Windows & GCC/INTEL/MSVC & 64-bit/SSE4.2/32-bit
Computer Chess Forum
on Sat Jan 05, 2013 9:10 pm by
on Sun Jan 06, 2013 3:44 pm by
Games Completed = 1000 of 1000 (Avg game length = 14.747 sec)
Settings = RR/32MB/2000ms+100ms/M 700000cp for 7 moves, D 150 moves/EPD:E:\LittleBlitzer\openings.epd(4000)
Time = 4411 sec elapsed, 0 sec remaining
1. Gull II x64 JA Intel 504.5/1000 245-236-519 (L: m=236 t=0 i=0 a=0) (D: r=385 i=80 f=35 s=3 a=16) (tpm=111.4 d=12.69 nps=1687675)
2. Gull II x64 Original 495.5/1000 236-245-519 (L: m=245 t=0 i=0 a=0) (D: r=385 i=80 f=35 s=3 a=16) (tpm=111.6 d=12.63 nps=1616145)
on Sat Mar 09, 2013 12:49 pm by
on Sat Mar 09, 2013 1:29 pm by
on Sun Mar 10, 2013 9:23 am by
@jnhes wrote:Gull R375
on Sun Mar 10, 2013 9:32 am by
on Sun Mar 10, 2013 6:51 pm by
on Sun Mar 10, 2013 7:48 pm by
on Mon Mar 11, 2013 11:54 am by
on Mon Mar 11, 2013 1:50 pm by
is this an official release?
on Mon Jun 17, 2013 5:10 pm by
on Mon Jun 17, 2013 5:24 pm by
Gull 2.1 @ 17.06.2013
Evaluation weights are optimized with the use of automated tuning (source code included).
Gull's evaluation is no longer almost identical to that of Ivanhoe.
Minor search & time management & SMP efficiency enhancements.
On SMP & Hyper-Threading:
By default Gull limits the number of processes to the number of phisical CPU cores availible.
This is done in order to avoid ELO drop due to significant SMP search overhead at bullet time controls.
However for long games & analysis it is recommended to turn HT on and manually set the number of processes (UCI option "Threads")
to equal the number of logical CPUs.
On the automated tuning (technical):
Tuning is done in two steps:
1. an estimate of the gradient is calculated;
2. line search is performed.
SPSA (just move a bit along the gradient direction instead of performing a line search) may also be quite effective,
though it's difficult to pick good parameters.
Fairly simple exit criteria are used for the line search. I don't employ SPRT because it's guaranteed to be optimal only for
elo=elo1 vs elo=elo0 hypothesis, whereas in this case we are interested in elo>elo0 vs elo<=elo0 test. However it's still possible
for SPRT (or some other test) to be better for some elo-elo0 values. Shouldn't be very difficult to run a numerical comparision
between various tests.
on Mon Jun 17, 2013 5:53 pm by
on Mon Jun 17, 2013 11:01 pm by
on Tue Jun 18, 2013 2:15 pm by
@Fauzi wrote:When I installed the 32 bit version of Gull 2.1 in my Arena Gui, got this error message:
" The program can't start because MSVCP110.dll is missing from your computer. Try reinstalling the program to fix this problem
The Visual C++ Redistributable Packages install runtime components that are required to run C++ applications built with Visual Studio 2012.
on Tue Jun 18, 2013 2:28 pm by
on Tue Jun 18, 2013 2:35 pm by
@ChessDrone wrote:New compilation that fix the DLL missing I think: Gull 2.1.zip modified < 5 hours ago 308.9 kB
[mention][/mention] wrote:I recompiled the 32-bit version so it doesn't require any additional DLLs now
on Sat Jun 22, 2013 9:48 am by
on Wed Jun 26, 2013 2:52 am by
http://www.chess2u.com/t7288-cegt-rating-lists-june-23th-2013#43928@CEGT wrote:Gull is running for our blitz list:
Gull 2.1 x64 1CPU = ELO 2962 from 800 games (+5 / +10)
Gull R375 x64 1CPU = ELO 2957 from 1600 games
on Mon Aug 12, 2013 10:12 am by
Program Elo + - Games Score Av.Op. Draws
1 Gull 2.1 x64 Trap : 2957 20 20 500 52.0 % 2943 55.2 %
2 Gull 2.1 x64 : 2943 20 20 500 48.0 % 2957 55.2 %
1 Gull 2.1 x64 Trap : 2957 500 (+122,=276,-102), 52.0 %
2 Gull 2.1 x64 : 2943 500 (+102,=276,-122), 48.0 %
on Mon Aug 12, 2013 10:47 am by
on Mon Aug 12, 2013 11:12 am by
Try this one:@CEGT wrote:thank you,
but AVX is not good for my I7 920
Last edited by majortom on Tue Aug 13, 2013 10:03 am; edited 1 time in total
on Mon Aug 12, 2013 11:33 am by
on Sat Aug 17, 2013 11:18 am by
5-15 ELO gain in 1CPU ponder off mode is expected. More in ponder on and SMP modes.
on Sat Aug 17, 2013 11:52 am by