• State
  • Locked Locked
  • Replies 4 replies
  • Subscribers 82 subscribers
  • Views 1782 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.

FPU, TMU of 28377

vesgine
Expert 1315 points

1. on SPRUHM8E, it mentioned that "The TMU extends the capabilities of a C28x+FPU by adding instructions and leveraging existing FPU instructions to speed up the execution of common trigonometric and arithmetic operations". Does it mean that TMU is only supported by C28x+FPU but not CLA?

2. Since the FPU extends the capabilities of the C28x fixed-point CPU, So there are two ways of implementing floating-point tasks.

   i. Just use the native FPU;

   ii. Use IQmath libraries

   So which method will be more efficient under the same condition?

  • 0
    Guru 192875 points

    Hi,

    vesgine said:
    Does it mean that TMU is only supported by C28x+FPU but not CLA?

    Yes, as there doesn't seem to be any info on TMU support for CLA.

    vesgine said:
       So which method will be more efficient under the same condition?

    This depends as in few cases iQMath is better than FPU else FPU is very powerful in itself.

    Regards,

    Gautam

  • 0 in reply to Gautam Iyer
    Expert 1315 points
    Hi Gautam,

    As it depends on cases, would you please show me the details? Are there any execution efficiency comparison info of IQMath and FPU? Or, is there any suggestion about choosing floating-point method?
  • 0 in reply to vesgine
    Guru 192875 points
    There's no comparison doc as such from TI. But you can search the forum for some info on the same. Here's one of them:
    e2e.ti.com/.../167007
  • 0
    TI__Mastermind 20850 points

    Hi,

    vesgine said:
    Does it mean that TMU is only supported by C28x+FPU but not CLA?

    Yes, as Gautam pointed out its only supported by the C28x+FPU32

    vesgine said:

    2. Since the FPU extends the capabilities of the C28x fixed-point CPU, So there are two ways of implementing floating-point tasks.

       i. Just use the native FPU;

       ii. Use IQmath libraries

       So which method will be more efficient under the same condition?

    With the TMU you get a significant bump in computation time if you do any trigonomentric, sqrt and division. If you dont use these functions, you generally see better performance using the FPU. This is not always the case, sometimes IQMath will give you comparable perfomance - but in general you get better performance using the FPU.

    You do get better precision with IQMath than with FPU. this is just due to the representation of numbers. I believe there is a very good presentation on the difference between IQ and float on our processors.wiki.ti.com. I think its in the c2000 training material.