IvanHoe 999949 Beta

[Tom Barrister]

Just to add, I get a set of WinRAR error messages for each zip file stating 'unknown method'?

I had the same problem with an earlier version of 7Zip. My understanding is that you need to download the current version of 7Zip, which is free. Downloading the current version of WinRar may (but probably won't) also work.

[Razor]

I will try downloading latest version of WinRAR; I already have latest version of 7zip

[Razor]

That fixed it!

[Razor]

Question: when comparing H1.03a to I999949 what conclusions can be drawn from the depths reached, i.e., if H1.03a states a depth count of 32/70 is this the same as a 32.44 depth count declared by I999949? If the first two digits of the depth count are different for both engines, then which of the examples given can be seen as deeper?

BTW: on reasonably identical machines, using the same GUI application, Hash Memory, same position, etc., H1.03a reached 32/70 in 196mins whereas I99949 reached 32.44 in 598mins

[kingliveson]

IvanHoe 999949j Linux/Windows 32 and 64 bit builds are now available at http://chess.cygnitec.com/engine/ivanhoe/

The archive has an Intel build for both Linux and Windows, GCC, and Visual Studio.

Rather than overwriting the archive on the server, since I dont like to have more than a single build of the same release there, so I'll just attach a new build here. Hash Clear function in Arena should now work and this build should be faster. Windows Only.

Note: if you have a hard time extracting the file, get the latest 7-zip.

edit: file removed and re-attached.

[ernest]

...Attachments

Your attachment is only the Microsoft compile.
Why this choice, since the Cygnitec has both the Intel and Microsoft compiles?

Also, are those KLO compiles?

[kingliveson]

...Attachments

Your attachment is only the Microsoft compile.
Why this choice, since the Cygnitec has both the Intel and Microsoft compiles?

Also, are those KLO compiles?

Windows compile, meaning no new Linux build,though the Linux build is included in the archive.

[ernest]

...Attachments

Your attachment is only the Microsoft compile.
Why this choice, since the Cygnitec has both the Intel and Microsoft compiles?

Also, are those KLO compiles?

Windows compile, meaning no new Linux build,though the Linux build is included in the archive.

Of course Windows compile, but you did not answer my questions...

[kingliveson]

...Attachments

Your attachment is only the Microsoft compile.
Why this choice, since the Cygnitec has both the Intel and Microsoft compiles?

Also, are those KLO compiles?

Windows compile, meaning no new Linux build,though the Linux build is included in the archive.

Of course Windows compile, but you did not answer my questions...

All the builds on the server are roughly 70/30 percent Microsoft/Intel respectively. This new build is probably direction rest of the builds will go; Intel for x64 and Microsoft for x86. Am sure there will be another Thanksgiving day where all builds from both compilers will be available.

[kingliveson]

Your attachment is only the Microsoft compile.
Why this choice, since the Cygnitec has both the Intel and Microsoft compiles?

The attachment is a combination of both Intel and Microsoft (see above post). The reason is because after doing some tests, This new build is faster than all the other builds (~5%), and performs better. And also non-popcnt build closed the speed gap when compared to Houdini 1-cpu -- a difference of ~2-3% from ~7-10%. Normally I would just overwrite what's on the server, especially after a few corrections. It was tested against compiles by others (1 minute 150 games only) and the result was K>T>B>K. Recommendation would be to use the newer build, and hope your question was answered this time.