gate/xilinx slice

when

Here we go again...

There have been many discussions on design gate equivalents and how the "gates" in a part relate between manufacturers. You can't really relate the slice count in a design to gates, because the gate count depends on what resources you use from each slice. For example some slices may implement a 4-input lookup that would require several 2-input gates to implement plus a flip-flop which could be built of several more gates. Other slices could implement dual-port RAM that would count as a whole bunch of gates, or a route-through that counts as no gates.

At one time the tools reported the equivalent gate count of a design and I have no idea what algorithm they used to get it. I often had designs in an XCV50 that reported 150,000 gate equivalents or three times the supposed gates available in the part.

I think most of us have given up on counting gates and let the marketing guys argue over who has more. If you're not planning to move your FPGA design to an ASIC, you probably don't need to know the gate count and if you are moving to an ASIC, I'm not sure the tools will give you an accurate estimate anyway.

Good luck, Gabor

"> I think most of us have given up on counting gates and let the

yes marketing guy what we couldn't do without them ^^

If you're not planning to move your FPGA

No it's not to move from a fpga to an asic , it's just for my resume i wanted to indicate the size in gate of an IP I developped during my internship

Alexis

and

an

I see, so it's marketing after all :) By all means use the largest number you can find!

At some point gate count becomes irrelevant. Some of the things that really matter in terms of complexity are:

- the device family, size and percentage utilization

- number of clock domains and their frequencies

- the function and complexity of the core processing logic

- what kind of interfaces exist and the rates they run at e.g. DDR, bit skew compensation schemes etc.

- programmability/reconfiguration features of the design e.g. microcoded control, CSRs etc.

Hardware design is moving in the same direction as software design in terms of functional complexity. Nobody in the software world will these days use number of lines of asm code as a metric (unless they're low level firmware guys maybe).

my 2 cents

-Paul