This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

Linux/DRA746: 1080 video with @60fps are playing in slow motion.

Part Number: DRA746

Tool/software: Linux

Hello Team, 

We use custom h/w and software based on DRA746 and VSDK 3.04. 

We are facing one issue with HD video playback if video files framerate is 60fps, then if play in slow motion, if we convert that video from 60fps to 30fps then that file plays properly.

As per your previous inputs TI supports one IVA which can decode 1x1080@60 or 2x1080p@30.

In our full loaded system(navigation is running in parallel) when we play HD video @60 fps then playback is not smooth. But if we play just video playback by removing all other loads then it plays properly(through command line), but in between flicking is there.

Sharing you the original video and observation video through portal.

Regards,

Ikshwaku

  • Hi Ikshwaku,
    Can you point me to the duration where it is not smooth in the observation video?
    I decoded this with gst-play-1.0 and it is able to decode in 2.02 minutes which is same as the duration of the stream.

    Thanks
    Ram
  • Hello Ram,

    Observation video is for flickering we are observing, if video played without load(HMI, navigation). That you can observer at 2, 4 and 8 sec in observation video.
    Sharing another observation video where video playback is not smooth.

    Regards,
    Ikshwaku
  • Hi Ikshwaku,
    Can you check if you can execute this and get data?
    processors.wiki.ti.com/index.php

    Please also check if you can execute this to visually get the data.
    processors.wiki.ti.com/index.php

    Thanks
    Ram
  • Hi Ikshwaku,
    Please try to get the L3 bandwidth details from the statcollector and share with us.

    Please go through these two documents on tuning L3 bandwidth with BW regulation and limiting.

    www.ti.com/.../sprabx1a.pdf
    www.ti.com/.../sprabx0.pdf

    First try this option and check if there is any improvement. This configures IVAHD BW regulation at 1.5Gbps

    target$omapconf dump 0x44805000 0x44805014
    target$omapconf write 0x44805008 0xB5
    target$omapconf write 0x4480500C 0x5DC
    target$omapconf dump 0x44805000 0x44805014---> Che ck if the registers are updated

    Next option is to limit Bandwidth of VPE and GC320. I am not sure if any of the usecase is involving GC320 hardware.
    Can you confirm if GC320 hardware is used for any of the usecase?

    Thanks
    Ram
  • Hello Ram,


    root@mmt2020-a880:~/glsdkstatcoll# glsdkstatcoll -f config.ini
    -sh: glsdkstatcoll: command not found
    root@mmt2020-a880:~/glsdkstatcoll# ./glsdkstatcoll -f config.ini
    NOTE: BANDWIDTH is not enabled, ignoring DELAY
    NOTE: STATCOLL is not enabled, ignoring DELAY
    NOTE: BANDWIDTH is not enabled, ignoring EMIF_PERF_CFG1
    NOTE: STATCOLL is not enabled, ignoring EMIF_PERF_CFG1
    NOTE: BANDWIDTH is not enabled, ignoring EMIF_PERF_CFG2
    NOTE: STATCOLL is not enabled, ignoring EMIF_PERF_CFG2
    NOTE: BANDWIDTH is not enabled, ignoring TOTAL_TIME
    NOTE: BANDWIDTH is not enabled, ignoring INTERVAL_US


    COMPLETED: Parsing of the user specified parameters..

    Configuring device now..

    STATISTICS COLLECTOR option chosen
    ------------------------------------------------

    ------------------------------------------------
    Compile time = Jan 31 2019 10:53:10
    ------------------------------------------------

    ERROR: Unknown initiator

    Any idea for this error?

    Regards,
    Ikshwaku
  • Hi Ikswaku,
    Looks like you have not copied initiator.cfg
    Can you share your config.ini and initiator.cfg?

    Thanks
    Ram
  • I have copied these files the working dir.
    Sharing through portal.

    Regards,
    Ikshwaku
  • Hi Ikshwaku,

    Can you try with attached initiator.cfg? I have removed the initiator BB2D and VIP which are are not required for your usecase and added VPE.

    Please try again with this and let me know.

    It will generate statcollector.csv . I will get the data for same stream on EVM.

    initiators.zip

    Thanks

    Ram

  • Hi Ikshwaku,
    In config.ini change TOTAL_TIME to 120 seconds since the stream Lamberghini is 2 minutes.
  • Sharing the generated .csv file.

