Post Go back to editing

System Crossbars (SCB)

Category: Datasheet/Specs
Product Number: ADSP-21569

Hi.

I have some questions about System Crossbars (SCB).

1. Where is the Sharc+ core SCB on the Figure 41-3: SCB Block Diagram?

2. How can C0_I get access to L1 according Figure 41-5: SCB Block Diagram - Interconnections?

3. On 41–12 says

S2 slave port of SHARCs is intended primarily for MDMA between the L1 of the 2 SHARCs. Hence, only the Max BW MDMA is given access to S2 ports while all others masters are given access to S1 port.This arrangement also means that S1 and S2 port now share the same address space as against the ADSP-2156x family of processors where an alias address space was maintained.

Can you explain about the sentence which contains "the 2 SHARCs"?

Can you explain about the sentence which contains "as against the ADSP-2156x family"?

4. How do SCB Arbitration work in the case which describe below.

My case is a simple interaction between the internal core action and external devices. So my program do some work which are based on data from external devices and internal state machine variables.

At the same time:

4.1. MDMA channels copy L1-to-L1 memory : from Source_MDMA_Destination_CORE_DataBufL1 to Destination_MDMA_Source_CORE_DataBufL1.

4.2. The SHARC+ copy L1-to-L1 memory :  from Destination_MDMA_Source_CORE_DataBufL1 to Source_MDMA_Destination_CORE_DataBufL1.

4.3. Source_MDMA_Destination_CORE_DataBufL1and Destination_MDMA_Source_CORE_DataBufL1 have the same memory bank.

Do SCB Arbitration work every 32 bit L1 memory transfer or every L1 memory transfer which has some package length?

Best regards,

Daim

  • Hi.

    I add a new question about System Crossbars (SCB).

    5.  Figure 41-5: SCB Block Diagram - Interconnections don't explain: How can the SHARC+ core do 128 bit L1 memory transfer thorough 64 bit PM and 64 bit DM buses for one core clock.

    Best regards,

    Daim

  • Hi Daim,

    We are checking on this and will get back to you as soon as possible.

    Regards,
    Nandini C

  • Hi!

    I think the full SCB Block Diagram is based on Figure 37-5: SCB Block Diagram - Interconnections of HRM and Figure 7-1: Access Types of PRM and Figure 8-2: Memory Cache Signal Flows of PRM.

    The data core port C0_D access to L1 through L1 DM ports of memory blocks0, blocks1, blocks2. blocks3.

    The instruction core port C0_I access to L1 through L1 PM ports of memory blocks0, blocks1, blocks2. blocks3.

    Requestors (DDE DMA Channels and Non DDE DMA Channels) access to L1 through L1 S1 and S2 ports.

    Am I right?

    Regards

    daim.

  • Hi,

    Yes, your understanding is correct.

    Please refer the below link for SCB arbitration:
    SCB arbitration/priority scheme 

    When the DMA and the Core attempt to access the same L1 memory block at the same time, it may result in a memory access conflict(block conflict).

    To ensure continuous data streams, mechanisms like ping-pong buffers, together with chained DMA transfers, can be implemented. Designs should ensure that while the DMA moves data to the primary memory block, the core processes the secondary block's data. Then, after the DMA interrupt is generated, the memory block processing between core and DMA is flipped which prevents memory block conflicts between the core and DMA.

    We recommend referring to the Memory Block Arbitration section in the PRM for detailed information about the limitations of accessing the same block and the available mitigation options.

    Regards,
    Nandini C

Before You Switch


Switching languages will make ADI Explorer unavailable. Resume your session by switching back to English and reopening ADI Explorer.