17 Mar
2005
17 Mar
'05
4:14 p.m.
Tony wrote:
And how long did it take you to design all those parts? (If you're just
sticking pre-designed modules together, that's little more than PC
assembling, and doesn't count :-)). I find I can design and debug _as I
wire_.
So not much better then. The FPGAs have
presumably become more complex as
the PCs to compile them get faster...
[...]
Oh yes. But the point it is took many, many,
compiles to get the darn
thing to work. I am sure I could have hand-wired it in less time than
that.
For a Spartan 3 FPGA (XC3S400), in less than five minutes I can
compile a design that includes a 32-bit RISC processor core, a VGA
controller, a PS/2 mouse controller, and an SDRAM interface.
If I wanted to build the equivalent out of SSI/MSI chips, it would
take me weeks if not months.
I suspect that the actual time in going from idea to working prototype is
much the same in both cases. Or at least it was when I had to use FPGAs.
If you need to make a change to a hand-wired
design, it takes,
perhaps, 5 minutes. To do it to the FPGA design means another overnight
compile.
Depends on the change. If it's simple, the times might be comparable.
For a complex change, doing it on the FPGA is much faster. For instance,
suppose I decide that I want to change the load and store byte
instructions to be little-endian rather than big-endian. With an HDL IF I want to do things like experiment with adding
extra pipeline stages,
the FPGA is an even bigger win.
And when it doesn't work, and you can't probe the relevant signal in the
FPGA, the board of TTL/ECL is a much bigger win...
-tony