Timothy Miller wrote:
> I was hoping to shoe-horn everything into one blockram.  I just don't
> think we need 512 registers, although we could need more than 512
> instructions.

I don't think you need more than 32 registers, which is why a little
distributed RAM works fine for that.  You can dual-port the block
RAM between constants and program.

> What are your pipeline stages?  I'm thinking there's:
> - fetch
> - read
> - compute
> - write

That's right.

> With a unified model, there could be four accesses at once (to a
> unified RAM).  With a separate program file, there are three, as you
> say.  Now, if we were to make one operand always an accumulator, we
> could reduce it to two.

You could, but it doesn't seem like that much help.  The amount of
distributed RAM for the register file is small (assuming that you
don't really need hundreds of registers), and will probably work
out better for an ASIC as well.

Anyhow, to get 3-read, 1-write, you need 3 dual-port memories, or
to run a single one for three clocks.  For 4-read, 1-write, you need
four, or four clocks.  However, even if the nanocontroller runs at
75 MHz, I think it may be possible to run the blockRAM at 225 or
possibly 300 MHz, as long as the clocks are sychronous (i.e, divide
down by a DCM).

I was just really concerned about the 8-read, 2-write statement, but
it sounds like 2-read, 1-write will be OK.  And that there's no reason
why the instructions should require exclusive use of the CPU for five
clocks each (minimum), when pipelining is very easy.

> I've done a VGA text display that refreshed at 18Hz, and it was VERY
> usable.  While I agree that it would be NICE to have better
> performance, I really don't think it's necessary.

Maybe.  I agree with whoever observed that the text mode will be the
first thing many people see, and that some people (myself included)
actually still used text mode a moderate amount.  While that doesn't
justify thowing a huge amount of the FPGA resources at it, if 20% more
logic gets a 5x improvement (without a huge schedule hit), I think
that's worthwhile.

Best regards,
Eric

_______________________________________________
Open-graphics mailing list
[email protected]
http://lists.duskglow.com/mailman/listinfo/open-graphics
List service provided by Duskglow Consulting, LLC (www.duskglow.com)

Reply via email to