RTOS/TMS320C6657: EDMA3_DRV_create() fails

Pablo Stickar

Part Number: TMS320C6657

Tool/software: TI-RTOS

Hello everyone

I'm still having problems bringing EDMA into life. My code looks like:

static unsigned int edma3InstanceId = 0, edma3RegionId = 1;
extern EDMA3_DRV_GblConfigParams sampleEdma3GblCfgParams[];                              // Also tried: extern EDMA3_DRV_GblConfigParams sampleEdma3GblCfgParams;
extern EDMA3_DRV_InstanceInitConfig sampleInstInitConfig[][EDMA3_MAX_REGIONS]; // also tried: extern EDMA3_DRV_InstanceInitConfig sampleInstInitConfig;

// And in the same source file:

void edma_config(EDMA_T *EdmaObj)
{
  EDMA3_DRV_Result edma3Result = EDMA3_DRV_SOK;
  // Set global config to defaults in sample Params (located in bios_edma3_drv_sample_C6455_cfg.c)
  EDMA3_DRV_GblConfigParams *globalConfig = &sampleEdma3GblCfgParams; // + edma3InstanceId;
  // Used in DRV_create() function to specify master/slave
  EDMA3_DRV_MiscParam miscParam;
  // Used for EDMA3_DRV_open() API and initCfg structure, defined in <edma3_drv.h>
  EDMA3_DRV_InitConfig initCfg;
  // Located in bios_edma3_drv_sample_C6455_cfg.c
  EDMA3_DRV_InstanceInitConfig *instanceConfig = &sampleInstInitConfig; // [edma3InstanceId] + edma3RegionId;
  // Declare Config structure used to initialize the Driver Instance (needed by DRV_open)
  initCfg.isMaster = TRUE;
  initCfg.regionId = edma3RegionId;	        // User must specify region (LLD/DRV does not support "global" region)
  initCfg.drvSemHandle = (void*)&os_sem; 	// Os_sem create in semaphore.c
  initCfg.drvInstInitConfig = instanceConfig;   // Device-specific configuration - resources owned by region edma3regionId
  initCfg.gblerrCb = NULL; 			// Callback function to catch channel controller errors such as TCC error,
                                                // queue threshold exceeded, etc.
  initCfg.gblerrData = NULL; 			// any data required by the error callback function
  miscParam.isSlave = FALSE;                    // ARM + DSP, specify who is the master, single-chip = use FALSE

  // EDMA3 Driver Create (called only once for each EDMA3 hardware instance)
  edma3Result = EDMA3_DRV_create( edma3InstanceId, globalConfig, (void *)&miscParam );
  if( edma3Result != EDMA3_DRV_SOK )
  {
    Log_info0( "e@edma_config:EDMA3_DRV_create()" );
    return;
  }
}

The return value is not EDMA3_DRV_SOK . the context ma matter, so here goes a zip of the whole project (CCS7.3, pdk_2_0_7, edma3_lld_2_12_05_29):

gpio_edma1.zip

I've made same clean up in main.c, so only the edma3 functionality remains, but some more source files are there.

If anyone has a tip, it will be very helpful to me.

Kind regards,

Pablo

over 6 years ago

0 Cvetolin Shulev-XID over 6 years ago

TI__Guru 65405 points

Hi Pablo,

I've forwarded your query to the software experts. Their feedback should be posted here.

BR
Tsvetolin Shulev

0 Pablo Stickar over 6 years ago in reply to Cvetolin Shulev-XID

Intellectual 350 points

Hello

the reason for the problem is actually very silly: I coded this function call in a loop, and the corresponding "release" is outside.

I hope nobody spent too much time on this. Kind regards,

0 lding over 6 years ago in reply to Cvetolin Shulev-XID

TI__Guru* 95265 points

Hi,

For the C6657 with PDK 2.0.7, there is an example for MCBSP. In sample_init.c, it calls edma3init()----->EDMA3_DRV_create(). You can just use that code to see if you can create an EDMA instance (it should), then compare what differs in your code.

Regards, Eric

0 Pablo Stickar over 6 years ago in reply to lding

Intellectual 350 points

Thank you Eric for pointing me to that example. I find many answers there. And also a question:

That example contains a full definition of the EDMA3_RM_GblConfigParams[] array. Is it a definition tuned up for that particular example (MCBSP)? Or is it about C6657 properties that are common to any application using the EDMA3-LLD? In my previous code, I declared that array extern, and I though that the real definition came inside one of the edma-lld libraries I linked against (the linker gave no error). May be more than one definition is possible, but then I'm surprised that they have the same name. Do you have at hand a similar definition valid for the C6678 ?

