http://www.surveyor.com/blackfin/srv-blackfin-122908.zip
    or
http://code.google.com/p/surveyor-srv1-firmware/
 
The latest SRV-1 firmware includes something unexpected - a very compact, portable chess engine called TSCP ( http://www.tckerrigan.com/Chess/TSCP ).  The TSCP program code fits into slightly more than 9kB.  It resides in SDRAM, but could instead be loaded into L1 cache for significantly better performance.  
 
Besides being an interesting exercise in porting new functions to the SRV-1, part of the concept was that someone might eventually want to add vision processing functions to actually recognize chess piece positions on a board and even a manipulator to move the pieces.  If someone went to that much trouble, they might want a stronger chess engine or an adaptive engine that learned, but TSCP is a good starting point.
 
In order to support the book.txt book openings used with TSCP, we added some firmware commands to read/write the SRV-1's flash memory.  The commands 'zWxx' and 'zRxx', where xx is the 64kB sector numbered from 02-63 (00 and 01 are the boot sectors, and are blocked from access).  We set up TSCP to automatically load the book openings from sector 05, so book.txt needs to be uploaded via XMODEM and then written to flash using 'zW05'
 
The other significant change to this version of firmware is a fix for the motor direction function setPWM() which caused problems with RoboRealm.
 
=============
update - the latest version of srv-blackfin-122908.zip runs chess from L1 and fixes serious memory indexing problem.  also, book openings now work properly

Sorry - it was in the zip archives but we forgot to add it to the svn archive.  It's there now -
    http://code.google.com/p/surveyor-srv1-firmware/source/browse/trunk/blackfin/srv /chess.h
 
I need to note that the chess move generator is misbehaving, and I have not isolated the problem.  At first, I thought it might be the rand() function, and changed to a linear feedback shift register with a period of 2^32, but it's still happening. TSCP is mature code, so we must somehow be clobbering memory to cause this problem.