[cctalk] Re: FORTH fans rejoice

The NC4000 was fun to play with, still. I recall comparing my 4MHz processor to Intels 80386 at 33mHz. I wrote some optimized code to sort 1000 integers. The NX4000 ran in about 19 millisconds. I don't recall exactly what the 386 did but it was 2 or 3 times slower. I'd added one specialized hardware/code to my boards. I made it so that if I wrote to 0FFFF and read it, it would swap the high and low order 8 bits. It was relatively efficient because -1 was a short constant instead of a full address. I needed the 8 bits for my drivers for the floppy and hard drive. I'd also add a couple parts so that on boot, I could load the core to shadow RAM and then switch to full RAM. That allowed me to do patching while experimenting with new core code. I'll trying to recall the manufacture of the processor board( it is in storage ). The other fun thing was that it could recompile the ROM image from the hard disk in less than 15 seconds. The 4000 had an interesting tradeoff in power to speed. The ALU would calculate everything on each cycle and then you'd pick the answer through a mux. This allowed things like add or subtract to start working before the opcode was even fetched, for the add or subtract. Long carries still consumed more than two cycles on faster clock speeds. One threw a NOP infront of the + instruction to allow it to complete. It didn't make sense to me until John Ribble explained it to me. Dwight