To: Pilgrim who wrote (6482 ) 6/6/1999 12:45:00 AM From: bob Respond to of 8581
From USENET. Subject: Re: PSC1000 Java Chip Date: 1999/06/03 Author: Keith Wootten <Keith@wootten.demon.co.uk> Posting History In article <7j425s$mko$1@news4.svr.pol.co.uk>, Clive Arthur <clive@turpe ntine.freeserve.co.uk> writes >Is anyone using this chip? It looks interesting, does it perform as well as >claimed? > >Thanks >-- >Clive Arthur >Development Engineer I've been using it for a while, and I have heard recently that other lesser known companies also use it. Apparently - and I don't know the validity of these rumours - Boeing are using it in some sort of factory automation system, and Sun Micro use it in their JINI development boards. I use a Forth on the PSC1000 and it's very, very fast. I don't know anything about its Java performance; for me life's too short to get into that. For a fast hard real-time system, major advantages of the PSC1000 architecture are its interrupt latency and its deterministic behaviour. Forth Inc have some benchmarks at www.forth.com which are interesting. The chip has a lot of support logic included, and a minimal system can use a PSC1000, a single DRAM chip, an 8 bit EPROM and an oscillator. A full 32 bit SRAM based design gives the best performance - with a 32 bit data bus width, four 8 bit opcodes are loaded in one cycle, helping to ease the memory bottleneck. 8 bit opcodes are all that is required as the source and destination for most operations is implicit - it's the stack. It's supposedly $10 in quantity - Patriot sell them for $25 in small numbers. This is excellent value. They have full technical details at www.ptsc.com. Don't dismiss it because it's different. It will seem at first that a stack architecture *must* be inferior to the register based architecture you know and love, but there can be significant advantages if used properly. Cheers -- Keith Wootten