    Regards,

    Ikshwaku

  • Hi Ikshwaku,

    Please generate this data  a) when only this specific steam is played and b) video playback + navigation is running. Create separate csv files for these two cases.

    This will help to differentiate the components used.

    Thanks

    Ram

  • Hello Ram,

    Shared the .csv files for follwoing cases:

    1. non working video + navigation
    2. only non working video
    3. only working video

    Issue is seen with both 1 and 2 .

    Regards,
    Ikshwaku
  • One more point to add, executed above usecases through HMI.

    Regards,
    Ikshwaku
  • Hi Ikshwaku,
    The data looks not correct.
    Can you please try to generate the data by following this user guide

    processors.wiki.ti.com/index.php

    I collected statistics while executing Lamerghini stream . I am getting
    -------------------------------------------------------------------
    Initiator Average Peak Average(active)
    -------------------------------------------------------------------
    STATCOL_EMIF1_SYS 1341.0 1759.0 1341.0 (3998)
    STATCOL_EMIF2_SYS 1299.0 1693.0 1299.0 (3998)
    STATCOL_IVA 1242.0 2016.0 1248.0 (3977)
    STATCOL_DSS 499.0 569.0 498.0 (3998)
    STATCOL_GPU_P1 249.0 552.0 249.0 (3998)
    STATCOL_GPU_P2 263.0 582.0 263.0 (3998)
    STATCOL_MA_MPU_P1 8.0 32.0 8.0 (3998)
    STATCOL_MA_MPU_P2 14.0 64.0 14.0 (3998)
    STATCOL_VPE_P1 185.0 310.0 189.0 (3914)
    STATCOL_VPE_P2 188.0 317.0 192.0 (3914)
    -------------------------------------------------------------------
    -------------------------------------------------------------------

    and for a 1920x1080 @ 30fps stream, I get
    -------------------------------------------------------------------
    Initiator Average Peak Average(active)
    -------------------------------------------------------------------
    STATCOL_EMIF1_SYS 865.0 1578.0 865.0 (3998)
    STATCOL_EMIF2_SYS 847.0 1527.0 847.0 (3998)
    STATCOL_IVA 506.0 1629.0 796.0 (2543)
    STATCOL_DSS 499.0 651.0 499.0 (3998)
    STATCOL_GPU_P1 248.0 552.0 248.0 (3998)
    STATCOL_GPU_P2 262.0 582.0 262.0 (3998)
    STATCOL_MA_MPU_P1 15.0 69.0 15.0 (3998)
    STATCOL_MA_MPU_P2 33.0 167.0 33.0 (3998)
    STATCOL_VPE_P1 92.0 283.0 150.0 (2446)
    STATCOL_VPE_P2 94.0 294.0 154.0 (2446)
    -------------------------------------------------------------------
    -------------------------------------------------------------------


    This shows that IVAHD's DDR access is doubled due to 60fps decoding and similarly for VPE.
    Can you try this with gst-play-1.0 instead of HMI?

    Did you try to experiment with the omapconf setting I mentioned earlier?

    Thanks
    Ram
  • Hi Ikshwaku,

    Can you check if you can execute this application on your system?

    This is a desktop-shell based wayland client application to get DDR Bandwidth information graphically.

    soc-ddr-bw-visualizer.zip

    Thanks

    Ram

  • Hello Ram,


    For Non working Video:

    -------------------------------------------------------------------
    Initiator Average Peak Average(active)
    -------------------------------------------------------------------
    STATCOL_EMIF1_SYS 111.0 384.0 111.0 (3998)
    STATCOL_EMIF2_SYS 108.0 381.0 108.0 (3998)
    STATCOL_IVA 68.0 408.0 69.0 (3936)
    STATCOL_DSS 21.0 112.0 44.0 (1930)
    STATCOL_GPU_P1 2.0 197.0 2.0 (3867)
    STATCOL_GPU_P2 1.0 201.0 3.0 (2010)
    STATCOL_MA_MPU_P1 66.0 256.0 66.0 (3998)
    STATCOL_MA_MPU_P2 43.0 171.0 43.0 (3998)
    STATCOL_VPE_P1 0.0 61.0 0.0 (1450)
    STATCOL_VPE_P2 1.0 63.0 3.0 (1972)
    -------------------------------------------------------------------
    -------------------------------------------------------------------

    For 1920x1080 @24fps:

