JP7.2 / R39.2 — Does GMSL Camera HAL support YUV422 (UYVY) ICP?

Robotics & Edge Computing Jetson Systems Jetson Thor
1

Hi NVIDIA team,

We are bringing up a Maxim MAX96724 + MAX9295 + Sony ISX031 module on Jetson AGX Thor, JetPack 7.2 (L4T R39.2) using SIPL + UDDF (Camera HAL / -H).

Working reference (legacy V4L2):
UYVY 1920×1536 @ 30 fps. VI logs show csi_stream=2, vc=1, VI_PIXFMT_FORMAT_T_U8_Y8__V8_Y8. Deserializer DT is YUV422-8 on VC1 (0x1E/0x5E).

SIPL issue:

  • nvsipl_query JSON inputFormat values such as yuv422 / uyvy are not accepted (fall back to RAW12). The SIPL Query JSON Guide lists only raw* for GMSL.
  • Forcing NVSIPL_CAP_INPUT_FORMAT_TYPE_YUV422 + NVSIPL_PIXEL_ORDER_UYVY after parse causes SetPlatformConfig (Camera HAL) to fail with NVSIPL_STATUS_ERROR.
  • uddf::ddi::PixPackingType only defines RAW10/12/16; libnvcamerahal.so contains "Unsupported input format: %d".
  • With inputFormat: raw16, Init/Start succeed and GMSL locks, but VI is programmed as T_R16 / vc=0 and we get NOTIF_WARN_ICP_CAPTURE_TIMEOUT (0 frames).

Questions:

  1. Is YUV422-8 / UYVY ICP supported on the GMSL Camera HAL path in JP7.2?
  2. If yes, what are the correct JSON inputFormat / cfa tokens (or required API overrides)?
  3. If not, is support planned, or is the recommended approach to keep V4L2 for YUV GMSL cameras?

Thank you.

3

同一个问题，插眼等待中

4

Looks like camera HAL don’t support YUV422.

However, for the YUV sensor you can just by pass the SIPL to get the YUV frame data right?

5

Yes, we can get the YUV frame by V4L2.

Do you have any plan to add YUV422 to GMSL Camera HAL?

6

Hi, here’s some information from internal discussion.

Does GMSL support YUV sensors?

Yes — YUV422 8-bit (packed/SP) and YUV422_10 (SP)

GMSL JSON tokens?

“422” + cfa yuyv/uyvy/…; “422_10” + cfa yuv/yvu