• State TI Thinks Resolved
  • Locked Locked
  • Replies 3 replies
  • Answers 1 answer
  • Subscribers 82 subscribers
  • Views 430 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.

TMS320F280025: IQ math execuate efficiency

Emma Wang
Expert 5200 points
Part Number: TMS320F280025

Hi expert,

My customer is porting from F28034 to F280025.

And they use IQmath for their code, due to tight schedule, they can not change to TMU all. If they use [IQ to float – sin/cos using TMU – float to IQ] for the critical algorithm. Will this method save time?

Or could you kindly tell the speed of  [IQ to float] or [float to IQ]?

BR

Emma

  • 0
    TI__Expert 5410 points

    Hi Emma,

    Please refer Table 5.3 in IQ Math User and FASTRTS guide for the benchmarks. 

    _IQ24sin takes 42 cycles. (assume N=24. cycles will vary for other values of N)

    _IQN is a macro and takes 0 cycles.

    IQNtoF takes 22 cycles

    Running sin using FASTRTS and TMU takes 63 cycles for double precision and 38 cycles for single precision.

    Hence total cycles for approach with TMU is 85 cycles for double precision and 60 cycles for single precision

    So in either case (single or double precision), it would be faster to use _IQsin.

    Hope this helped.

    -Shantanu

  • 0 in reply to Shanty
    TI__Expert 5200 points

    Shantanu，

    thanks for your quick response. it is helpful.

    however, I  test the __sin(test_a) function in my code, it only cost me 16 cycles.

    so where did you get the result of 38 cycles.

    and I found only sin() function without " __" can also be more quicker after I enable TMU.

    I am kind of confused on what is the real TMU instruction.

    • Is the TMU0/TMU1 affected?
    • or the __ and TMU0/TMU1 affected?

    BR

    Emma

  • 0 in reply to Emma Wang
    TI__Expert 5410 points

    Emma,

    I got those numbers from the FASTRTS user guides in the benchmarking section. I am talking about the sin() function provided by FASTRTS using the FPU and tmu0. Can you verify if we are talking about the same APIs? I am not sure about the API with "_". Which library is this from?

    -Shantanu