• State
  • Locked Locked
  • Replies 2 replies
  • Subscribers 99 subscribers
  • Views 198 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.

TDA4VH-Q1: VXLIB remap performance optimization

hongyao.jin
Expert 1230 points

Part Number: TDA4VH-Q1
Other Parts Discussed in Thread: TDA4VH

Tool/software:

TDA4VH custom board

SDK 11.00

freertos+linux

The four-way surround view data we obtained from the cameras needs to undergo the processing shown in the above figure.

Currently, we are attempting to optimize the performance of the "stitch" node.

After our investigation, we found that VXLIB_remapBilinear_bc_i8u_i32f_o8u will consume 32ms. The input size is 1920*1536*4, and the output size is 512*512.

1.We would like to know if VXLIB_remapBilinear_bc_i8u_i32f_o8u has reached its limit for our specific case? Is there still room for improvement?

2.Or is there a more efficient function available to replace VXLIB_remapBilinear_bc_i8u_i32f_o8u?

3.At the same time, we are also attempting to remove the LDC_CC node. Therefore, we would like to know if the remap function in the VXLIB library supports converting UYVY to NV12 while performing the remap operation, because the input of the preproc node must be NV12.

  • 0
    TI__Expert 8650 points

    Hi,

    Let me review your question and get back to you. From my initial assessment, it appears that the API you're using is not optimized for C7x. I will look into it further and get back to you shortly.

    Regards,
    Sivadeep

  • 0
    TI__Intellectual 1940 points

    Hi,

    hongyao.jin said:
    1.We would like to know if VXLIB_remapBilinear_bc_i8u_i32f_o8u has reached its limit for our specific case? Is there still room for improvement?




    The VXLIB_remapBilinear_bc_i8u_i32f_o8u function you are using is optimised for c66x not for c7x processors.

    hongyao.jin said:
    2.Or is there a more efficient function available to replace VXLIB_remapBilinear_bc_i8u_i32f_o8u?

    No,there is no other alternative function in VXLIB which can replace the purpose of  VXLIB_remapBilinear_bc_i8u_i32f_o8u.You can improve the performance by processing the data from L2 RAM.

    hongyao.jin said:
    3.At the same time, we are also attempting to remove the LDC_CC node. Therefore, we would like to know if the remap function in the VXLIB library supports converting UYVY to NV12 while performing the remap operation, because the input of the preproc node must be NV12.

    The standard VXLIB remap function (VXLIB_remapBilinear_bc_i8u_i32f_o8u) does not natively support color space conversion from UYVY to NV12 as part of remap.

    Regards,
    Shabary S Sundar