    -------------------------------------------------------------------
    Initiator Average Peak Average(active)
    -------------------------------------------------------------------
    STATCOL_EMIF1_SYS 68.0 415.0 68.0 (7764)
    STATCOL_EMIF2_SYS 70.0 428.0 70.0 (7764)
    STATCOL_IVA 18.0 217.0 31.0 (4508)
    STATCOL_DSS 22.0 125.0 43.0 (3950)
    STATCOL_GPU_P1 12.0 263.0 12.0 (7562)
    STATCOL_GPU_P2 12.0 264.0 26.0 (3762)
    STATCOL_MA_MPU_P1 43.0 261.0 43.0 (7764)
    STATCOL_MA_MPU_P2 24.0 190.0 24.0 (7762)
    STATCOL_VPE_P1 0.0 43.0 0.0 (1719)
    STATCOL_VPE_P2 1.0 48.0 3.0 (3568)
    -------------------------------------------------------------------
    -------------------------------------------------------------------


    Did you try to experiment with the omapconf setting I mentioned earlier?
    --> will check it today.

    Regards,
    Ikshwaku
  • Hi Ikshwaku,
    I tested glsdkstatcollector and soc-ddr-bw-visualizer on M4 based display server based SDK. Both are not showing correct information.
    These tools are developed for PSDKLA and it shows correct L3 BW result on PSDKLA .

    I will continue to get the data on PSDKLA3.04 SDK for this 60fps stream.

    Thanks
    Ramprasad
  • Hello Ram,

    Should I try the omapconf setting ??

    Regards,
    Ikshwaku
  • Hi Ikshwkau,
    Please start with omapconf experiments for
    a) video playback alone
    b) video playback + navigation usecase

    Is the issue happening for only this specific 60fps stream? Are you not observing issue for any other 1080p@60fps stream?

    Thanks
    Ram
  • Hello Ram,

    We have not tried for any other 1080p@60fps, but if we convert this video to 30fps then issue is not observed.

    Regards,
    Ikshwaku
  • Hello Ram,

    I have tried following settings, but no improvement:

    target$omapconf dump 0x44805000 0x44805014
    target$omapconf write 0x44805008 0xB5
    target$omapconf write 0x4480500C 0x5DC
    target$omapconf dump 0x44805000 0x44805014---> Che ck if the registers are updated

    Regards,
    Ikshwaku
  • Hi Ikshwaku,
    1) Did you execute these commands just before launching video playback of 60fps stream?
    2) Have you taken care of the patches from Subhajit + Prasad on vision-sdk?

    I will share the next set of experiments

    Thanks
  • Hi Ikshwaku,
    Vision-sdk has option to print the performance statistics when a usecase is running with "p" option.
    Can you check if you can get these details from vision-sdk? These statistics seems matching with glsdkstatcollect on PSDKLA

    Thanks
    Ram
  • Hello Ram,

    1) Did you execute these commands just before launching video playback of 60fps stream?
    --> Yes I have executed the playback just after updating the omapconf commands.


    2) Have you taken care of the patches from Subhajit + Prasad on vision-sdk?
    --> I have one updated VSDK with that also issue is seen.


    Regards,
    Ikshwaku
  • Hello Ram,

    I have shared you logs, I collected as per gokuls input.
    Can you have a look and let me know if you are expecting these logs?

    Regards,
    Ikshwaku
  • Hi Ikshwaku,
    Thanks for sharing this . This shows zero traffic for IVA and VPE(SCI_IVA , SCI_VPE_P1, SCI_VPE_P2). Similary GPU data is also not seems to be correct.
    Please get this data after you playback the video for atleast 1 minute.

    Please get this data for a) 1080p@60 video playback and b)1080p@60 video playback + Navi
  • Hi Ikshwaku,

    Please refer the attached statistics data. This is got for same 1080p@60fps playback.

