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.

8127 algorithm optimization

Hello everyone !

I'm using IPNC_RDK ver3.5 with 8127 .I have implemented an alg and run with it successfully. I have tested it and it takes more time  than i can accept. I have also found it tales about 110ms to run a function called RGB2GRAY which is convert color space from RGB to GRAY (with resolution 352*288).I think it is too much long for me .For the reason,i think it is because my data is in DDR and it takes a lot of time to load data from memory .Now I don't know how to verfiy and improve it ,so is there anyone can give me some advice.

regards

yu!

  • Hi Yu,

    Are you running this algorithm on the Cortex-A8 ARM, or C674x DSP? Have you enabled the cache?

    See if the below links will be in help:

    http://processors.wiki.ti.com/index.php/Optimized_Sort_Algorithms_For_DSP

    http://processors.wiki.ti.com/index.php/TMS320C6000_DSP_Optimization_Workshop

    http://e2e.ti.com/support/dsp/davinci_digital_media_processors/f/717/t/118850.aspx

    http://e2e.ti.com/support/dsp/davinci_digital_media_processors/f/717/t/317885.aspx

    http://e2e.ti.com/support/dsp/davinci_digital_media_processors/f/99/t/60797.aspx

    http://e2e.ti.com/support/dsp/davinci_digital_media_processors/f/99/t/59795.aspx

    http://e2e.ti.com/support/dsp/tms320c6000_high_performance_dsps/f/115/t/184645.aspx

    Regards,
    Pavel

  • Thanks for your reply!

    I'm running my algorithm on the C674x dsp.My input is a video data written in a .txt file stored in the A8 .It was in the CMEM memory(0x85000000--0x89000000).And I pass the data address to the dsp so i can manipulation it..Now i can't use CCS for i don't have a proper emulator .I am not very familiar with all kinds of configuration files ,so i don't really know how to enable cache .As you know ,the time RGB2GRAY function costs quite a long time(100ms).In my intuition,I feel it is because loading data costs time(as in the CMEM).But in my alg's other part ,there is also similar operation using the video data,it seems they don't need much time .It really confuses me.

    regards

    yu