This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

Slow ccs v6.1.2 compilation in Simulink

We have managed to make Simulink build, load and run a project on our F28377D. However, no matter which configuration we use it take 4 minutes to complete this task even for the simplest project, on our >>brand new<< PC. We have tried "faster builds", "optimization" etc. parameters with trivial improvements. On top of that, even when the project is finally compiled, it waits for 1.5 minutes to complete "Downloading to target hardware".

1) Could you please advise me why is it so slow? F28335 with CCSv3.3 builds, loads and runs in less than 0.5min!

2) Is there an option in Simulink Options/Code Generation (below) that I can activate parallel builds in Behavior of CCS at least?

Thank you for your time,

Panagis Vovos

Lecturer
University of Patras

  • Hi Panagis,

    This is clearly a MathWorks question, so you may get a better response posting on MATLAB Answers, or contacting MathWorks Technical Support.

    I'll also ask about this through my sources.

    Thanks,

    -Brian

  • Hello Brian,

    Thank you for your reply. I have posted my question to Mathworks Technical Support. However, my 2nd question could be answered by TI, since I couldn't find a command-line argument that would order parallel compilation. Command line arguments are just specified in the "options" pane in Simulink.

    Thank you for your time,

    Panagis

  • Hi Panagis,

    I learned a couple reasons why the build process for F28377D is taking more time:

    1) Our build infrastructure is compiling files one by one. As we have more files included for F28377D compared to F28335 it will take little more time for F28377D.

    2) For F28377D by default we are enabling Boot from flash option. Loading code to flash will take always more time than loading code to RAM. For F28335 default behavior is RAM. Also loading code to Flash is taking more time for new processors even with CCS.

    We also see an increase of build time in our testing environment from CCS, but we need to investigate this further.

    Thanks,

    -Brian

  • Thank you Brian for your reply. However, we have not set our code to load from flash. Your 1st explanation made us spend money for a super computer, which loads code in approx. 30sexonds. I do not think this is exactly efficient.
    Thanks again,

    Panagis
  • Hello Panagis,

    Do you require further support on this?

    Best Regards,
    Adam Dunhoft
  • Hello Adam,

    Thank you for your interest. I do not think compiling on a super computer is the optimum solution to slow compilation. I am pretty sure, there is a small issue somewhere in the process. A more detailed description of the reasons causing the problem from TI side (CCS) or Mathworks (Simulink/Matlab) would be highly appreciated from the DSP community and allow us to investigate it further.

    Best regards,

    Panagis Vovos

  • Problem solved:

    CCS doesn't use parallel building by default. You have to set it in command line by appending the -j X (where X:CPU cores) at the end of the gmake command. To do that with Simulink's fast prototyping you have to append it at the end of the Code Generation/Build Configuration/Specify/make tool options of the model configurations parameters.

    Regards,

    Panagis