• State Resolved
  • Locked Locked
  • Replies 6 replies
  • Answers 2 answers
  • Subscribers 100 subscribers
  • Views 490 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.

Compiler: Optimization using EVE

Neha Bhosale69
Expert 1085 points

Tool/software: TI C/C++ Compiler

Hello Everyone,

My concern is optimization of LUT generation. The LUT creation has floating point operations. Can you suggest some ways to obtain optimization for LUT creation on DSP or EVE(if possible).

Also I would like to know if there is any function available in DSPLIB or IMGLIB which performs pixel interpolation(bilinear).

Thanks,

Neha

  • 0
    TI__Guru**** 253230 points

    I presume LUT means lookup tables.  Unfortunately, we compiler experts lack knowledge of LUT, displib, and imglib.  Exactly which TI DSP do you plan to use?  That will help me redirect your question to the best forum.

    Thanks and regards,

    -George

  • 0 in reply to George Mock
    Expert 1085 points
    Thank you for the quick response.
    Yes, you are right LUT means lookup tables(for image). I am working on TDA2px and currently using DSP1 for processing all operations. I would also like to use EVE, but as all operations are on floating points I am not able to.
    Can you suggest any optimization that can be done on EVE and also on DSP?
  • 0 in reply to Neha Bhosale69
    TI__Guru 56820 points

    Neha,
        EVE supports a maximum 8 way LUT which means that you can load 8 element from Lookup table in single cycle ( given that you are able to store 8 copies of your lookup table in EVE's internal memory). But as EVE is a fixed point processor, you will have to convert your lookup table to fixed point. You can use Q format for representing the floating point LUT to fixed point. 

    Regards,

    Anshu

  • 0 in reply to Anshu Jain
    Expert 1085 points
    Hello Anshu,

    Thank you for the reply.
    May I know if there is any function available in DSPLIB or IMGLIB which performs pixel interpolation(bilinear).
  • 0 in reply to Neha Bhosale69
    TI__Expert 5655 points
    Neha,

    Can you give more information about LUT generation. Are you referring generating a LUT as part of histogram equalization? You may want to consider looking at the VXLIB optimized C66x DSP library that is delivered as part of the Processor SDK - Vision release. There are some source examples of LUT generation as part of histogram equalization, as well as other remap function which use a floating point LUT as input and generates interpolated output pixels from an input image based on this LUT. From your description of LUT generation however, I wasn't clear if you were referring to this or not.

    Jesse
  • 0 in reply to Jesse Villarreal
    Expert 1085 points
    Thank you for the response.
    Your referring point is exactly right, and I want to do the same processing .I will check for VXLIB optimized C66x DSP library.
    Your response is of great help.