OpenMP QMSS manual setup

Hi, everyone,

I'm using TMS320C6670 DSP and trying to implement OpenMP in my project. But I need to do a manual QMSS initialization since it's used for other peripherals (Ethernet, FFTC). So I took OpenMP 2_01_16_03 Hello_world example and modify it a bit:

#define PAD     (64)
#define PNUM    (16384)
#define OPAD    (8)
#define TRNUM   (4)

#pragma DATA_SECTION(iqs, ".inputbuff");
#pragma DATA_SECTION(data, ".inputbuff");
#pragma DATA_ALIGN(iqs, 128);
#pragma DATA_ALIGN(data, 128);
#pragma DATA_ALIGN(win, 128);
float    data[2*(PNUM+PAD)];
float   iqs[2*PNUM];
float   win[PNUM];

inline static float * restrict windowing(
    int         n          // number of sampled data
)
{
    short i;
    float ampl = (float)(1./(1<<0));

    for (i=0; i < n; i++)
    win[i] = (float) ampl;

    return win;
}
int main (int argc, char *argv[]) {

  int i, n_treds,c_l_n,cl_pc,num_p_pc,iq_pairs;
  double st_o=0.0,fn_o=0.0;
  float *win;

  win = windowing(PNUM);

  omp_set_num_threads(TRNUM);

  iq_pairs=64/sizeof(float); // number of IQ pairs in one cache line (128 bytes)
  c_l_n= (PNUM+iq_pairs-1)/iq_pairs;// cache lines number
  cl_pc= (c_l_n +3)/4; // cache lines per core number.
  num_p_pc = cl_pc*iq_pairs; // number of IQ pairs per core

  st_o = omp_get_wtime();
#pragma omp parallel shared (data,iqs,win) private (i)
  {
	  int id;
      float * restrict pout;
      const float * restrict pin,* restrict pw;
	  id=omp_get_thread_num();

      pout=&data[PAD+id*num_p_pc];
      pin=&iqs[id*num_p_pc];
      pw=&win[id*num_p_pc];
#pragma omp for
      for (i=0;i<num_p_pc/2;i++)
      {
    	  pout[2*i]=pin[2*i]*pw[i];
    	  pout[2*i+1]=pin[2*i+1]*pw[i];
      }
  }
  fn_o = omp_get_wtime();
  printf("Calculation time = %f mus\n", (fn_o-st_o)*1000000);
}

If i using standard example platform and config file with this code everything calculating just fine, ~4 times faster than sequential code.

But I need to specify additional regions for other peripherals, so I made some changes to the config file (as they given in the OpenMP user guide):

    var ompSettings = xdc.useModule("ti.runtime.openmp.Settings");
    ompSettings.runtimeInitializesQmss  =  false;
    var OpenMP = xdc.useModule('ti.runtime.ompbios.OpenMP');    
    OpenMP.qmssMemRegionIndex = 1; // region 0 is occupied by Ethernet
    OpenMP.qmssFirstDescIdxInLinkingRam = 32; // 0 region is 32 descriptor long


// after shared region and heapOM config added that lines

// __TI_omp_start_rtsc_mode configures the runtime and calls main
    var Startup = xdc.useModule('xdc.runtime.Startup');
    
    Startup.lastFxns.$add('&qmssInitOmp');
    
    Startup.lastFxns.$add('&__TI_omp_initialize_rtsc_mode');

Where the QMSS initialization function look like:

int qmssInitOmp (void)
{
    int               result = 0;
    Qmss_MemRegInfo     memCfg;
    Qmss_InitCfg        qmssInitConfig;

    // Set up QMSS configuration
    if (DNUM==0)
    {

    memset (&qmssInitConfig, 0, sizeof (Qmss_InitCfg));
    // Use internal linking RAM
    qmssInitConfig.linkingRAM0Base  =  0;
    qmssInitConfig.linkingRAM0Size  =  0;
    qmssInitConfig.linkingRAM1Base  =  0x0;
    qmssInitConfig.maxDescNum       =  NUM_HOST_DESC+256;

    qmssInitConfig.pdspFirmware[0].pdspId = Qmss_PdspId_PDSP1;
    qmssInitConfig.pdspFirmware[0].firmware = (void *) &acc48_le;
    qmssInitConfig.pdspFirmware[0].size = sizeof (acc48_le);

    // Initialize the Queue Manager
    result = Qmss_init (&qmssInitConfig, qmssGblCfgParams);
    if(result != QMSS_SOK) {
        #if DEBUG_ERRORS
        System_printf("Error initializing Queue Manager SubSystem, Error code : %d\n", result);
        #endif // DEBUG_ERRORS
        return -1;
    }
    // Initialize and setup CPSW Host Descriptors
    memset (gHostDesc, 0, SIZE_HOST_DESC * NUM_HOST_DESC);
    memCfg.descBase         =  (unsigned int *) Convert_CoreLocal2GlobalAddr ((unsigned int) gHostDesc);
    memCfg.descSize         =  SIZE_HOST_DESC;
    memCfg.descNum          =  NUM_HOST_DESC;
    memCfg.manageDescFlag   =  Qmss_ManageDesc_MANAGE_DESCRIPTOR;
    memCfg.memRegion        =  Qmss_MemRegion_MEMORY_REGION0;
    memCfg.startIndex       =  0;

    // Insert Host Descriptor memory region
    result = Qmss_insertMemoryRegion(&memCfg);
    if(result == QMSS_MEMREGION_ALREADY_INITIALIZED) {
        #if DEBUG_ERRORS
        System_printf("Memory Region %d already Initialized \n", memCfg.memRegion);
        #endif // DEBUG_ERRORS
    } else if(result < QMSS_SOK) {
        #if DEBUG_ERRORS
        System_printf("Error: Inserting memory region for Eth %d, Error code : %d\n", memCfg.memRegion, result);
        #endif // DEBUG_ERRORS
        return -1;
    }

    Qmss_start();
    }
    return 0;
}

So when I made all includes and so on for the QMSS the project builds without errors, but during run time, then the core 0 enters #pragma omp zone it just hangs at the lines 228, 229 of the tomp_util.h file :

        while (mysense != barrier->sense)
            tomp_completePendingTasks();

I think I made some mistake with QMSS init but I can't find it by m own. Thank you in adwance.

Best Regards,

Pavlo!