Stats

  • Locked Locked
  • Replies 10
  • Subscribers 126
  • Views 19998
  • 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

Multithreading not working

Rob Gregoire

I'm simulating a 15 stage SAR ADC and even though the output log indicates that 4 processors are used with multi-threading I am not getting a reduction in Elapsed Time compared to CPU time. I even tried it with a couple ADCs in parallel to make sure the ADC is big enough but I still get no improvement.

It seems to be related to the process design kit. I can simulate a similar ADC in another process and get speed improvement. I'm using IC 6. Any ideas on what to check or what would be helpful in debugging? My internet search only turned up a question from 2006 that indicated that it doesn't work with some CMOS models.

  • What version of MMSIM are you working with?

    Always safer to go with simulation>options>analog and define threads under n threads in ADE. This number should be equal to no of logical cores.

  • What version of MMSIM are you working with?

    Always safer to go with simulation>options>analog and define threads under n threads in ADE. This number should be equal to no of logical cores.

  • Thanks for your response. To answer you questions:

    MMSIM131. I am already. defining number of threads under simulations->Options->Analog (turning multithreading on, nthreads=4)

    Some more information (red emphasis mine):  Using IC6.1.5

    Right before the transient analysis is listed in the output log there is a notice stating:

    Notice from spectre during initial setup.
        Multithreading Enabled: 4 threads on system with 24 available processors.
            
        58 notices suppressed.


    Time for parsing: CPU = 173.973 ms, elapsed = 266.827 ms.
    Time accumulated: CPU = 2.12368 s, elapsed = 2.34084 s.
    Peak resident memory used = 96.9 Mbytes.

    Entering remote command mode using MPSC service (spectre, ipi, v0.0, spectre0_10789_18, ).

    ============================================================

    And at the end of the simulation:

        tran: time = 410.9 ns    (99.6 %), step = 3.464 ps     (840 u%)
        tran: time = 411.1 ns    (99.7 %), step = 1.216 ps     (295 u%)
        tran: time = 411.4 ns    (99.7 %), step = 979.1 fs     (237 u%)
    Number of accepted tran steps =             124642

    Notice from spectre during transient analysis `tran'.
        Trapezoidal ringing is detected during tran analysis.
            Please use method=trap for better results and performance.

    Initial condition solution time: CPU = 4.46432 s, elapsed = 4.46464 s.
    Intrinsic tran analysis time:    CPU = 11.3365 ks, elapsed = 11.3387 ks.
    Total time required for tran analysis `tran': CPU = 11.341 ks (3h 9m 1s), elapsed = 11.3433 ks (3h 9m 3s).
    Time accumulated: CPU = 11.3436 ks (3h 9m 4s), elapsed = 11.3462 ks (3h 9m 6s).
    Peak resident memory used = 372 Mbytes.

    Any help would be appreciated. This is the first time my team has run into this one. Again, if I run a similar design in another process I do get a reduction in elapsed time vs CPU time (and the elapsed time is correct).

    --rg

  • I suspect you are only turning on the multithreading in spectre rather than using APS. Multithreading in spectre itself was only supported for a very limited amount of the processing for a limited number of devices, and nowadays it is not supported at all (I hope that we will at some point get the simulator to abort with a clear message rather than make you think it is supposed to do something).

    Whilst it is possible to set the number of threads as a simulator option, the right way to do this is via Setup->High Performance Simulation in ADE. You should turn on "APS" and then you can set the number of threads on this form (make sure you are not also setting it on the Simulation Options Analog form).

    If you are not turning on APS, I would expect the improvement to be limited (if anything), because that is the obsoleted mode that I refer to above.

    Regards,

    Andrew.

  • Andrew, your presence here is a gift that keeps on giving. That was it.

  • No. The first (limited) release of APS was in MMSIM7.1 (DC and transient only). Given that MMSIM7.0 was released in 2007 (so more than 7 years ago), and there have been 7 major releases of the simulator since then), why would you want to use such an old version? It makes no sense at all.

    Regards,

    Andrew.

  • Hi analogy,

    > Is APS supported in MMSIM 7.0 ?

    From my search, and Andrew will know for sure, APS is supported for "vanilla" transient simulations and harmonic balance simulations (not shooting) in MMSIM 7.2. I did not find any references to APS support in MMSIM 7.0.

    Shawn

  • Oh, our school is using Cadence 6.1.6 and MMSIM 7.0 .I have asked to upgrade to a newer version of MMSIM, but it might take a while.No wonder I cannot use ASU's predictive model files, because I believe, BSIM CMG was first added in cadence e 7.2, isn't it ?

  • Indeed - we've already had that conversation in http://community.cadence.com/cadence_technology_forums/f/38/t/31531

    Andrew

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.