DM365 & IPNC Appro : Capture driver problem in YUV420

Hi Reda,

Sorry, i could not uinderstand your description. Are you trying YUV420SP to YUV420SP resizing or are you trying to do something else? For any analysis, please share images of the YUV data that you got out from the resizer and also the input YUV data. At one place, you mention that you get black image which is your original data. Does this mean RSZ did not dump out anything?

reda38 said:

OUT START ADDR = 0x84dda000
OUT LINE  OFFSET = 640
OUT YUV420 COUT START ADDR = 0x84e25000
OUT YUV420 COUT LINE  OFFSET = 640
OUT BASE ADDR = 0x84dda000
OUT START LINE = 0
OUT END LINE = 480
OUT YUV420 COUT BASE ADDR = 0x84e25000
OUT YUV420 COUT START LINE = 0
OUT YUV420 COUT END LINE = 240

I dont see anything wrong in the above values. Do you find some issues with these?

Regards,

Anshuman

Hello Anshuman

>Sorry, i could not understand your description. Are you trying YUV420SP to YUV420SP resizing or are you trying to do something else?


I am trying VGA input in YUV422 to VGA output in YUV420SP. For the moment the resizing ratio is 1:1.
But I plan also to use resize ratio 1:4 for the reszB, once this rszA issue is fixed.


>For any analysis, please share images of the YUV data that you got out from the resizer
>and also the input YUV data.

OK. I will send them to you on monday. I have switched to color bars for debugging.

>At one place, you mention that you get black image which is your original data. Does this
>mean RSZ did not dump out anything?

Yes it looked like it is the case. I checked the code, and some significant enable flags were not set
in the drv_capture.c and drv_ipipe.c.
Maybe a minor bug due to a test case.

>I dont see anything wrong in the above values. Do you find some issues with these?

Neither do I. It was just to double check these values.

Regards,

Hi,

Understood your usecase. I assume you want ISIF-->IPIPEIF-->RSZ-->DDR data flow and you have already verified ISIF output to be good. Why are you concerned about drv_ipipe.c? I am attaching a few files which i think are enough for your usecase.

What do you mean by "significant enable flags" not being set?

Regards,

Anshuman

Hello Anshuman,

>Understood your usecase. I assume you want ISIF-->IPIPEIF-->RSZ-->DDR data flow and you have already verified ISIF output to be good.

yes, i verified isif to ddr, with drv_isif.c test module

ddr to ipipeif and then resizer with drv_resz.c test module

and now i want to validate all the chain with no temporary storage in ddr. This chain is run in the test module of drv_capture.c which calls drv_ipipe.c

>Why are you concerned about drv_ipipe.c?

drv_ipipe.c sets the parameters for the resizer and is called by drv_capture.c as said before.

>I am attaching a few files which i think are enough for your usecase.

Thank you i will check this on monday (not at work today).

>What do you mean by "significant enable flags" not being set?

the code i have was set so that the resz.setup.enable flags from configuration structures (from my memory since i do not have the files today) were set to FALSE instead of TRUE.

Thank you

Hello Anshuman,
I had a look at the files you sent and I noticed some points :

drv_ipipe.c:
-----------

1-in Ipipe_Open()
******************

"   for(rszId=DRV_IPIPE_RSZ_A; rszId<=DRV_IPIPE_RSZ_B; rszId++) {

    gDRV_ipipeObj.rszInfo[rszId].enable = config->rszOutConfig[rszId].enable; "


the resizer A or B looks enabled by its bit in a global structure.

But when you look in DRV_ipipeRszSetParams() which is called in DRV_ipipeSetParams() itself called in int DRV_ipipeOpen()
you see this:


"  gDRV_ipipeObj.rszSetup.enable         = TRUE;"

and

"  if(gDRV_ipipeObj.config.rszOutConfig[0].enable)
    gDRV_ipipeObj.rszSetup.rszAConfig     = &gDRV_ipipeObj.rszOutConfig[0];

  if(gDRV_ipipeObj.config.rszOutConfig[1].enable)
    gDRV_ipipeObj.rszSetup.rszBConfig     = &gDRV_ipipeObj.rszOutConfig[1];
"

Again it looked OK.
The problem comes after, from the coming loop which sets the hardware:

"  for(i=0; i<CSL_RSZ_CH_MAX; i++)
  {
    if(!gDRV_ipipeObj.config.rszOutConfig[i].enable)
      continue;

    pOutConfig = &gDRV_ipipeObj.rszOutConfig[i];
    ...
"

there is a copy to a temporary pOutConfig structure to change bits for each resizer.


and then later in the code I see:

"    pOutConfig->enable                 = FALSE;"

the enableflag is reset here !

alls flags/values are configured and then near the end we get :

"status = CSL_rszHwSetup(&gCSL_rszHndl, &gDRV_ipipeObj.rszSetup);"

Meaning that the enable bit will be written to FALSE into the hardware.

