• State Resolved
  • Locked Locked
  • Replies 6 replies
  • Subscribers 99 subscribers
  • Views 239 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.

question about link performance

bin li2
Expert 1290 points

hi,

     recently i think about a interesting question . for example , i use a simple chain:

capture link -> dei link

capture use two instance to process multi-data on one queue. it means capture only send queue0  to dei0 link.

if i create two queue in capture link ,and then ceate two dei links ,dei0 and dei1. then capture send two queue data to dei0 and dei1 seperatly.

i dont know which can get  better performance?

thanks alot

  • 0
    TI__Guru 59700 points

    If you use two different DEI IP instances (DEI_HQ_0 and DEI_0) then performance will be better using 2 queues. If you are using the same DEI IP (DEI_0 and DEI_1 or DEI_HQ_0 and DEI_HQ_1) then there is not significant difference between using two link instances versus using 1 link instance. Using 1 link instance will be slightly better in terms of performance because there is no task switch between the two link instances but it will not be noticeable difference in performance.

  • 0 in reply to Shankari G
    Expert 1290 points

    thanks a lot.   after reading the link ,i know that memory difference between two plans.

    but now i still confuse, i have to design a chain which contains capture,nsf, osd and dei. i use vip0 portA 16bits to capture progressive frames and vip1 portA to capture interlaced frames. so i have to use two queues in capture . like below

    capture intance 0 (vip0 portA 16bits 1080P queue0) -> nsf0 -> osd0 -> dei0(bypass)

    capture intance 1 (vip1 portA 8bits 576I queue1) -> nsf1 -> osd1 -> dei1(not bypass)

    it is my idea, but i think it's not good.  because the dei i should use two , one is bypass mode ,and one is not . so i cant think better ideas.

    can you give me some suggestions? thank you !

  • 0 in reply to bin li2
    Expert 1290 points

    my new idea :

    capture(one queue) -> osd -> nsf (output two queues: queue 0) -> dei0

                                                     ->  nsf(output two queues: queue1) -> dei1

    is it better? but i dont know whether osd module can process interlaced datas?

     

    i tried this plan . and the osd is not clear when input interlaces datas ....the progressive datas is ok!

    how to improve it ? i have no idea.

  • 0 in reply to bin li2
    TI__Guru 59700 points

    If you require one dei in bypass mode and one dei in dei enabled mode you require 2 link instances. There is no other option.

    On 816x there are two different DEI IP so use one DEI for DEI bypass (DEI_HQ_0) and one DEI for DEI enabled (DEI_0).

    Add OSD after DEI. Why are you adding NSF ? Is it for 422 to 420SP format conversion ? If so you don't need NSF. Your chain can be:

    Capture --> que0 --> DEI_HQ0 (bypass) --> Merge --> OSD

                 --> que1 --> DEI_0 (enabled)     -->

     

  • 0 in reply to Badri Narayanan
    Expert 1290 points

    thank you very much . i use nsf to reduce the noise for pictures. because i can see many noised on screen. it is another question for me but not the most critical.

    now my confused question is how to use osd with dei for frame/field input at the same time.

    and you answered my another similar question . thank you very much!