    stat_video_playback.txt
    [HOST] [IPU2  ]  86584.302232 s:        STATISTIC          Avg Data        Peak Data
     [HOST] [IPU2  ]  86584.302324 s:        COLLECTOR          MB/s            MB/s      
     [HOST] [IPU2  ]  86584.302415 s:  --------------------------------------------------
     [HOST] [IPU2  ]  86584.302507 s:  SCI_EMIF1 RD+WR      |   1152.433356   1934.048780
     [HOST] [IPU2  ]  86584.302629 s:  SCI_EMIF2 RD+WR      |   1142.831088   1897.721641
     [HOST] [IPU2  ]  86584.302751 s:  SCI_EMIF1 RD ONLY    |    768.202325   1358.294448
     [HOST] [IPU2  ]  86584.303147 s:  SCI_EMIF1 WR ONLY    |    384.572418    780.787769
     [HOST] [IPU2  ]  86584.303269 s:  SCI_EMIF2 RD ONLY    |    757.191509   1328.350764
     [HOST] [IPU2  ]  86584.303391 s:  SCI_EMIF2 WR ONLY    |    386.163515    772.306578
     [HOST] [IPU2  ]  86584.303483 s:  SCI_MA_MPU_P1        |     13.286201    413.235720
     [HOST] [IPU2  ]  86584.303605 s:  SCI_MA_MPU_P2        |     12.698753    393.798873
     [HOST] [IPU2  ]  86584.303727 s:  SCI_DSS              |    496.678825    521.818255
     [HOST] [IPU2  ]  86584.303879 s:  SCI_IPU1             |      0.000000      0.000000
     [HOST] [IPU2  ]  86584.303971 s:  SCI_VIP1_P1          |      0.000000      0.000000
     [HOST] [IPU2  ]  86584.304093 s:  SCI_VIP1_P2          |      0.000000      0.000000
     [HOST] [IPU2  ]  86584.304184 s:  SCI_VPE_P1           |      0.000000      0.000000
     [HOST] [IPU2  ]  86584.304306 s:  SCI_VPE_P2           |      0.000000      0.000000
     [HOST] [IPU2  ]  86584.304398 s:  SCI_DSP1_MDMA        |      0.680605      0.974451
     [HOST] [IPU2  ]  86584.304520 s:  SCI_DSP1_EDMA        |      0.000000      0.000000
     [HOST] [IPU2  ]  86584.304611 s:  SCI_DSP2_MDMA        |      0.680458      0.974451
     [HOST] [IPU2  ]  86584.304733 s:  SCI_DSP2_EDMA        |      0.000000      0.000000
     [HOST] [IPU2  ]  86584.304855 s:  SCI_EVE1_TC0         |      2.733627     10.227908
     [HOST] [IPU2  ]  86584.305069 s:  SCI_EVE1_TC1         |      0.000000      0.000000
     [HOST] [IPU2  ]  86584.305191 s:  SCI_EVE2_TC0         |      2.925691     11.141143
     [HOST] [IPU2  ]  86584.305282 s:  SCI_EVE2_TC1         |      0.000000      0.000000
     [HOST] [IPU2  ]  86584.305404 s:  SCI_EDMA_TC0_RD      |      0.000000      0.000000
     [HOST] [IPU2  ]  86584.305496 s:  SCI_EDMA_TC0_WR      |      0.000000      0.000000
     [HOST] [IPU2  ]  86584.305618 s:  SCI_EDMA_TC1_RD      |      0.084218      0.369636
     [HOST] [IPU2  ]  86584.305709 s:  SCI_EDMA_TC1_WR      |      0.084240      0.369636
     [HOST] [IPU2  ]  86584.305862 s:  SCI_VIP2_P1          |      0.000000      0.000000
     [HOST] [IPU2  ]  86584.305984 s:  SCI_VIP2_P2          |      0.000000      0.000000
     [HOST] [IPU2  ]  86584.306075 s:  SCI_VIP3_P1          |      0.000000      0.000000
     [HOST] [IPU2  ]  86584.306197 s:  SCI_VIP3_P2          |      0.000000      0.000000
     [HOST] [IPU2  ]  86584.306319 s:  SCI_EVE3_TC0         |      2.885446      6.657707
     [HOST] [IPU2  ]  86584.306411 s:  SCI_EVE3_TC1         |      0.000000      0.000000
     [HOST] [IPU2  ]  86584.306533 s:  SCI_EVE4_TC0         |      2.890280      6.726803
     [HOST] [IPU2  ]  86584.306624 s:  SCI_EVE4_TC1         |      0.000000      0.000000
     [HOST] [IPU2  ]  86584.306746 s:  SCI_IVA              |   1184.602851   2381.367979
     [HOST] [IPU2  ]  86584.306899 s:  SCI_GPU_P1           |    300.064972   1139.275911
     [HOST] [IPU2  ]  86584.307021 s:  SCI_GPU_P2           |    265.469539   1022.587616
     [HOST] [IPU2  ]  86584.307143 s:  SCI_GMAC_SW          |      0.002171      0.147624
     [HOST] [IPU2  ]  86584.307234 s:  SCI_OCMC_RAM1        |      0.000000      0.000000
     [HOST] [IPU2  ]  86584.307356 s:  SCI_OCMC_RAM2        |      0.000000      0.000000
     [HOST] [IPU2  ]  86584.307448 s:  SCI_OCMC_RAM3        |      0.000000      0.000000
    

    Thanks

    Ram

  • Hi Ikshwaku,
    You log has this print "There may be a timestamping problem, or this computer is too slow."
    Are you often seeing this for this video playback? Is this observed without Navigation launch ?

