Error Message when calling HeapMemMP_alloc()

Hi Dominic,

Which version of IPC are you using? Would I be able to rebuild your test case to reproduce the error? If so could you upload/e-mail me the test case (.c and .cfg files).  You can use the ROV tool (Tools -> ROV) and navigate to the view for the ti.sdo.ipc.heaps.HeapMemMP module to find out how much free memory the heap contains.  Using this tool could allow you to determine whether this error is only occurring when the heap runs out of memory.

Regards,

Shreyas

Hi Shreyas,

I am using IPC version 1.20.00.23, and unfortunately I don't have a test case that I can email to you, but I can walk you through of what I did.

In the configuration file, I have specified a shared memory region using DDR2 as the following in bold with starting address of  0xe0000000 and the size of 0x0C000000.

/* Shared Memory base address and length */
var SHAREDMEMDDR2           = 0xe0000000;
var SHAREDMEMSIZEDDR2       = 0x0C000000;
 
var SharedRegionDDR2 = xdc.useModule('ti.sdo.ipc.SharedRegion');
SharedRegionDDR2.numEntries = 4;
SharedRegionDDR2.translate = true;   
SharedRegionDDR2.setEntryMeta(1,
    { base: SHAREDMEMDDR2,
      len:  SHAREDMEMSIZEDDR2,
      ownerProcId: 0,
      isValid: true,
      cacheLineSize: 128,
      createHeap: true,
      name: "DDR2_MEM",
    });   

In Core 0 I have created the memory handler as the following in bold :

#define DDR2_STARTING_ADDR          0xE0000000
#define MAX_DDR2_SHARED_BUF_SIZE   0x0C000000
#define DDR2_HEAP_NAME   "DDR2HeapBuf"

    HeapMemMP_Params              heapMemParams;
    HeapMemMP_Handle              gHeapMemDDR2Handle = NULL;
       
    // Create a heap in DDR2
    HeapMemMP_Params_init(&heapMemParams);
    heapMemParams.regionId       = 1;
    heapMemParams.name           = DDR2_HEAP_NAME;
    heapMemParams.sharedAddr     = (void*)DDR2_STARTING_ADDR;
    heapMemParams.sharedBufSize  = MAX_DDR2_SHARED_BUF_SIZE;

    gHeapMemDDR2Handle = HeapMemMP_create(&heapMemParams);

In Core 1,2,3,4,5. I have been able to open the heap successfully using the #define DDR2_HEAP_NAME

#define DDR2_HEAP_NAME   "DDR2HeapBuf"

      do
        {
            heapStatus = HeapMemMP_open(DDR2_HEAP_NAME, &gHeapMemDDR2Handle);
        } while( HeapMemMP_S_SUCCESS != heapStatus );    

However, I didn't have much luck allocating different memory in core 1,2,3,4,5 using the provided API. If I call this API for a few times in every core, it seems to work fine. However, if I were to call it more than 8 times or more, it seems to pass on 1 core but fail on the other.

int bufSize;

int align;

void* pBuf = NULL;

*pBuf = HeapMemMP_alloc(&gHeapMemDDR2Handle, bufSize, align);

Thanks much,

Dominic

Dominic,

I didn't find any problems with your code.  However, I have a couple of more questions:

1) What value(s) of 'align' & 'bufSize' are you passing to HeapMemMP_alloc?

2) If you have ROV running, could you open the ROV view for ti.sdo.ipc.heaps.HeapMemMP and post a screenshot of the 'Detailed' and 'FreeList' views after you get the Assert?  This would help me determine the integrity of the heap at that point.

Thanks,

Shreyas

Hi Shreyas,

Since Dominic is taking days off for a couple weeks, I'll continue to discuss with your on this issue to get things going.

I haven't got a chance to use the ROV tool yet, but I can provide answer for your question 1:

We are calling HeapMemMP_alloc() multiple times (around 20 times, is there a limit for it?) at system initialization, as Dominic mentioned before, with bufSize set to different sizes (ranging from 75KB to 1200B), and align always set to sizeof(long long).

Do you see any problem with them?

Once I get a chance to try ROV, I will provide your the screen shot.

Thanks,

-- Louis

Hi Louis,

I'll try to reproduce the problem given the information that both of you have provided so far.  Please provide the ROV screenshot if you get it running.

Regards,

Shreyas