Stats

  • Locked Locked
  • Replies 3
  • Subscribers 143
  • Views 13944
  • Members are here 0
This discussion has been locked.
You can no longer post new replies to this discussion. If you have a question you can start a new discussion

possible to do Analog simulation using High speed clusters and Parallel processing?

AJ123654
Hello All,

I mostly use scripts to do Spectre simulation. We here have a high speed cluster and we would like to leverage it to reduce the speed time of simulation. So I have decided to join a course which uses openMPI or C ( not really sure about the language ) to parallelize the code.

Does anybody here have any experience on using cadence spectre simulation on high speed clusters? or if anybody can give me feedback on how useful this course would with respect to cadence spectre simulation.

Cheers
Anil
  • Spectre already has an option called "APS" which gives access to enhanced performance without sacrificing accuracy, as well as being able to take advantage of multi-core machines which can give major performance increases overall. It also has capability (early access in MMSIM10.1 but enhanced in MMSIM11.1) which allow you to distribute parallel performance using multiple separate machines which can help in certain situations (contact Cadence to find out more about that).

    I doubt going on the course you describe would help, because it will be focused on how you write programs to take advantage of multi-core architectures. Not much you can do if you have existing executables - you'd need the source code and also understanding of how to make the algorithms parallelizable (which is by far the hardest bit compared with the mechanics of using any API) - so clearly that won't help. That's the part that Cadence has done for you.

    Andrew

  • As Andrew says, APS is the way to go if you can. I'm afraid I have nothing useful to add other than that APS is absolutely fantastic and we're now doing simulations we never would've dreamt of before. Thanks Cadence!

  •  Unfortunately I am not in a possition to influence decisions on License. But thanks for giving your feedback.

     I was thinking more on the lines of parallelizing data independent loops. Most of the time for Corner simulations, we end up running the loop for different corners which can be parallelized. Each iteration can be thrown on a seperate core and then the data integrated. I realize that Cadence does have support for distributed computing which can be used in such situations. Maybe I got to explore that.

    Any idea if SKILL is compatible with C? as these languages are used to work wth Clusters. I guess this is a very long shot!

     Cheers

    Anil

Community Guidelines

The Cadence Design Communities support Cadence users and technologists interacting to exchange ideas, news, technical information, and best practices to solve problems and get the most from Cadence technology. The community is open to everyone, and to provide the most value, we require participants to follow our Community Guidelines that facilitate a quality exchange of ideas and information. By accessing, contributing, using or downloading any materials from the site, you agree to be bound by the full Community Guidelines.