• State Resolved
  • Locked Locked
  • Replies 4 replies
  • Subscribers 83 subscribers
  • Views 675 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.

Loading floating point value in HRPWM

sunil barot
Intellectual 580 points

Other Parts Discussed in Thread: TMS320F28035

I want to load HRPWM period from the result of some floating point calculations done in the CLA of TMS320F28035.  Can anybody suggest an example code or method ? 

  • 0
    Prodigy 220 points

    The simplest way is to just use the MF32TOI32 instruction (or MF32TOUI32 ) to convert your float to a signed or unsigned integer respectively.

    Keep in mind that your floats have 24 bits of precision and also that the high resolution registers are significant in bits 15:8 and the specifications require that writes to bits 7:0 must be zero. 

    Generally, you want to do your calculations so that the float produces a positive 24 bit result that you will convert to unsigned integer and then left shift 8 bits with a MLSL32 instruction to align it. You can then write this result directly to your HRPWM period register with a single 32 bit write.

    Hope this helps.

     

     

     

     

     

     

  • 0
    Intellectual 580 points

    Thank you Ritchie,

    The problem got modified a bit from what it was.

    The CLA computes the frequency information in the form of clock counts (period). It transfers the  counts to CPU as a float variable. For example it computed the period to be 333.4105 counts that is represented by hex as 0x43A6B48C. I want to load HRPWM  with minimum amount of manipulation of the data. I do not want to loose the fraction. Ideally there should be a routine that excepts the float and loads HRPWM. Let me know your way.

     

    sunil

  • 0 in reply to sunil barot
    Intellectual 430 points

    Generally, it's easiest to treat this as 8 bits assuming you are using auto-conversion and the sfo library.

    You can Multiply your floating point count by 256 and then convert it to integer. This will produce a 24 bit result.

    Then  you can Multiply the integer by 256 with a shift left 8. This will give you a 32 bit integer result.

    You can then save this for the CPU or directly write it out to the HRPWM period register.

    Here is a code snippet that will do this 

    MMOVF32 MR0,#333.4105 ;simulate initial value

    MMOVF32 MR1,#256.0

    MMPYF32 MR0,MR0,MR1

    MF32TOUI32 MR0,MR0

    MLSL32 MR0,#8

    MMOV32 @_EPwm1Regs.TBPRDM.all,MR0

     

     

    It's far faster and easier to do this in the CLA becuase the CPU is fixed point and the CLA is native floating point.