*? There is no speed difference, you just exclude a lot of AMD users!
2. Why is there no generic x64 windows binary in the official download. You have a working compile. You just have to put it into the download zip file?

* I don' think that it is that important. The 3% difference are how many Elo? 2, 3 ... way below any statistical possibilities I have and ANY human can see! But of course the difference between the x64 SSE42 and the 32 bit is huge (20-25%) The normal 64 bit compile is needed.

Thanks for Version 4, right now, and that's much to early, I guess it will stay below CCT but it is a big improvement anyhow!

Bye
Ingo

[Albert Silver]

As you know, windows users do not compile by themselves and 90% (at least) of the Stockfish user are windows users.

I have two problems with your decision:

1. Why did you go for a full SSE 4.2 support insted of a POPCOUNT only*? There is no speed difference, you just exclude a lot of AMD users!
2. Why is there no generic x64 windows binary in the official download. You have a working compile. You just have to put it into the download zip file?

* I don' think that it is that important. The 3% difference are how many Elo? 2, 3 ... way below any statistical possibilities I have and ANY human can see! But of course the difference between the x64 SSE42 and the 32 bit is huge (20-25%) The normal 64 bit compile is needed.

Thanks for Version 4, right now, and that's much to early, I guess it will stay below CCT but it is a big improvement anyhow!

Bye
Ingo

You test on an AMD, right? You can find what you are looking for here:

http://abrok.eu/stockfish/

[IWB]

You test on an AMD, right? You can find what you are looking for here:

http://abrok.eu/stockfish/

I can't find an x64 POPcount compile there ...!

But thats not the point, I was there and use the x64 compile but why is this not in the official download (at least it wasn't this morning) and why use the full SSE 4.2 set while the POPcount is running everywhere and is as fast. Basicaly you make a penalty for all AMD users down to the old Athlon without any need.

Just assume you make a x64 with POPcount, an x64, and a 32 bit. Same work and same speed as now and a lot more happy users without any additional trouble than now!?

(Idealy you make a x64 compile with autodetection of POPcount).

But don't get that wrong, I am happy to have stockfish for free!!!
(especially as I am offering something for free by myself and nonetheless people complain about it!)

BYe
Ingo

[Eelco de Groot]

You test on an AMD, right? You can find what you are looking for here:

http://abrok.eu/stockfish/

I can't find an x64 POPcount compile there ...!

But thats not the point, I was there and use the x64 compile but why is this not in the official download (at least it wasn't this morning) and why use the full SSE 4.2 set while the POPcount is running everywhere and is as fast. Basicaly you make a penalty for all AMD users down to the old Athlon without any need.

Just assume you make a x64 with POPcount, an x64, and a 32 bit. Same work and same speed as now and a lot more happy users without any additional trouble than now!?

(Idealy you make a x64 compile with autodetection of POPcount).

But don't get that wrong, I am happy to have stockfish for free!!!
(especially as I am offering something for free by myself and nonetheless people complain about it!)

BYe
Ingo

Hello Ingo,

Probably the programmer had to set the limit somewhere, and for the release, Marco wanted to limit the number of different versions in the official release of Stockfish 4. But this site is just an independant initiative from programmer Roman Korba. It is highly appreciated, but the responsibility is Roman's, what versions he chooses. See his first post in the CSS forum. I have not checked there if he posted there recently. I think a popcount version specifically for AMD Phenom maybe could be handled by the makefile, but you would have to ask Marco about it, and maybe you would have to compile it yourself with GCC or MinGW but that really is very simple, well to be fair it is not so simple for the 64 bit optimized version, it is a bit more complicated to set up the compiler for that, at least that is what Gary and Marco said.

From the results so far, I think the compile you have is doing fine! Thanks for running this, it is interesting to see all the different results, against multiple engines!

Eelco

[IWB]

Hello Ingo,

Probably the programmer had to set the limit somewhere, and for the release, Marco wanted to limit the number of different versions in the official release of Stockfish 4. But this site is just an independant initiative from programmer Roman Korba. It is highly appreciated, but the responsibility is Roman's, what versions he chooses. See his first post in the CSS forum. I have not checked there if he posted there recently. I think a popcount version specifically for AMD Phenom maybe could be handled by the makefile, but you would have to ask Marco about it, and maybe you would have to compile it yourself with GCC or MinGW but that really is very simple, well to be fair it is not so simple for the 64 bit optimized version, it is a bit more complicated to set up the compiler for that, at least that is what Gary and Marco said.

From the results so far, I think the compile you have is doing fine! Thanks for running this, it is interesting to see all the different results, against multiple engines!

Eelco

Ahh, I did not know that this wasn't offical. Somehow I had the impression ..., thx! Of course Marco can't put that into it's download.

Still, the other possibility is the same work, with the same result and pleasing more people ...

Yep, the compile I use is just 3% slower on an AMD 8350 and an i5.2500, basicaly there is no difference.

Thx for th einfo
Ingo

[Marek Soszynski]

The majority of open source software is designed to make money, even if not directly. Often it is through support or value added services or simply add placements on websites.

Really? In any case, it's an irrelevant point unless you want to claim that the majority of open source chess software is designed to make money.

[Peter Skinner]

Aggressiveness 120

Cowardice 110

You'll not the only one following closely the computer chess scene
Dr.D

I have tried those settings in the past and never found them to be very good. Either Stockfish wins in brutal fashion or it gets crushed just as brutally.

Is there a place where people are trying different settings and posting results? I would like to add my own..

Peter