Kind regards,

Pablo

0 lding over 6 years ago in reply to Pablo Stickar

TI__Guru* 95265 points

Hi,

The EDMA3_RM_GblConfigParams[] is a global array. In the MCBSP sample, it is defined in mcbsp\example\c6657\MCBSPDigLpbk\sample_c6657_cfg.c. The same is defined in EDMA LLD \ti\sdo\edma3\drv\sample\src\platforms\sample_c6657_cfg.c (the contents are exact the same). When both are defined in the application, the one in source code will be linked instead the other in the library. I didn't see the reason why need a separate sample_c6657_cfg.c in the example.

For the same definition for C6678, you can find it under EDMA LLD: packages\ti\sdo\edma3\drv\sample\src\platforms

Regards, Eric

0 Pablo Stickar over 6 years ago in reply to lding

Intellectual 350 points

Hello

thank you! I like it that I'm finally understanding some details.

I hope I can still ask one more little question here: I would like to understand what the following two arrays represent (from sample_c6657_int_reg.c, lines 67-73):

ccXferHostInt[NUM_EDMA3_INSTANCES][NUM_DSPS] = {{0u, 20u}};
edma3ErrHostInt[NUM_EDMA3_INSTANCES][NUM_DSPS]= {{1u, 21u}};

The other arrays appear to be taken from the CIC0-Input-Event table. Are they hand-picked events of the CIC0/1 tables?

Thanks and regards,

0 lding over 6 years ago in reply to Pablo Stickar

TI__Guru* 95265 points

Hi,

These are host interrupt, as you saw they are used in function CpIntc_enableHostInt(). So they are hand-picked from C6657 primary corepac interrupt table.

Regards, Eric

0 Pablo Stickar over 6 years ago in reply to lding

Intellectual 350 points

Hello

there are a couple of things that I cannot understand in the example. Do please tell me if I should open separate threads for these questions.

[1]

The region number. I've read in the tutorials that region 0 (zero) is for the supervisor, and that it is not supported by the LLD. So I used region 1. But the example uses the core number as the region. Hence the first core is a supervisor? The example definitely makes use of the LLD, and the first few lines in main() basically state that the core number is zero.

[2]

The global structures EDMA3_RM_GblConfigParams and EDMA3_RM_InstanceInitConfig. We've already discussed that the example provides a definition of them. In my opinion, the definition provided by the library is constant and a user-program cannot modifiy it. Or can it? The example modifies the contents of both structures (source mcbspMasterDigLpbk.c, lines 946-957). It looks like the dma channels should be enabled by hand. I understand that those lines assume both edma-instance zero and region zero, but I do not understand the subscript "1" there:

uint32_t temp;

// ...

/* Enabling the dmaChannel MCBSP Events to EDMA (36-40) */
temp=sampleEdma3GblCfgParams[0].dmaChannelHwEvtMap[1];
sampleEdma3GblCfgParams[0].dmaChannelHwEvtMap[1]= temp | 0x000000F0;

/* Let RM Own the dmaChannels (36-40) */
temp=sampleInstInitConfig[0][0].ownDmaChannels[1];
sampleInstInitConfig[0][0].ownDmaChannels[1]=temp | 0x000000F0;

/* Let RM Own the TCCs (36-40) */
temp=sampleInstInitConfig[0][0].ownTccs[1];
sampleInstInitConfig[0][0].ownTccs[1]= temp | 0x000000F0;

/* Initialize EDMA3 library */
hEdma[0] = edma3init(0, &edmaResult);

I wonder if channels 36-40 are something special, or just a selection as anyone else (I assume the latter is right).

Thanks and regards,

Pablo

0 Pablo Stickar over 6 years ago in reply to Pablo Stickar

Intellectual 350 points

Ohhh, nooo.
In the example provided with the pdk_c665x_2_0_7, there are two variables with the same name (sampleEdma3GblCfgParams) but different type. One is EDMA3_DRV_GblConfigParams and the other one EDMA3_RM_GblConfigParams.
Same story with the instance config (sampleInstInitConfig).
Needless to say, I fail to understand this trick.
Regards,
p.

Processors

Processors forum

RTOS/TMS320C6657: EDMA3_DRV_create() fails