0
\$\begingroup\$

I have a hardware structure (described in logic diagram consisting of adders and multipliers) that is extremely regular, i.e., one small hardware is repeated again and again in the structure with regular pattern of interconnects. for example, see the schematic below.

Image 1

The hardware on the right is formed using multiple copies of the small hardware element on the left. Now, suppose I have another structure that has the same number of Z-elements (in the above example image) but it doesn't follow any regular pattern of interconnects between different layers, i.e., the interconnections between Layer 1 and 2 are random. My question is which of these hardware will be better (in terms of power efficiency, area or any other respect) or they will both be identical in terms of performance (power, area etc.) ? Please answer both with an FPGA- and ASIC-based implementation platform in mind. Any references are appreciated.

\$\endgroup\$
  • \$\begingroup\$ This is so incredibly vague I don't know where to even start. \$\endgroup\$ – Matt Young Aug 25 '12 at 19:51
  • \$\begingroup\$ Can you post a schematic ??? for booth circuits & describe it more ?? Arrange your question ??? \$\endgroup\$ – xsari3x Aug 25 '12 at 19:55
  • \$\begingroup\$ @xsari3x schematic added. \$\endgroup\$ – ubaabd Aug 25 '12 at 20:53
  • \$\begingroup\$ All I can say is "Huh?". \$\endgroup\$ – Olin Lathrop Aug 25 '12 at 20:59
  • \$\begingroup\$ @OlinLathrop May I know which part of the question I was unable to explain? I don't belong to hardware domain and wanted to know if given some hardware resources, a hardware structure that is regular can be regarded as better than a hardware structure that is irregular even if they use the same hardware resources, i.e., the same number of adders and multipliers. May be I am not using the correct terminology so I would appreciate any help. \$\endgroup\$ – ubaabd Aug 25 '12 at 21:08
3
\$\begingroup\$

You can find the answer yourself by synthesizing the two circuits and comparing the results. Everything depends on the exact circuit, the synthesis tool, P&R tools and the target technology.

It is dangerous to generalize performance (or ask others to predict your performance) based on vague input.

\$\endgroup\$
  • \$\begingroup\$ I thought it would be straight forward that regularity in the structure will 'ALWAYS' be a good thing. I actually belong to algorithm design domain so not good at hardware simulations. I was just wondering if I produce regularity in the structure (by tweaking the algorithm), can I claim without going into further details that it is a better way considering I use the same resources. I guess its not that simple given your input. Thanks for the answer by the way. \$\endgroup\$ – ubaabd Aug 26 '12 at 11:07
  • \$\begingroup\$ Yeah, it is not that straightforward. If you are developing algorithms, the best way is to do a lot of experiments with a hardware synthesizer. There are some "rules of thumb" around, but many of them have been outdated with the increase quality of synthesizers. Use the scientific method: hypothesis -> experiment -> evaluation. \$\endgroup\$ – Philippe Aug 26 '12 at 15:22

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.