[Concerto] IPC: MessageQ from Flash - without Debugger

Scott,

First of all: Thanks for your long answer!

Then, I am currently using sysbios v. 6.32.05.54 (from the name of the folder in my windows-installation). I am downloading the latest version, v. 6.33.05.46 just now, and once I figured out how to update it, I will try it with this version. ;)

And tell me if I get this right: I don't have to do any low-level calls when I use sysbios, correct? So, no need to call FlashInit(), no need for the low-level IPC command to startup the c28 core, and no SystemClock-Calls like SysCtlClockConfigSet(....)? But on the other hand, I do need lowlevel Calls for the GPIO-configuration, because there is no GPIO support for Concerto in sysbios yet, correct?

OK, I will try the newer Sysbios version now, and report on my advances later :)

UPDATE:

I updated my software successfully, created a clean new project, and after some try&error, it really works now! - From Flash and everything!! yay :)

The next thing is, I want to transfer some "real" information between the two cores, not only the MessageID. So, I defined a custom Message:

typedef struct MyMsg {
MessageQ_MsgHeader header; // Required
Int32 myInt;
char myString[15];
} MyMsg;

Then, I changed the commands (to the new data type MyMsg) to allocate Heap and to open the MsgQ and so on, and it still works well.

But how do I now write something into my new defined message, and how can I read this out on the other core?

Scott, could you (or someone else) post me a "code-snippet" on how I could write for example the value 4711 in my "myInt"-field in the structure, and another "code-snippet" on how I would read this out on the other core?

As always, I am very thankful for every answer. You do a great work!

- Philipp L.

Philipp,

I’m glad to hear you are booting now!  

Right, the Boot modules for M3 and C28 can handle the low-level details to manage Flash booting.  The M3 Boot module can also configure the PLLs and clock dividers. 

The configuration of the Flash controllers (setting wait states and enabling of caches while executing from RAM) has been recently added, but not released externally yet.  This will be available in an XDCtools release scheduled for the end of the month.  Until then, you can leave the Flash controllers in their default configuration for booting.

And yes, GPIO is not managed by SYS/BIOS.

I will need to search for some MessageQ snippets to send…

Scott

Here's an (uncompiled) example:

Void example()
{
MyMsg msg;
...

msg = MessageQ_alloc(HEAPID, sizeof(MyMsg));

MessageQ_setReplyQueue(msgqHandle, (MessageQ_Msg)msg);

msg.myInt = 0x4321;
MessageQ_put(queueId, (MessageQ_Msg)msg);

MessageQ_get(msgqHandle, (MessageQ_Msg)&msg, MessageQ_FOREVER);
System_printf("received: 0x%x\n", msg.myInt);
}

Ooops, those MessageQ_* calls should have &msg in them as in

MessageQ_put(queueId, (MessageQ_Msg)&msg);

Gack! Try again (memo to Mark: don't try to respond while on conf call :)

typedef struct MyMsg {
MessageQ_MsgHeader header; // Required
Int32 myInt;
char myString[15];
} MyMsg;

Void example()
{
MyMsg *msg;
...

msg = (MyMsg *)MessageQ_alloc(HEAPID, sizeof(MyMsg));

MessageQ_setReplyQueue(msgqHandle, (MessageQ_Msg)msg);

msg->myInt = 0x4321;
MessageQ_put(queueId, (MessageQ_Msg)msg);

MessageQ_get(msgqHandle, (MessageQ_Msg *)&msg, MessageQ_FOREVER);

System_printf("received: 0x%x\n", msg->myInt);
}

Ok, big thanks to you two, Mark and Scott :)

It works. Perfectly.

The only problem is: when I make a char[xx] (string)-field in the struct, I have the problem, that the c28 has only 16bit data width, and the m3 can do 8bit.

That means, char is defined as 16bit on c28, and as 8bit on m3.

So, if I send the string "HELLO" from c28 to m3 over a string field in my struct (as shown in my earlier post), I receive this string: "H\0E\0L\0L\0O\0". 

My ideas for a solution are:
- I could write a special "strcpy" function on c28, which would put two 8bit chars into one 16bit char field, which would be interpreted correctly on the m3
- Perhaps I could define the char[xx] field on the m3 as int[xx], and after receiving do some kind of casting to char / or filter the '\0' out somehow else.

What would you recommend? Or is there an exisiting solution? I think, performance in my application is more important on the c28 side, so the second solution would be better for the performance - but it would take more of the shared memory...

-Philipp

Philipp,

You could take either approach, depending on whether C28 performance or shared memory footprint is a higher priority for you.

In case it helps, I’ve included below a function that does packing on the C28 side. The C compiler provides a __byte() intrinsic that you might use too, to simplify the below code.

Scott

------

/*
 *  ======== NameServerBlock_strncpy ========
 *  Copies the source string into the destination string as a
 *  packed string.  The length includes the null terminating
 *  character.
 */
Char* NameServerBlock_strncpy(Char *dest, Char *src, SizeT len)
{
    Int i;
    UInt half = len >> 1;

    /* copy and pack the string */
    for (i = 0; i < half; i++) {
        dest[i] = src[i << 1] | (src[(i << 1) + 1] << 8);
    }

    /* The terminating character */
    if (len & 1) {
        /*
         *  (len = odd) ==> name has even number of characters.
         *  The next byte is (len >> 1), so zero that out.
         */
        dest[half] = '\0';
    }
    else {
        /*
         *  (len = even) ==> name has odd number of characters.
         *  Zero out the upper half of the last byte.
         */
        dest[half - 1] |= ('\0' << 8);
    }

    return (dest);
}

Scott,

Thank you very much. Again :) I'll work with this.

I'll mark this thread as "answered" then. (I'll take one of your answers from above, because the actual topic of this thread was already answered there..)

Philipp

Philipp L wrote:

 >I updated my software successfully, created a clean new project, and after some try&error, it really works now! - From Flash and everything!! yay :)

I'm experiencing similar problems.  I can't get the IPC messaging example to boot from flash. (same SW versions as you mentioned bios_6_33_05_46 IPC: ipc_1_24_03_32

Could you share some details what you did to make it work ?

How did you determine if the app runs ok ?  I added some blinking leds to it, to visually see that both cpus run and complete the test run.

Here is what I did, and say

I took the example, add some blinking leds to it. (GPIO).  (works fine in debugger)

  1 enabled bios C3: boot from flash  + boot C28:  

  2 eanbled bios C28: boot from flash

From the leds it appears that the system keeps rebooting. (some where after the MessageQ_open() call on the ARM. unsure where)

Hello,

I need to go on a business trip in just two hours, so I can't help you right now.

If your problem still exists next week, I can try to help you on Monday! Perhaps the "MCU SDK" for Concerto can solve your problem, too. I had no chance yet to try that, but it seems promising!

http://www.ti.com/product/f28m35h52c --> Technical Docs --> User Guides --> MCU SDK

But I can tell you one thing right now: "How did you determine if the app runs ok ?" --> Also with LEDs, and on the M3 side, I made a console output to UART/USB, and from there I could see on my console program on the PC when the M3 resetted. The c28 side was more difficult. When the IPC worked, I could use the same console (send message c28 > m3, then m3 > console > PC), but before, I also only had LEDs.

Thanks, I got it working.

This set of versions worked for me http://e2e.ti.com/support/embedded/bios/f/355/t/213533.aspx