• State Resolved
  • Locked Locked
  • Replies 11 replies
  • Subscribers 104 subscribers
  • Views 1386 views
  • Users 0 members are here
Support feedback
Options
Options
Similar topics

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.

AM335x ISDK Know issue and limitation

UNA
Mastermind 7980 points
Other Parts Discussed in Thread: SYSBIOS

Hi,

I check the known issues and limitations about AM335x ISDK.

AM335x SYSBIOS Industrial SDK 01.01.00.05 Release Notes
http://processors.wiki.ti.com/index.php/AM335x_SYSBIOS_Industrial_SDK_01.01.00.05_Release_Notes#Known_Issues_And_Limitations

I found the Ethernet/IP issue(SDOCM00106849).
Workaround for this issue has been described as follows.

"Increase the size of ICSS Driver Tx Queues from 3Kb to 6 Kb"


Please tell me how to increase the size of Tx Queses.
Do I need to modify the source code?
If so, which source file in ISDK should I modify?


Best regards,
H.U

  • 0
    TI__Guru**** 393215 points

    Moving this to the SYS/BIOS forum.

  • 0 in reply to Biser Gatchev-XID
    Guru 10570 points

    Hello,

    I also need this information.
    Can I have your response?

    Best regards, RY

  • 0 in reply to RY9983
    TI__Intellectual 1050 points

    RY-San,

    To increase the ICSS Driver TX queues from 3KB to 6KB, we need to modify 'icss_switch.h' file which can be found at “(IA_SDK_HOME)\public\sdk\os_drivers\include\ICSS" directory.

    Replace the following code

    // Physical Port queue size. Same for both ports

    #define QUEUE_1_SIZE                                97           // Network Management high

    #define QUEUE_2_SIZE                                97           // Network Management low

    #define QUEUE_3_SIZE                                97           // Protocol specific

    #define QUEUE_4_SIZE                                97           // NRT (IP,ARP, ICMP, …)

    with this 

    // Physical Port queue size. Same for both ports

    #define QUEUE_1_SIZE                                64           // Network Management high

    #define QUEUE_2_SIZE                                64           // Network Management low

    #define QUEUE_3_SIZE                                64           // Protocol specific

    #define QUEUE_4_SIZE                                194           // NRT (IP,ARP, ICMP, …)

    Rebuild the os_drivers and then ethernetip_adapter application for the changes to take effect.

    Regards,
    Nijin

  • 0 in reply to Nijin P
    Guru 10570 points

    Nijin-san,

    Thank you so much!!
    Sorry, I can not verify your post for some reason.

    BTW, I can not understand the meaning of numerical value.
    It means modification from 3KB to 6KB?

    Best regards RY

  • 0 in reply to RY9983
    TI__Intellectual 1050 points

    RY-San,

    We have 4 physical port queues (Q1 to Q4) which are same for both ports. To calculate the queue size, we have to multiply queue size offset with the 'Block Size' which is 32 bytes as derived from ICSS architecture.

    By default, we have the queue size offset as 97 which stands for 3 KB (97*32 Bytes). It is 97 instead of 96 because we use one offset to detect FIFO is full or not.

    During Queue optimization, we increments the Q4 size from 3 KB to 6 KB with a proportionate decrease in first 3 queue sizes so that the total queue size (12 KB) remains constant before and after the optimization process.

    With this optimization, we can see that the NDK performance (recv.exe) is increased considerably. 

    Hope these info helps.

    Regards,
    Nijin

  • 0 in reply to Nijin P
    Guru 10570 points

    Nijin-san,

    Thank you very much !!

    Best regards, RY

  • 0 in reply to RY9983
    Guru 10570 points
    Nijin-san,
    Thank you so much for your comment!

    I would like to clarify:
    - Do you have a plan to fix this issue in next ISDK release(ISDK v1.1.0.6)?
    - Can I think this workaround is sufficient reliable? (Queue size changing)

    Because,
    our another project team encountered unstable system trouble in PROFINET after they modified QUEUE_4_SIZE as 194 byte.
    Note: In Ethernet/IP, we have never encountered issue.
    We understand its workaround is for Ethernet/IP, and not PROFINET.
    However, we would mind it because this issue is not described in TI's document currently.

    Could you please your comment?

    Best regards, RY
  • 0 in reply to RY9983
    TI__Expert 7360 points
    RY,

    You cannot increase the queue size in Profinet because of memory constraints. This is the reason why it's not set by default.

    It should work fine with Ethernet/IP.

    Regards,
    Vinesh
  • 0 in reply to Vinesh Balan
    Guru 10570 points
    Vinesh-san,
    Thanks so much.
    It will not update next release of ISDK. Right?
    Best regards, RY
  • 0 in reply to RY9983
    TI__Expert 7360 points

    RY-san,

    It will not be update as it may effect other protocols.

    Regards,

    Vinesh

  • 0 in reply to Vinesh Balan
    Guru 10570 points
    Vinesh-san,
    Thank you so much !!
    Best regards, RY