Added on 31 Aug: The main changes:
1) Adding of new UCI option LogFile;
If set to 1 shows successes and failures while
creating Large page; if set to 2 writes all UCI
commands from GUI to engine. The name
of log file looks like "Ivanhoe XXX-log.txt";
2)Deleting of UCI option "TitanicMode" -- this option automatically
becomes OFF when the number of threads is changed to 1 ("Robbolito mode");
3) The option "QuickSearch" from v.63Mod6U is restored under the name "HighPrioritySearch".
It speeds up the engine additionally
up to 15-20% and may turn out to be useful in engine-engine matvhes on chess sites.
Like "Large pages" it uses additional resources
of OS to speed up search -- this time additional
processor time. But does not steal this extra time from the opponent during opponent's
search -- so is fair in this sense.
4) Auto load/detect triple/total RB checkboxes from v.63Mod6U is restored.
Loading of triple bases is very time consuming: it takes
about 40 secs on my note! So it is desirable to keep together
with checkbox for it (for users uccustomed to load them at any match)
and the button for those who switch them on only occasionally (like me).
The button for unloading of these bases is kept just for the case.
Because I noticed, that sometimes clicking the load button in one session
is somehow memorized by GUI like Fritz and in the next session triple bases are loaded without "load" button is pressed!
Added on 30 Aug: I replaced yesterday's version
with a fixed one -- there was a garbage left
in yesterday's x32 compile. Besides, I've added
one more engine 52gvUS-x64, which differs from
52fUS-64 in one respect: all old fashioned
C allocations of memory are replaced by modern
windows mechanism of VirtualAlloc(). Because
I can not decide which of these versions is quicker (if at all): the results of last 2 series
of 200 games match with Houdini are dubious,
not permitting to decide it uniquelly.
The promised source is uploaded as well --
it compiles now both with
Intel C++ compiler and MS VS 2010.
It is compatible with the original v.999952a
in 2 respects:
1) If one compiles it under Linux it will reproduce
the original v.999952a code;
2) If one compiles it under Windows it will
reproduce the original Windows code of 999952a
if one comments the only line
#define ADDONS
inside the file WinDefs++.h.
For this reason I propose (to whom it may concern) to lay this source in the foundation
of all following official decembrist's
distributions. If they admit this proposal,
this will greatly facilitate merging of all new
windows addons in Windows port of all subsequent decembris's modificatios.
On the other side, some (if not all) of these
addons can be realized in Linux port as well.
~~~~~~~~~~~~~~~~#
Well said Vlad
