[Custom IQ data with MIDAS 2.0]: DSP MMU Error Fault! MMU_IRQSTATUS = [0x1]. Virtual DSP addr reference that generated the interrupt = [0xfe00fe00]

Hi Honey

Your understanding looks correct and lower scanline or sample sizes, as long as they meet module requirements (multiplicity, max. size), should work if the higher scanline/sample sizes work. 

One thing I noticed from this log, is that the UNIVERSAL_process call that is failing seems to be related to the DPU call. Also I notice before the MMU error, there is another error:

"Wallfilter file = ./userdata/LUTs/TI_Carotid_WallFilterCoeff_001.bin
Error in reading DPU Wall Coeff file ./userdata/LUTs/TI_Carotid_WallFilterCoeff_001.bin"

It doesn't seem to find the wall filter coefficient file. It could be that the files doesn't exist or have something to do with the permissions on that .bin file. Have a look and ensure that the wall filter coeff file can be ready successfully. Also remember that if you are changing the number of ensembles, you will need to generate a wall filter coefficient corresponding to that. 

If the above doesn't work, another check you can do quickly to help us get to the root of this, is to modify the loadmodules.sh, mud.tcf, etc. to be exact for the 384x640 use-case. Then also try the same for the 112x1024 use-case and rerun. See if those work. 

Hi Uday,

Thank you for the suggestions. But unfortunately, all of those modifications didn't help.

I am pasting the log files, loadmodules.sh and mud.tcf for the 380x640 use-case below and 112x1024 use-case in next post. Please analyse it. I need to solve it as soon as possible :(

1. For 384x640 use-case, generated the wall filter coefficient file, the following is the log:-

============ Texas Instruments Inc ==============

Edited

Thanks,

Honey S

For the 112x1024 use-case, the following are the log files:-

Log deleted

Thanks,

Honey S

It looks like the log in the case where you are using the loadmodules.sh, etc. to be exact as the dataset has no errors, so let's first make sure that this works correctly.

1. How many input frames of custom input data are you providing? 

2. When you say the output is hanging, do you mean you see only 1 output frame being continuously displayed and no running video? 

3. If yes, does the output frame look as expected? If no, do you see anything else on the screen in this case or is it always just blank?

In either case, the way to debug further, is to write the contents of the input frame, output frame after BPU, output frame after DPU and output frame after SCU to different files, so you can analyze these frames to see if they look correct. To do this, you can use something like the 'logBuffer' function below and write the contents to a binary file, and then analyze it in MATLAB. Let me know if these look as expected.

/******************************************************************************
 * logBuffer
 * Logs the contents of a buffer in a binary file
 ******************************************************************************/
int logBuffer( unsigned char *pBuffer, int bufferSize, const char * filename, const char * mode)
{
	FILE *file;

	file = fopen(filename, mode);

	if (file!=NULL) {   
		fwrite(pBuffer, 1, bufferSize, file);
		printf("Logging %s \n", filename);
	}
	else {
		printf("Creating BPUIn Frame failed");
		return -1;
	}

	fclose(file);

	return 0;

}

Hi Uday,

Please see my reply to your queries. The following are tested for Bmode data 384x640 and Color mode data 256x512x12 ensembles.

1. How many input frames of custom input data are you providing? 

>> Bmode data - 14 frames, Color data - tested with 10 frames and 5 frames.

2.  When you say the output is hanging, do you mean you see only 1 output frame being continuously displayed and no running video? 

>> No frame displayed. The screen just freeze when I press "Start" on MIDAS GUI.

3. If yes, does the output frame look as expected? If no, do you see anything else on the screen in this case or is it always just blank?

>> The screen just freeze when I press "Start" on MIDAS GUI showing only MIDAS DEMO Menu.

4.  In either case, the way to debug further, is to write the contents of the input frame, output frame after BPU, output frame after DPU and output frame after SCU to different files, so you can analyze these frames to see if they look correct. To do this, you can use something like the 'logBuffer' function below and write the contents to a binary file, and then analyze it in MATLAB. Let me know if these look as expected.

>> When I enabled logging to buffers, I got only 3 buffers - bpuout, dpuout and scubout. bpuout and scubout seems to be okay. But dpuout seems to be incorrect.

FYI:- I could successfully run Linear Bmode data 384x640 and Color mode data 128x416x12 ensembles data. Also I could successfully run Bmode data 384x736 and color mode data 128x416x12 ensembles. These all worked with modified memory map targeted for Bmode data 512x1024 and color data 256x512x12 ensembles.

But still issues with Bmode data 384x640, Color data 256x512x12 ensembles and Bmode data 112x1024 and Color mode data 256x512x12 ensembles.

By the way, I am now starting to test with Sector data. I will update with the observations while running those too.. 

I need to fix the issues very soon. :(

Thanks,

Honey S

Hi Uday,

I have finally resolved the issue "DSP MMU error fault". On further debugging, I found that my issue is isolated to the scu unit. So I decided to rebuild the scu dsp code.

I didn't do any modification to source code except the one which I mentioned earlier for the sector data where the logic for checking sector compliance is wrong. Then I recompiled "scu" in code composer studio and rebuilt the midas application with the new scu package. 

Thus now my linear and sector data work with no DSP MMU error fault. I assume some instability in the older scu package. 

Thanks for your support. 

Thanks,

Honey S

Hi Honey

Thanks for the update. Glad your custom datasets are working with MIDAS 2.0 now. Could you please mention the code change you had to make for the 'sector compliance' test in SCU.