• State TI Thinks Resolved
  • Locked Locked
  • Replies 1 reply
  • Answers 1 answer
  • Subscribers 101 subscribers
  • Views 959 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

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.

LU inverse matrix

C. García
Intellectual 830 points

 

Hi,

I ran from the DSPLIB 3_4_0_0 version the following routines in order to get 128 matrix inverse calculation:

DSPF_sp_lud and DSPF_sp_lud_inv on C6678 EVM (TMDSEVM6678LE) running at 1 GHz frequency, Little endian ELF format. I took the TI original project  DSPF_sp_lud_inv_66_LE_ELF

I received the following result:

DSPF_sp_lud_inv=30 mSec   The TI test report written DSPF_sp_lud_inv= 15318956 cycles means 15.3 mSec.

1.- Please, could you explain me why if I run the TI project I receive twice the Test Report time for the same routine?

 I have ran also the code from the book Numerical Recipies in C (William H. Press, Saul A. Teukolsky) second edition and the time I received was 6.9 mSec.

2- Could you tell me if there is any advantage for using one routine or another because the time I received using the book's routine is significantly slower?

Thanks in advance.

Carlos.

  • 0
    TI__Expert 4680 points

    Carlos,

    Please share the numbers you have measured by running the algorithm. The numbers listed on the DSPlib report is for flat memory model and as per the test results matrix upto 64 shows similar performance on device  and flat memory system. There are some optimization for memory will be needed for bigger size matrix.

    Please share the algorithm you have used and confirm whether you are measuring the performance on device for 128 matrix size.

    Regards

    Asheesh