Vlad0 writes ...
Finally seemingly obezglyuchil version Mod5U. Novaya version Mod5Ua
in the attachment below.
What's new?
1. Fixed bug with avtoopredleniem Popcnt.
2. The number of cores and logical processors are now
determined correctly. At least
intelskih percent. Retrieved on i7 quad:
nuclei, 4 logical processors - 8, Popcnt: Yes.
As it should be.
3. For those who like to include and HT in the number of processors
UCI has introduced a new variable AddHT. If you put it
daw, then a default number of processors should be established
not on the number of nuclei, and include yourself and all the logical processors of the HT
(Unless of course he is present on your computer and turned on). Do not know
whether this feature: PC with HT I have not. But the experiment with the definition of the number of cores and Popcnt on i7 I spent in sending engine
Skype calls to his friend who has a new laptop with a i7 Quad.
4. Fixed bug with long auto-sensing robbobaz, in the case
if the engine put in the wrong place (eg in the folder Desktop).
Search shall be conducted on all subfolders
parent folder (which would take about 10 minutes)
but only to a shallow depth.
5. Fixed bug (present in the Decembrist source),
consisting in the fact that the cache of Total Robbie bases can not be changed:
He was always 1M, regardless of what is displayed
at UCI parameter.
6. Yesterday I made the first PGO optimization, but in distribution
including a compilation of the usual optimization. Although PGO and accelerates
engine per cent at 1.5 - 2, but executable is obvious glitch.
I ran tonight blitsevy match of the 300 parties
between PGO ModB10c Ivan and Peter Pan. Somewhere after 60 parties
Ivan lost consciousness completely disable anything not respond,
so ModB10c performance won about 100 games, until I noticed
this outrage and do not stop beating my defenseless Ivan.
From Mod5a and Mod5U (not turbo) this phenomenon (and even single
Exception) did not happen on my competence had never, even though the amount they
played enough of more than 1000 games.
It is interesting that "coaching" so-called instrumental version of Ivan on PGO
test "go movetime 120000" test gave quite impossible result:
the best first move of white (from the initial position): e8-e4??? Ie reason Gluck
is, apparently, somewhere in the PGO Optimizer VS 2010 RC.
Shl also added a link
IvanHoe63Mod5Ua (Vlad0) & Source.7z - 360.7 Kb
ZZY Forgot to add (although I already wrote about it several times, but to many for some reason did not make it):
My compilation series Uxx not need any external libraries: all they need
they are inside the executable.
http://depositfiles.com/files/1euiaj1zhhttp://www2.speedy-share.com/zr8yrhc...Source.7z.htmlhttp://www.zshare.net/download/75406271876c1ff8/