    I am not able to see this print when I playback video.

    Thanks
    Ram
  • You log has this print "There may be a timestamping problem, or this computer is too slow."
    --> yes these logs comes couple of times

    Please get this data after you playback the video for atleast 1 minute.
    --> I have captured the logs 5-6 times at different time instances. and all are present in log file

    Regards,
    Ikshwaku
  • Hi Ikshwaku,

    >>I have captured the logs 5-6 times at different time instances. and all are present in log file
    yes, there are 7 time instances this data was taken. But in all cases IVA, VPE statistics is 0 and GPU stat is very less.
    Can you give a try again and confirm you get non-zero data?

    Please do these 5 experiments for both video playback with/without navigation app.
    Experiment 1 was already tried out and did not show any improvement.

    1. Regulate IVAHD BandWidth to 1500MBps
    omapconf dump 0x44805000 0x44805014
    omapconf write 0x44805008 0xB5
    omapconf write 0x4480500C 0x5DC

    2.Change DSP1/2 priorities to 0. Change IPU2, IVA and VPE priorities to 1
    omapconf write 0x4e000624 0x09888888
    omapconf write 0x4e00062C 0x00000090
    omapconf write 0x4e000630 0x90000000


    3.Increase EMIF system port priority, increase priorities of IPU2, IVA and VPE
    omapconf write 0x40D00040 0xC0000000
    omapconf write 0x4e000624 0x09888888
    omapconf write 0x4e00062C 0x00000090
    omapconf write 0x4e000630 0x90000000

    Experiment 1 : Regulate IVA BW(1)-----> This is done and no improvement
    Experiment 2 = Increase IVA Priority (2)
    Experiment 3 = Regulate IVA BW + Increase Priority(1 + 2)
    Experiment 4 = Increase EMIF Priority and IVA priority (3)
    Experiment 5 = Regulate IVA BW + Increase EMIF Priority and IVA priority(1 + 3)

    Thanks
    Ram
  • Hi Ikshwaku,
    As discussed please try these two configurations if the remaining 4 experiments don't show improvement.

    # GPU (limit each port to 750MB/s)
    omapconf write 0x44805B08 0x1A
    omapconf write 0x44805B0C 0x2
    omapconf write 0x44805B10 0
    omapconf write 0x44805B14 1
    omapconf write 0x44805C08 0x1A
    omapconf write 0x44805C0C 0x2
    omapconf write 0x44805C10 0
    omapconf write 0x44805C14 1

    # VPE (limit each port to 750MB/s)
    omapconf write 0x44804008 0x1A
    omapconf write 0x4480400C 0x2
    omapconf write 0x44804010 0
    omapconf write 0x44804014 1
    omapconf write 0x44804108 0x1A
    omapconf write 0x4480410C 0x2
    omapconf write 0x44804110 0
    omapconf write 0x44804114 1

    Experiment 6 = BW Limit VPE
    Experiment 7 = BW Limit GPU
    Experiment 8 = BW Limit VPE + GPU

    Thanks
    Ram
  • Hello Ram,

    I have tried these setting, but no improvement.

    Regards,
    Ikshwaku
  • Hi Ikshwaku,
    Can you quickly try this.
    1) export GST_DEBUG=ducati:5 and execute the pipeline and check if you are seeing the logs for ducati plugin.
    This is to confirm if IVAHD is getting used or not as statistics don't show for IVAHD

    2) Generate pipeline graphs .
    export GST_DEBUG_DUMP_DOT_DIR=/tmp/

    after playback starts you can .dot files in /tmp/.
    You can share these files to me.

    Thanks
    Ram
  • Hello Ram,

    1) export GST_DEBUG=ducati:5 and execute the pipeline and check if you are seeing the logs for ducati plugin.
    --> yes getting the ducati logs after enabling the debug flag

    Shared the dot file.

    Regards,
    Ikshwaku
  • Hi Ikshwaku,
    I verified the dot file you shared. It looks proper. Ducati and vpe plugins are properly configured.

    Did you check if vision-sdk statistics show proper results for a small resolution streams and 30fps streams?

