Dedicated Chess Computer Emulations

Discussion of anything and everything relating to chess playing software and machines.

Moderators: hgm, Harvey Williamson, bob

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
Post Reply
Steve B
Posts: 3697
Joined: Tue Jul 31, 2007 2:26 pm
Contact:

Dedicated Chess Computer Emulations

Post by Steve B » Fri Jun 18, 2010 11:03 am

For the first time ever...
emulations for the Mephisto World Champion programs for the years 1984-1991 and the Lang London Update(1996) are now incorporated into a dedicated chess computer

the computers are the Phoenix Technology Resurrection II and Revelation computers which run on a Xscale PXA255 processor-500 Mhz-32MB Dram(8-16 is used for Hash Tables)


i did some initial testing of the speedups achieved using the emulations vs the actual modules(68000 Motorola Processor)....

the speedup is dependent on the use of hash tables ,and the actual number of possible moves..so speedups will vary depending on the phase of the game(Opening,middle and endgame)
i tried a simple test in the opening phase..
for the 9 Emulations(Glasgow to London)only the 68000 modules were tested and not the 020 or 030 modules


Test Methodology:
time it takes to complete a full x ply search after white plays 1.Nh3(or 1.Na3 depending on which move is not in the engines book)
there are 20 possible black replies
For Glasgow to Roma i used a full 5 Ply Search
For Almeria to London i used a full 6 Ply Search

when switching between emulations in the Revelation ..Ram seems to be effected
as a result .. each test was performed by completely closing down the Rev after each test and loading a new program rather then just switching programs
each test result shown was performed twice for both the module and the emulation as a double check

of course a more accurate test would be all phases of the game for each module ..but as i work for a living this will have to do for now...when i am retired and living on a dessert island i will perform a more complete and accurate test

Rev-Glasgow.....595 Sec.
Glasgow Module.....3330 Sec.
speedup achieved...5.6x

Rev-Amsterdam.......470 Sec.
Amsterdam Module......1695 Sec.
speedup achieved...3.61x

Rev-Dallas.....575 Sec.
Dallas Module.....2040 Sec.
speedup achieved...3.55x

Rev-Roma......595 Sec.
Roma Module.......2075 Sec.
speedup achieved...3.49x

Rev-Almeria.....194 Sec.
Almeria Module.......597 Sec.
speedup achieved...3.1x

Rev-Portorose.....180 Sec.
Portorose Module.......538 Sec.
speedup achieved...3.0x

Rev-Lyon.....172 Sec.
Lyon Module.....510 Sec.
speedup achieved...2.97x

both 1.Nh3 and 1.Na3 are in the Vancouver and London's book so here the move sequence tested was 1.Na3 e5 2.Nb1
there are 30 possible black replies

Rev-Vancouver.....352 Sec.
Vancouver Module.....1395 Sec.
speedup achieved...3.96x

in preparation to test the London modules i was reminded of the fact that i do not own the London 68000 modules while i do own the 020 and 030 London modules
i therefore tested the speedup achieved against the London 68020 modules

Rev-London 68000.....264 Sec.
London 68020 Module.....451 Sec
speedup achieved...1.71x

Richard Lang On Steroids Regards
Steve

Post Reply