Notify Event Registering

Jibim,

I don't think you should have to call Notify_eventAvailable() on the ARM side in order to get your notification from the DSP.  The error on the DSP side indicates that the ARM has not registered for the notification at the time the DSP has called Notify_sendEvent().  Where inside of the MessageQ example did you put these calls?  Did you move the location of the Notify_registerEvent() when adding the call to Notify_eventAvailable()?

Best regards,

    Janet

Hi Janet,

Thanks for the reply.

I put these inside messageQ example. The Event registering is done before Ipc_readConfig() and wait is done just before the cleanup section. Because I not getting any error in ARM side, I believe the event registering is done properly eve without calling Notify_eventAvailable(). 

Is there any documents explaining the API call sequence to be followed in ARM & DSP side from IPC (Notify, messageQ, RingIO etc..)?

Thanks

Jibin

Jibin,

The SysLink User's guide and SysLink API guide have documentation on the IPC modules.  You can access these guides from the links in the SysLink release notes.  I will try to reproduce the problem you are seeing.

Best regards,

    Janet

Jibim,

I tried to reproduce the failure you are seeing, using the TI816X platform.  On the hlos side of the messageQ app (in MessageQApp.c), I added the lines in red to the function MessageQApp_threadHandler:

    Osal_printf("MessageQApp_threadHandler entered\n");

    lProcId = MultiProc_self();

    semSyncHandle = OsalSemaphore_create(OsalSemaphore_Type_Counting);
    if (semSyncHandle == NULL) {
        Osal_printf ( "Error: Failed OsalSemaphore_create \n");
        return;
    }

    status = Notify_registerEvent(rProcId, 0, APPNOTIFY_EVENT_NO_SYNC,
            (Notify_FnNotifyCbck) rtosInfocallback, (Ptr)semSyncHandle);

    /* Read the config info written by the slave. */
    do {
        status = Ipc_readConfig(rProcId, APP_INFO_TAG, &aInfo,
                sizeof(App_Info));
    } while (status == Ipc_E_FAIL);

    ...

    Osal_printf("Waiting for notification from processor %d\n", rProcId);

    OsalSemaphore_pend (semSyncHandle, OSALSEMAPHORE_WAIT_FOREVER);

    switch (cleanupCode) {

    ...

On the rtos side, I added the lines below in red, just before cleanup.  I had to change the Task_sleep() value to 3000 to get a 3 second delay:

    Task_sleep(3000);

    System_printf("Sending a Notify to ARM\n");

    do {
        status = Notify_sendEvent (rProcId, 0, (APPNOTIFY_EVENT_NO_SYNC),
                0x1234, TRUE);
        System_printf("Sendt a Notify to ARM status = %d\n",status);
    } while (status < 0);

    /* Cleanup. */

I did not need to call Notify_eventAvailable() for this to work.  This is the output I got from running the message sample:

# ./messageapp_debug 1 DSP ./message_dsp.out

MessageQApp sample application
Entered MessageQApp_startup
Loading and starting procId [0] with [./messageq_dsp.out]
Attached to slave procId 0.
Loaded file ./messageq_dsp.out on slave procId 0.
Started slave procId 0.
SlaveLoader_startup status [111693824]
After Ipc_loadcallback:
    status [0x0]
After Ipc_startcallback:
    status [0x97d2000]
Leaving MessageQApp_startup 97d2000
Entered MessageQApp_execute
MessageQApp_threadHandler entered
Registering heapId 0 with MessageQ for procId: 0
MessageQ_create name MSGQ_30  status [0x0] : procId [0]
Sending synchronizaion notification to ProcId: 0
Sent synchronizaion notification to ProcId: 0
MessageQ_open Status [0x0] : procId [0]
MessageQApp_queueId  [0x0] : procId [0]
Sending a message #100 to 0
Sending a message #200 to 0
Sending a message #300 to 0
Sending a message #400 to 0
Sending a message #500 to 0
Sending a message #600 to 0
Sending a message #700 to 0
Sending a message #800 to 0
Sending a message #900 to 0
Sending a message #1000 to 0
Waiting for notification from processor 0
Received event 0x1234 from procId [0]!:
Leaving MessageQApp_threadHandler 0
Leaving MessageQApp_execute
Entered MessageQApp_shutdown()
After Ipc_stopCallback status: [0x97d2000]
Shutting down procId [0]
Stopped slave procId 0.
Unloaded slave procId 0.
Detached from slave procId 0.
SlaveLoader_shutdown status [0]
Leaving MessageQApp_shutdown() (0x0)

(I renamed the dsp executable to message_dsp.out)  What value did you set APPNOTIFY_EVENT_NO_SYNC to?  I set it to 20 in my test.

Best regards,

    Janet

Hi Janet,

Thanks for the reply.

The clock frequency was 32Khz and actual frequency is 20Mhz.  I changed this and now the notify is working fine. But I have another problem now.

I also integrated RingIO example to messageQ example as I did for Notify. But the DSP side is giving error RingIO_E_NOTFOUND fro RingIO_open(). Can you help me on this?

Thanks

Jibin

Hi Jibin,

I'm glad to hear that changing the value for the clock frequency fixed your Notify problem.  Regarding the RingIO_open() failure, were you able to run the RingIO sample as is, and not integrated into the MessageQ example?   In your case, it may be that the DSP is returning RingIO_E_NOTFOUND because the RingIO object has not been created on the GPP.   Are you calling RingIO_create() on the GPP?

The RingIO sample on the rtos side calls RingIO_open() in a loop:

    do {
        Error_init (&eb) ;
        status  = RingIO_open (RingIOApp_instName,
                               &rioOpenParams,
                               NULL,
                               &rioReaderHandle);
        if ((rioReaderHandle == NULL)|| (status < 0)){
            System_printf ("ERROR:Failed to open RingIO in Read mode."
                           "status 0x%x\n",status) ;
        }
    } while (rioReaderHandle == NULL) ;

This is probably to wait for the GPP side to create the RingIO object.  Are you also calling RingIO_open() in such a loop?

Best regards,

   Janet

Hi Janet,

In Arm side everything is ok. No error in MessageQ as well as RingIO. But in DSP messageQ is working but RingIO is giving not found error.

From my experiments I found that If I open RingIO before MessageQ, RingIO is working but MessageQ will not work. It seems to be some conflict between the two.

Do you have any pointers on this?

Thanks

Jibin

Jibin,

I am not aware of any conflicts between MessageQ and RingIO.  Does MessageQ on the DSP side also return NOTFOUND when you open the RingIO object first?  It sounds like there's a timing issue.  What are the names of the RingIO and MessageQ objects?  And, are you looping on the DSP side until you successfully open the object?

Best regards,

    Janet

Hi Janet,

The issue is solved by calling Syslink_setup() from the task in which the RingIO is accessed. Thanks for your valuable support.

I have one more question for you. Do you have any information on McASP driver for DM8148 in SYSBIOS? This will help me to reduce number of RingIOs ...

Thanks

Jibin

Hi Jibin,

Glad to hear you got your example running.  You may want to post your question regarding the McASP driver in a new thread on on BIOS forum, since I don't know the answer to this question.

Best regards,

    Janet