    I talked to expert on this issue. Since issue is not reproduced with vision-sdk, it was recommended 1) on TI's vision-SDK setup add one module at a time or 2) to remove one by one module or to narrow down which module is causing the issue.
    You can comment all services from service start-up script and enable one by one to check which service is causing the issue.


    Thanks
    Ram
  • Hello Ram,

    I have observed one behavior: on basic SDK build(without hmi, nav, carplay.. etc) if i check the stats its sometimes show zero usage for IVAHD, but video plays properly. But on complete build video playback is not smooth with IVA usage as zero.

    Last time I have check that time it shows the values, but today i have checked and its showing as follow for all the video resolutions:

    [0][ 41.705] 44.519490 s: SCI_IVA | 0.000000 0.000000
    [0][ 41.706] 44.519703 s: SCI_GPU_P1 | 0.023248 0.026397
    [0][ 41.706] 44.519978 s: SCI_GPU_P2 | 0.052317 0.059393


    Regards,
    Ikshwaku
  • Hello Ram,

    We have identified one if we remove DPSIPC driver, then video playback is smooth on 7inch display.
    DSPIPC is a IPC driver which enables audio manager to send commands to j6 DSP.

    Do you have inputs for this behaviour..??


    Regards,
    Ikshwaku
  • Hi Ikshwaku,
    Can you try to get details of these ?
    1) What HW IPs , components, softwares are being used in DSPIPC service? Does it use both the DDSPs r rSPs or one DSP
    2) Is there any priority set for this service?
    3) Collect DSP1 and DSP2 statistics from vision-sdk
    4) Get stats with statcollector, use only following entries in initiator.cfg

    STATCOL_EMIF1_SYS
    STATCOL_EMIF2_SYS
    STATCOL_DSP1_MDMA
    STATCOL_DSP1_EDMA
    STATCOL_DSP2_MDMA
    STATCOL_DSP2_EDMA
    STATCOL_DSP1_CFG
    STATCOL_DSP2_CFG

    Thanks
    Ram
  • Hello Ram ,

    Can you please clarify this "DDSPs r rSPs"?

    Regards,
    Ikshwaku
  • Hi Ikswaku,
    Sorry for the type. I wanted to know if both DSPs are used for this service or one DSP. If one DSP which DSP is used?

    Thanks
    Ram
  • Hello Ram,

    Can you share me the "glsdkstatcoll" for VSDK3.04, the one i have is from PSDKLA5?
    One more thing to add it looks like if we remove dspipc issue is not seen on 7 inch display setup, but after removing dspipc on 12 inch setup issues is reproducible.

    Regards,
    Ikshwaku
  • Hi Ikshwaku,

    Glsdkstatcollecor executable I share should work, do you see any issue?

    Please try to narrow down which other services are casuing the slow playback issue on 12 inch display.

    Ram

  • Hello Ram,

    I think you have only shared "initiators.cfg" not "Glsdkstatcollecor".

    Regards,
    Ikshwaku
  • Hi Ikshwaku,

    Please find the glsdkstatcollector binary and cfg files.

    glsdkstatcoll.zip

    Before getting the statistics can you do below experiments?

    1) Please get the CPU load a)just after booting b)when video playback happens for 60fps stream (no DSPIPC service) c) video playback for 60fps stream with DSPIPC service available.

    I discussed on this issue internally. We suspect audio could be the issue here. Audio playback for this stream could be causing video to drop.

    Is there a way for you to disable audio and check ?

    You can check gst-launch with audio-sink set to fakesink, and check if it reproduces the issue.

    Thanks

    Ram

  • Hi Ikshwaku,
    Can you do one profiling for IVA with gstreamer?
    Gstreamer ducati always calculates the time taken to decode a frame. but it is printed onn console only with debug level 5.

    You can set export GST_DEBUG=ducati:5
    and then execute the gst pipeline.
    You can see such prints
    ducati gstducatividdec.c:566:codec_process:<decoder> VIDDEC3_process took 11979099ns (11 ms)
    For this lamberighini stream , time taken is 11 to 12 ms with video playback smooth(You can check with working case).

    In non-working case, please get this data. Need to check if decoding time is more when DSPIPC service is running in background.

