• State Resolved
  • Locked Locked
  • Replies 7 replies
  • Subscribers 80 subscribers
  • Views 4597 views
  • Users 0 members are here
Support feedback
Options
Options
Related

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.

MSP430FR5994: LEA Setup/Operation

Lawrence Wong
Genius 15760 points
Part Number: MSP430FR5994
Other Parts Discussed in Thread: BOOSTXL-AUDIO,

Customer working with the FR5994 LaunchPad and Booster packs.  Having issues trying resolve is with the Filtering and Signal Processing with LEA example.

 

The users guide (section 4.1) says the complex FFT using LEA requires 5,578 cycles and 77,000 without using LEA.

 

When customer run the example code it’s showing 134792 cycles using LEA and 77,088 without LEA.

 

They did verify that DSPLIB.lib/include/DSPLib_lea.h has MSP_LEA_REVISION set to MSP_LEA_REVISION_B because the FR5994 on the LaunchPad is Rev C

The FIR mode is not working either but they have not investigated.

  • 0
    TI__Genius 15370 points
    Hi Lawrence,

    Did the customer make any changes to the example before trying to run the code? By default the code uses LEA unless the target device is not an FR5994. Can you have them check if MSP_DISABLE_LEA has been defined anywhere? Also, when the FFT executes can you have them check in msp_cmplx_fft_fixed_q15.c if the function within the "#if defined(MSP_USE_LEA)" is being executed or if it is the CPU version?

    Using the incorrect revision wouldn't cause the function to take longer to execute. Instead an error would cause the function to exit prematurely and the cycle count would be much lower than expected.

    Best regards,
    Caleb Overbay
  • 0 in reply to Caleb Overbay
    TI__Genius 15760 points

    Here are some updates from the customer:

     

    There is a code set include with the out of the box demos identified as  BOOSTXL-AUDIO_430BOOST-SHARP96_FFT_FIR_LEA_MSP430FR5994 and a 2nd set that is identified as TDIM Filter and Signal Processing with LEA which I downloaded from the TI site.

    Both behave the same with regard to the # of cycles with LEA, but there are quite a few code differences.

     

    Can you point me to the best code set to use?

     

    Regarding your questions, I’m looking in the Out of the box version of the code.

     

    MSP_DISABLE_LEA is NOT defined.

     

    In msp_cmplx_fft_fixed_q15.c the code within the #if defined(MSP_USE_LEA) is NOT compiled in.

     

    They did find MSP_USE_LEA defined as 1 in DSPLib_lea.h, however it appears that another #if in that file is not allowing the inclusion of #if defined(MSP_USE_LEA).

    In other words the preprocessor is not including the MPS_USE_LEA define, so that’s a big issue.

     

    They did a put a bit of jibberish that would cause a compile error just before the #if define(MSP_USE_LEA) and the code compiled fine so they are invesitigting further

     

     

  • 0 in reply to Lawrence Wong
    TI__Genius 15370 points
    Hi Lawrence,

    The fact that the #if defined(MSP_USE_LEA) is not compiled leads me to the suspicion that you may have an older version of either the FR5994 header file or an older version of DSPLib. Can you provide what CCS version you are using along with the DSPLib version?

    Best regards,
    Caleb Overbay
  • 0 in reply to Caleb Overbay
    TI__Genius 15760 points
    Caleb,

    CCS V6.2 with all the updates.

    The DSPLib came with the 2 software bundles (out of the box demos) and the TDIM so they are not sure what version it is.
  • 0 in reply to Caleb Overbay
    TI__Genius 15760 points

    Caleb,

    Do they need to upgrade to a newer CCS version?

    Regards,

    Lawrence

  • 0 in reply to Lawrence Wong
    TI__Genius 15370 points

    Hi Lawrence, 

    I recommend upgrading to the latest CCS version. I've run the code on this version and it's operating as expected. I think the root cause of this issue may be some kind of version mismatch.

    Best regards, 

    Caleb Overbay

  • 0 in reply to Caleb Overbay
    TI__Genius 15760 points
    Customer switched to CC7 and performance matched the documentation.

**Attention** This is a public forum