Winboard Forum

Here's a direct link to the new fixed build.
Intel compiler 9.1 p.g.o

http://www.sigmirror.com/files/599/PROPHET/Prophet_2.0_epsilon1%28fixed%29_win32.zip


Will be on my homepage tonight (the one available on my homepage at the moment is the un-fixed version. James found a bug in that version)

Jim.

Jim Ablett wrote:

Here's a direct link to the new fixed build.
Intel compiler 9.1 p.g.o

http://www.sigmirror.com/files/599/PROPHET/Prophet_2.0_epsilon1%28fixed%29_win32.zip


Will be on my homepage tonight (the one available on my homepage at the moment is the un-fixed version. James found a bug in that version)

Jim.

Hi Jim,

I tried Prophet for a new rapid test tourney and after a program
which should never lose on time lost 2 times on time only
against Prophet my investigation came to the conclusion
that Prophet _always_ ponders

Best regards,
Guenther

Correction Jim!
It seems I did not have yet the fixed version(mine was a day
or two earlier). The fixed version seems to work ok now with
ponder off!
Sorry for the false alert!

Best regards,
Guenther

Hi Guenther,

The original release of Prophet 2.0 Epsilon was recalled because
of some ponder commands missing, but obviously a problem still exists. I also made a change with the compiler used, going from
Gcc to Intel and using msvc compiled pthreads library in place of Gcc compiled pthreads.

It could be not releasing threads.
Could you test the GCC build?. I'd like to eliminate myself as the
culprit.


http://www.sigmirror.com/files/599/PROPHET/Prophet_2.0_epsilon1%28gcc%29_win32.zip



Regards,
Jim.

Jim Ablett wrote:Hi Guenther,

The original release of Prophet 2.0 Epsilon was recalled because
of some ponder commands missing, but obviously a problem still exists. I also made a change with the compiler used, going from
Gcc to Intel and using msvc compiled pthreads library in place of Gcc compiled pthreads.

It could be not releasing threads.
Could you test the GCC build?. I'd like to eliminate myself as the
culprit.


http://www.sigmirror.com/files/599/PROPHET/Prophet_2.0_epsilon1%28gcc%29_win32.zip



Regards,
Jim.

Hi Jim,

Both versions the first fixed one from 11/01 and the new one
seem to have no problems with ponder off.

As we are just here, I found some bugs in Echelon though
In my current test tourney it played once an illegal move
with a ghost rook and once it crashed with a stackdump
error. I will open a new thread for Echelon, please stay tuned.

Best regards,
Guenther

Jim Ablett wrote:Hi Guenther,

The original release of Prophet 2.0 Epsilon was recalled because
of some ponder commands missing, but obviously a problem still exists. I also made a change with the compiler used, going from
Gcc to Intel and using msvc compiled pthreads library in place of Gcc compiled pthreads.

It could be not releasing threads.
Could you test the GCC build?. I'd like to eliminate myself as the
culprit.


http://www.sigmirror.com/files/599/PROPHET/Prophet_2.0_epsilon1%28gcc%29_win32.zip



Regards,
Jim.

I should've just labeled the fix epsilon.2 to avoid any confusion. So, are you having trouble with pondering? I literally just switch a global flag "ponder_ok" off or on in response to the winboard easy/hard commands. I didn't test it. :-/ If there is a problem with pondering and it is with the "patched" epsilon.1, the problem is most likely an issue with my implementation of the Winboard protocol.

Why do you suspect Prophet is not releasing threads? The thread implementation is not very elegant - something I intend to clean up before releasing 2.0, but I think it's ok.

Hi James,

There's no problem now. Guenther was using the original build with
the 'hard/easy commands missing.

Nothing wrong with Prophet's thread handling either. I thought the problem may have been with the different 'win32 pthreads' library I had used to build with the Intel compiler. (before I found out Guenther was using the old version of Prophet of course).

Best regards,
Jim.