    Thanks
    Ram
  • Hello ram,

    With your provided glsdksatatcoll i am getting the error:

    root@mmt2020-a880:~/glsdkstat# ./glsdkstatcoll -f config.ini
    NOTE: BANDWIDTH is not enabled, ignoring DELAY
    NOTE: STATCOLL is not enabled, ignoring DELAY
    NOTE: BANDWIDTH is not enabled, ignoring EMIF_PERF_CFG1
    NOTE: STATCOLL is not enabled, ignoring EMIF_PERF_CFG1
    NOTE: BANDWIDTH is not enabled, ignoring EMIF_PERF_CFG2
    NOTE: STATCOLL is not enabled, ignoring EMIF_PERF_CFG2
    Invalid key found
    Supported keys are :
    DELAY
    EMIF_PERF_CFG1
    EMIF_PERF_CFG2
    BANDWIDTH
    STATCOLL
    TOTAL_TIME
    INTERVAL_US
    INITIATORS



    Regards,
    Ikshwaku
  • Hi Ikswaku,

    I tried with attached config.ini and initiator.cfg.

    Can you try with this one?

    glsdkstat.zip

    You should see this trace

    root@dra7xx-evm:~# glsdkstatcoll -f config.ini
    NOTE: BANDWIDTH is not enabled, ignoring DELAY
    NOTE: STATCOLL is not enabled, ignoring DELAY
    NOTE: BANDWIDTH is not enabled, ignoring EMIF_PERF_CFG1
    NOTE: STATCOLL is not enabled, ignoring EMIF_PERF_CFG1
    NOTE: BANDWIDTH is not enabled, ignoring EMIF_PERF_CFG2
    NOTE: STATCOLL is not enabled, ignoring EMIF_PERF_CFG2
    NOTE: BANDWIDTH is not enabled, ignoring TOTAL_TIME
    NOTE: BANDWIDTH is not enabled, ignoring INTERVAL_US


    COMPLETED: Parsing of the user specified parameters..

    Configuring device now..

    STATISTICS COLLECTOR option chosen
    ------------------------------------------------

    ------------------------------------------------
    Compile time = Dec 9 2017 02:24:49
    ------------------------------------------------

    Total configured initiators = 9
    SUCCESS: Mapped 0x45001000 to user space address 0xb6f7d000
    INTERVAL = 30000 usecs
    TOTAL TIME = 120 seconds
    TRACE SIZE = 4000 samples
    **************************************
    Going to initialize the L3 clocks
    **************************************
    SUCCESS: Initialized STAT COLLECTOR
    Initialized STATCOL_EMIF1_SYS
    Initialized STATCOL_EMIF2_SYS
    Initialized STATCOL_DSP1_MDMA
    Initialized STATCOL_DSP1_EDMA
    Initialized STATCOL_DSP2_MDMA
    Initialized STATCOL_DSP2_EDMA
    Initialized STATCOL_DSP1_CFG
    Initialized STATCOL_DSP2_CFG

    Could you check if you can see the issue if audio is disabled?

