FPGA vs. GPP anyone?

He probably means "general purpose processors". I'm sure with that detail you can now provide crystal clear guidance on "FPGA vs. GPP". :-)

--
Ben Jackson AD7GD

http://www.ben.com/

For what problem?

--
Phil Hays

sorry 'bout that. Yup, I was directing to the "general purpose procesor"

The answer is, it depends. You can find a detailed view of the tradeoffs here:

The basic point is, that each architecture is most efficient on a problem that matches it's instruction and data granularity. For a problem with 32-bit data that maps well to the instructions available in a given ISA and that does not have a lot more inherent parallelism than what the available ISA implementation offers a general purpose CPU is very good.

An ASIC implementation can exactly match the data types, instructions and spation parallelism that suits a problem, so it has a big advantage there, but it is inefficient in the sense that it must make a lot of worst case assumptions. (e.g. it implements always the same number of processing elements no matter what size the problem instance is.

The FPGA inherits essentially all advantages of the ASIC implementation at a price of a factor of 10x to 20x worse area/delay product. For certain applications the FPGA can more than compensate for that by its ability to tailor the hardware to the problem instance.

Kolja Sulimma

The easy answer is that general purpose processors are better for general purpose problems.

Assuming you are asking about FPGA for use as a processor, (see

) then they work well for specific problems. Specifically, if you want to do some simple operation a very large number of times an FPGA or array of FPGAs may be a good solution.

-- glen