Re: A chip too far? Where is your solution Mr Larkin?

On a sunny day (Wed, 20 Aug 2008 15:27:36 +0300) it happened Anssi Saari
<as@xxxxxx> wrote in <vg3ljyrna6f.fsf@xxxxxxxxxxxxxxxxxxxxxxx>:

Jan Panteltje <pNaonStpealmtje@xxxxxxxxx> writes:

FPGA is the perfect solution here.
So we have then mother boards with on the fly programmable logic as
part of the programs that run on it.

Hmm, I wonder if there are any consumer priced motherboards like that?
Some retro computers implemented with FPGAs exist, but they don't
exactly have 3 GHz C2Ds as the main CPU... More interesting FPGA
boards tend to be a little pricey, like the Xilinx ML-505, which is
shaped so that it can fit in an x4 PCIe slot.

Well, 'mass production' will help...
A Stratix 4 with 680k logic elements in quantities that we see for
PC sales would perhaps be possible.
What needs to be invented is a way to interface it to the CPU and
rest of the system, in such a way that each program can reserve and
use some of those gates at the same time.
Lots of IO I'd imagine.
Yes PCIe boards exist, I like this picture too:
http://www.heise.de/bilder/113681/1/1
It does not have to be just _one_ FPGA (picture is from this link in German:
http://www.heise.de/security/Von-Woerterbuechern-und-Regenboegen--/artikel/113681
)

There is a whole lot of issues to be solved, clock domains, what not,
but at least the FPGA approach should increase speed.
It is a proven concept... unlike using n cores on sequential code...

.