    Ram

  • Hello Ram ,

    I have checked with ducati enabled logs and getting following values:
    Line 114: ducati gstducatividdec.c:567:codec_process:<decoder> VIDDEC3_process took 15995554ns (15 ms)
    Line 119: ducati gstducatividdec.c:567:codec_process:<decoder> VIDDEC3_process took 15030568ns (15 ms)
    Line 125: ducati gstducatividdec.c:567:codec_process:<decoder> VIDDEC3_process took 16307501ns (16 ms)
    Line 126: ducati gstducatividdec.c:629:codec_process:<decoder> VIDDEC3_process outputID[0]: b48030a8
    Line 134: ducati gstducatividdec.c:567:codec_process:<decoder> VIDDEC3_process took 16074006ns (16 ms)
    Line 135: ducati gstducatividdec.c:629:codec_process:<decoder> VIDDEC3_process outputID[0]: b4803328
    Line 139: ducati gstducatividdec.c:769:codec_process:<decoder> VIDDEC3_process freeBufID[0]: b4803328
    Line 151: ducati gstducatividdec.c:567:codec_process:<decoder> VIDDEC3_process took 20598850ns (20 ms)
    Line 152: ducati gstducatividdec.c:629:codec_process:<decoder> VIDDEC3_process outputID[0]: b4803288
    Line 156: ducati gstducatividdec.c:769:codec_process:<decoder> VIDDEC3_process freeBufID[0]: b48030a8
    Line 165: ducati gstducatividdec.c:567:codec_process:<decoder> VIDDEC3_process took 16804820ns (16 ms)
    Line 166: ducati gstducatividdec.c:629:codec_process:<decoder> VIDDEC3_process outputID[0]: 00135a40
    Line 170: ducati gstducatividdec.c:769:codec_process:<decoder> VIDDEC3_process freeBufID[0]: 00135a40
    Line 177: ducati gstducatividdec.c:567:codec_process:<decoder> VIDDEC3_process took 16085529ns (16 ms)
    Line 178: ducati gstducatividdec.c:629:codec_process:<decoder> VIDDEC3_process outputID[0]: b4803008
    Line 182: ducati gstducatividdec.c:769:codec_process:<decoder> VIDDEC3_process freeBufID[0]: b4803008
    Line 189: ducati gstducatividdec.c:567:codec_process:<decoder> VIDDEC3_process took 15474851ns (15 ms)
    Line 190: ducati gstducatividdec.c:629:codec_process:<decoder> VIDDEC3_process outputID[0]: b4803148
    Line 194: ducati gstducatividdec.c:769:codec_process:<decoder> VIDDEC3_process freeBufID[0]: b4803288
    Line 203: ducati gstducatividdec.c:567:codec_process:<decoder> VIDDEC3_process took 16557601ns (16 ms)
    Line 204: ducati gstducatividdec.c:629:codec_process:<decoder> VIDDEC3_process outputID[0]: 00135ea0
    Line 208: ducati gstducatividdec.c:769:codec_process:<decoder> VIDDEC3_process freeBufID[0]: 00135ea0
    Line 216: ducati gstducatividdec.c:567:codec_process:<decoder> VIDDEC3_process took 17263168ns (17 ms)
    Line 217: ducati gstducatividdec.c:629:codec_process:<decoder> VIDDEC3_process outputID[0]: 00135cc0
    Line 221: ducati gstducatividdec.c:769:codec_process:<decoder> VIDDEC3_process freeBufID[0]: 00135cc0
    Line 228: ducati gstducatividdec.c:567:codec_process:<decoder> VIDDEC3_process took 15549407ns (15 ms)
    Line 229: ducati gstducatividdec.c:629:codec_process:<decoder> VIDDEC3_process outputID[0]: 00135f40
    Line 233: ducati gstducatividdec.c:769:codec_process:<decoder> VIDDEC3_process freeBufID[0]: b4803148
    Line 242: ducati gstducatividdec.c:567:codec_process:<decoder> VIDDEC3_process took 16491518ns (16 ms)
    Line 243: ducati gstducatividdec.c:629:codec_process:<decoder> VIDDEC3_process outputID[0]: 00135a40
    Line 247: ducati gstducatividdec.c:769:codec_process:<decoder> VIDDEC3_process freeBufID[0]: 00135a40
    Line 254: ducati gstducatividdec.c:567:codec_process:<decoder> VIDDEC3_process took 16598776ns (16 ms)
    Line 255: ducati gstducatividdec.c:629:codec_process:<decoder> VIDDEC3_process outputID[0]: b4803008
    Line 259: ducati gstducatividdec.c:769:codec_process:<decoder> VIDDEC3_process freeBufID[0]: b4803008
    Line 267: ducati gstducatividdec.c:567:codec_process:<decoder> VIDDEC3_process took 15430287ns (15 ms)
    Line 268: ducati gstducatividdec.c:629:codec_process:<decoder> VIDDEC3_process outputID[0]: b4803328
    Line 272: ducati gstducatividdec.c:769:codec_process:<decoder> VIDDEC3_process freeBufID[0]: 00135f40
    Line 280: ducati gstducatividdec.c:567:codec_process:<decoder> VIDDEC3_process took 17119139ns (17 ms)
    Line 281: ducati gstducatividdec.c:629:codec_process:<decoder> VIDDEC3_process outputID[0]: 00135ea0
    Line 285: ducati gstducatividdec.c:769:codec_process:<decoder> VIDDEC3_process freeBufID[0]: 00135ea0
    Line 292: ducati gstducatividdec.c:567:codec_process:<decoder> VIDDEC3_process took 17375679ns (17 ms)

    as per the logs it is taking 15,16,17,20 ms for decoding.

    And one more thing after removing DSPIPC, if we enable debug logs then issue can be seen, but if we are not enabling the debug logs then issue is not seen.

    Regards,
    ikshwaku
  • Also, i have verified by by adding audio-sink=fakesink in the pipeline, but issues is seen.

    Regards,
    Ikshwaku
  • Hello Ram,

    We have verified with video only file and with that also issue is seen.

    Regards,
    Ikshwaku