I do not remember ahving seen an enable bit somewhere else in calling procedure drv_capture.c


2-Second point:
***************

I see this in the code :

"#ifdef VIA_ISIF
  gDRV_ipipeObj.rszInConfig.inputDataPath      = CSL_RSZ_INPUT_DATA_PATH_IPIPEIF; // jbl
#else
  gDRV_ipipeObj.rszInConfig.inputDataPath      = CSL_RSZ_INPUT_DATA_PATH_IPIPE;
#endif
"

I had also to change it ,in the original code,  for my camera, but testing with the name of my camera instead of a define.


3-Third Point:
**************

in DRV_ipipeSetParams(), I see :

  "gDRV_ipipeObj.ipipeSetup.dataPath                   = CSL_IPIPE_IPIPE_DATA_PATH_BAYER_IN_YCBCR_RGB_OUT;"

   I have changed this to CSL_IPIPE_IPIPE_DATA_PATH_YCBCR_IN_YCBCR_RGB_OUT, no Bayer data is input indeed.

  
drv_isif.c:
-----------

1-First Point:
**************

I have also overwritten   sdrOutConfig.packMode with  CSL_CCDC_SDR_OUT_TYPE_8BITS_PER_PIXEL;

2-Second Point:
***************

Why in int DRV_isifSetParams()

have you changed

  sdrOutConfig.outWidth             = gDRV_isifObj.imgsModeInfo.validWidth;
  by
  sdrOutConfig.outWidth             = gDRV_isifObj.imgsModeInfo.validWidth*2; ?

Any way in the drv_capture test the data goes directly to ipipe fromisif, so sdram storage is not enabled.


For other parameters I will check on monday.

Thank you for files.

reda

Hi Reda,

Please go through the code flow of IPNC Ref Design.

reda38 said:

alls flags/values are configured and then near the end we get :

"status = CSL_rszHwSetup(&gCSL_rszHndl, &gDRV_ipipeObj.rszSetup);"

Meaning that the enable bit will be written to FALSE into the hardware.

I do not remember ahving seen an enable bit somewhere else in calling procedure drv_capture.c

If you refer to videoCaptureThr.c, you can see that the capture thread calls DRV_captureStart() after all initializations are done. During this time the driver goes and starts all the modules that come in data flow, including IPIPE and RSZ. There is no problem in the code for enabling these key bits. If they were not enabled, the whole code would have broken.

reda38 said:

I see this in the code :

"#ifdef VIA_ISIF
  gDRV_ipipeObj.rszInConfig.inputDataPath      = CSL_RSZ_INPUT_DATA_PATH_IPIPEIF; // jbl
#else
  gDRV_ipipeObj.rszInConfig.inputDataPath      = CSL_RSZ_INPUT_DATA_PATH_IPIPE;
#endif
"

I had also to change it ,in the original code,  for my camera, but testing with the name of my camera instead of a define.

Your changes are good.

reda38 said:

3-Third Point:
**************

in DRV_ipipeSetParams(), I see :

  "gDRV_ipipeObj.ipipeSetup.dataPath                   = CSL_IPIPE_IPIPE_DATA_PATH_BAYER_IN_YCBCR_RGB_OUT;"

   I have changed this to CSL_IPIPE_IPIPE_DATA_PATH_YCBCR_IN_YCBCR_RGB_OUT, no Bayer data is input indeed.

Again, you changes are correct. Our code was mostly enhanced for ISIF-->DDR-->RSZ-->DDR mode of data flow.

reda38 said:

2-Second Point:
***************

Why in int DRV_isifSetParams()

have you changed

  sdrOutConfig.outWidth             = gDRV_isifObj.imgsModeInfo.validWidth;
  by
  sdrOutConfig.outWidth             = gDRV_isifObj.imgsModeInfo.validWidth*2; ?

Any way in the drv_capture test the data goes directly to ipipe fromisif, so sdram storage is not enabled.

 As i mentioned earlier, we used DDR as our intermediate buffer, so the above chnages were needed. You need not do these changes for your usage.

Regards,

Anshuman

PS: Please mark this post as verified if you think it has answered your question. Thanks