Sridhar Ayengar wrote:
I'm wondering how a machine like this would compare performance-wise to
a new design using a Symbolics-compatible LISP machine blown into a
top-end FPGA. Would it easily fit? Would it perform well?
I think there's a lot of rom in the ivory, but I think the core logic
would fit. Given that you could do a 400mhz ddr-2 interface in today's
fpga's, and run with an internal clock of, well, maybe 266mhz on a good
day, it should be a lot faster.
I was hoping someday someone would pop up with a cad tape of the ivory.
It would be fun to turn it into verilog. I've never seen one, however.
I've seen schematics for the 3600, and they are complex. The "G"
machines where put into fpga's (probably standard cell) but I've never
seen cad files or schematics.
To be honest, running an emulation of the ivory on a dual core 64 bit
x86 chip at 3ghz is really fast and plenty snappy.
I did compile the original CADR into an fpga. It would synthesize with
the original clock speed. I don't recall how much margin there was; the
ALU paths are pretty slow. I have not actually run it in hardware, but
some day I will, just for fun. It would make a nice desktop curio in clear
plastic.
-brad