new chess computer: CT800
Posted: Mon Sep 05, 2016 11:46 pm
So here is a new dedicated chess computer, the CT800:
http://www.ct800.net
The raw playing strength of about ELO 2100 may seem unimpressive to quite some of you who are into engines in the 2800+ range. However, the power consumption of the total system under computing load is just 0.4W. That would barely suffice to power the mouse of a PC running Stockfish.
Some of you may remember the engine: it is based on NG-Play v9.86, but I made a lot of bugfixes and changes. The opening book alone has grown from a 1,000 plies to 17,000 plies, all of them checked with Shredder and entered manually. Together with a total rewrite from line-based to CRC-32 position based. Pawn eval and rook handling are also better than with the baseline version, playing stronger in quiet positions.
Actually, I am more into embedded systems than into chess programming. When I saw some exciting Cortex-M4 boards, I figured that I had to do something with them, so I came up with a chess computer. The system is not using any operating system; it is running bare-metal. I even ditched the C standard library. The hardware interface is not using libraries, either. I found it more interesting to do that low-level part as per the reference manual of the CPU.
The challenge was to get the software running on the ARM platform with just 192 kB RAM, including the interrupt stuff, various configuration options, time controls and the embedded user interface. The whole thing originally had 7,000 lines of source text, which has tripled by now. That is not counting the opening book and its compiler. Plus of course building up the hardware itself with lots of soldering, drilling, filing and so on. It is not just a concept, it is a real and working prototype.
However, the software itself can also be tested on any platform that has GCC available, like GNU/Linux or Windows with Cygwin. Documentation can always be better, but I put quite some effort into proper docs. Not only for the software, but also for the hardware.
George Georgopoulos, author of NG-Play, didn't put a dedicated licence to the software. He just stated that people are free to do anything as long as it stays open source. I felt that this is in the spirit of the GPL, so I put the project under GPLv3+. Another contribution is the KPK endgame table module by Marcel van Kervinck. Thanks to you both.
http://www.ct800.net
The raw playing strength of about ELO 2100 may seem unimpressive to quite some of you who are into engines in the 2800+ range. However, the power consumption of the total system under computing load is just 0.4W. That would barely suffice to power the mouse of a PC running Stockfish.
Some of you may remember the engine: it is based on NG-Play v9.86, but I made a lot of bugfixes and changes. The opening book alone has grown from a 1,000 plies to 17,000 plies, all of them checked with Shredder and entered manually. Together with a total rewrite from line-based to CRC-32 position based. Pawn eval and rook handling are also better than with the baseline version, playing stronger in quiet positions.
Actually, I am more into embedded systems than into chess programming. When I saw some exciting Cortex-M4 boards, I figured that I had to do something with them, so I came up with a chess computer. The system is not using any operating system; it is running bare-metal. I even ditched the C standard library. The hardware interface is not using libraries, either. I found it more interesting to do that low-level part as per the reference manual of the CPU.
The challenge was to get the software running on the ARM platform with just 192 kB RAM, including the interrupt stuff, various configuration options, time controls and the embedded user interface. The whole thing originally had 7,000 lines of source text, which has tripled by now. That is not counting the opening book and its compiler. Plus of course building up the hardware itself with lots of soldering, drilling, filing and so on. It is not just a concept, it is a real and working prototype.
However, the software itself can also be tested on any platform that has GCC available, like GNU/Linux or Windows with Cygwin. Documentation can always be better, but I put quite some effort into proper docs. Not only for the software, but also for the hardware.
George Georgopoulos, author of NG-Play, didn't put a dedicated licence to the software. He just stated that people are free to do anything as long as it stays open source. I felt that this is in the spirit of the GPL, so I put the project under GPLv3+. Another contribution is the KPK endgame table module by Marcel van Kervinck. Thanks to you both.