Post Go back to editing

why is the calculating efficiency is much different between fixed-point and float-point?

Category: Software
Product Number: ADSP-21489
Software Version: CCES-2.10.1

i write a demo code to test the calculating efficiency on platform sharc 21489. 

i found the float-point calculating is faster 60% than fixed-point calclating. The compiler configuration is same.

21489 should has fixed-point multiplication and fload-point multiplacation, so i do not understand why the calculating is much different.

  • Hi,

    SHARC is Floating point DSP. Floating-point DSPs are optimized for specialized, computationally intensive applications, whereas fixed-point DSPs are optimized for high-volume, general purpose applications. While it is possible to program fixed-point code for a floating-point processor and vice versa, significant performance limitations may arise. Performance efficiency can also impact power efficiency. For example, in a case where a floating-point processor is implemented to execute fixed-point tasks, the power draw might be greater than what it would be if a fixed-point processor had been implemented, which can impact power source and cooling requirements, as well as associated BOM costs.

    One alternative could be that you convert your input from fixed to floating point, do the processing in floating point on SHARC and then convert the result back to fixed point. Do you see any concerns with this approach?

    Regards,
    Divya.P

  • hi,divya

    Thanks for your answer and suggestion. I have finished the project coding now. it is used fixed-point. then i found this problem. Is it possible to optimize the calculating efficiency for fixed-point on platform sharc-21489? assembly instruction? I see sharc-21489 support 32bit fix-point in the product description. 

  • Hi,

    Yes, ADSP-21489 supports fixed point operations also.

    Please refer chapter 11 "COMPUTATION TYPES" which discuss about fixed point Computations. For information on syntax and opcodes, see Chapter 12,"COMPUTATION TYPE
    OPCODES" in below linked programming reference manual
    www.analog.com/.../adsp-2136x_2137x_214xx_pgr_rev2.4.pdf

    Regards,
    Divya.P