• State Resolved
  • Locked Locked
  • Replies 8 replies
  • Subscribers 82 subscribers
  • Views 3397 views
  • Users 0 members are here
Support feedback
Options
Options

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.

TMU or Fast RTS or FPU

Keyur Acharya
Expert 1740 points
Other Parts Discussed in Thread: TMS320F28075, CONTROLSUITE

Hi all,

I am using TMS320F28075 device. It has a TMU Type-0.

I have some confusion regarding the time of calculations for mathematical operations.

If i want to do some operations like clark transform, park transform, tan, and square root, which option would be most suitable?

1)FPU

2)TMU

3)Fast RTS library

  • 0
    Guru 192875 points
    Hi Keyur,

    TMU would be indeed effective in the above case. Please check this post: e2e.ti.com/.../1689616

    Regards,
    Gautam
  • 0 in reply to Gautam Iyer
    Expert 1740 points
    hi gautam,
    I checked the post which you mentioned here, the post is about TMU and FPU but there is no mention about "fast RTS library".
    I am confused whether to include fast RTS library if we want to use TMU?
    Which would be good?

    Regards,
    Keyur
  • 0 in reply to Keyur Acharya
    Guru 192875 points

    Keyur Acharya said:
    I am confused whether to include fast RTS library if we want to use TMU?

    I've used FPU+fast RTS library or FPU+TMU, but never FPU+TMU+Fast RTS! 

  • 0 in reply to Gautam Iyer
    Expert 1740 points
    That's what my question is !! :)
  • 0 in reply to Keyur Acharya
    Guru 192875 points
    Not tried but their respective functions should work but in case of colliding functions (supported by TMU and FastRTS) I'm unaware of who will take over. Do check whether there are any colliding functions at first!
  • 0
    TI__Mastermind 20850 points
    Hi keyur,

    Clark, park use sine and cosine. If you are doing these in floating point format you can turn on the hardware floating point unit (FPU). Now you have 3 options
    1. Call sin() and cos() from the standard run time support library (rts2800_fpu32.lib) that comes packaged with the CGT - these routines are more accurate than the fastRTS but take longer to execute as well
    2. add in the FPUfastRTS library (C:\ti\controlSUITE\libs\math\FPUfastRTS\V100\lib\rts2800_fpu32_fast_supplement.lib) to your project, give it search precedence over the normal RTS, and call the faster sin() cos() routines.
    3. if your hardware supports it, turn on the TMU support (--tmu_support=tmu0 and --fp_mode=relaxed) in addition to FPU support (--fpu_support=fpu32) and then call the sin(), cos()- what the compiler does in this case is replace those calls with TMU instructions. This is the fastest of the 3 options.
  • 0 in reply to Vishal_Coelho
    Guru 192875 points
    Thanks Vishal for the detailed description. Cleared few of my doubts :)
  • 0 in reply to Vishal_Coelho
    Expert 1740 points
    thanks for the clarification..