Re: [PATCH RFC 2/7] media: qcom: iris: gen2: add support for 10bit decoding

[Dikshita Agarwal <dikshita.agarwal@oss.qualcomm.com>]

On 4/17/2026 1:33 PM, Neil Armstrong wrote:
> On 4/17/26 09:22, Dikshita Agarwal wrote:
>>
>>
>> On 4/8/2026 10:13 PM, Neil Armstrong wrote:
>>> Add the necessary plumbing into the HFi Gen2 to signal the decoder
>>> the right 10bit pixel format and stride when in compressed mode.
>>>
>>> Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>
>>> ---
>>>   .../platform/qcom/iris/iris_hfi_gen2_command.c     | 71
>>> +++++++++++++++++++++-
>>>   .../platform/qcom/iris/iris_hfi_gen2_defines.h     |  1 +
>>>   drivers/media/platform/qcom/iris/iris_utils.c      |  4 +-
>>>   3 files changed, 72 insertions(+), 4 deletions(-)
>>>
>>> diff --git a/drivers/media/platform/qcom/iris/iris_hfi_gen2_command.c
>>> b/drivers/media/platform/qcom/iris/iris_hfi_gen2_command.c
>>> index 30bfd90d423b..8e547e390fa3 100644
>>> --- a/drivers/media/platform/qcom/iris/iris_hfi_gen2_command.c
>>> +++ b/drivers/media/platform/qcom/iris/iris_hfi_gen2_command.c
>>> @@ -481,8 +481,20 @@ static int iris_hfi_gen2_set_colorformat(struct
>>> iris_inst *inst, u32 plane)
>>>         if (inst->domain == DECODER) {
>>>           pixelformat = inst->fmt_dst->fmt.pix_mp.pixelformat;
>>> -        hfi_colorformat = pixelformat == V4L2_PIX_FMT_NV12 ?
>>> -            HFI_COLOR_FMT_NV12 : HFI_COLOR_FMT_NV12_UBWC;
>>> +        switch (pixelformat) {
>>> +        case V4L2_PIX_FMT_NV12:
>>> +            hfi_colorformat = HFI_COLOR_FMT_NV12;
>>> +            break;
>>> +        case V4L2_PIX_FMT_QC08C:
>>> +            hfi_colorformat = HFI_COLOR_FMT_NV12_UBWC;
>>> +            break;
>>> +        case V4L2_PIX_FMT_P010:
>>> +            hfi_colorformat = HFI_COLOR_FMT_P010;
>>> +            break;
>>> +        case V4L2_PIX_FMT_QC10C:
>>> +            hfi_colorformat = HFI_COLOR_FMT_TP10_UBWC;
>>> +            break;
>>> +        };
>>>       } else {
>>>           pixelformat = inst->fmt_src->fmt.pix_mp.pixelformat;
>>>           hfi_colorformat = pixelformat == V4L2_PIX_FMT_NV12 ?
>>> @@ -517,7 +529,8 @@ static int
>>> iris_hfi_gen2_set_linear_stride_scanline(struct iris_inst *inst, u32
>>>       stride_uv = stride_y;
>>>       scanline_uv = scanline_y / 2;
>>>   -    if (pixelformat != V4L2_PIX_FMT_NV12)
>>> +    if (pixelformat != V4L2_PIX_FMT_NV12 &&
>>> +        pixelformat != V4L2_PIX_FMT_P010)
>>>           return 0;
>>>         payload[0] = stride_y << 16 | scanline_y;
>>> @@ -532,6 +545,57 @@ static int
>>> iris_hfi_gen2_set_linear_stride_scanline(struct iris_inst *inst, u32
>>>                             sizeof(u64));
>>>   }
>>>   +static int iris_hfi_gen2_set_ubwc_stride_scanline(struct iris_inst
>>> *inst, u32 plane)
>>> +{
>>> +    u32 meta_stride_y, meta_scanline_y, meta_stride_uv, meta_scanline_uv;
>>> +    u32 stride_y, scanline_y, stride_uv, scanline_uv;
>>> +    u32 port = iris_hfi_gen2_get_port(inst, plane);
>>> +    u32 pixelformat, width, height;
>>> +    u32 payload[4];
>>> +
>>> +    pixelformat = inst->fmt_dst->fmt.pix_mp.pixelformat;
>>> +    width = inst->fmt_dst->fmt.pix_mp.width;
>>> +    height = inst->fmt_dst->fmt.pix_mp.height;
>>
>> This HFI is only applicable to AV1, you might see some corruption due to
>> this. Please check.
> 
> This is what I saw looking at donwstream, but I had not implemented this
> initially
> but I got some corruption with some different width/height, which was
> solved adding
> this command.

Even with your implementation, this API would only ever be triggered for
QC08C and QC10C formats, which I don’t believe are currently being tested
with your series. Since HFI defines this as AV1D‑specific, invoking it for
other codecs is unlikely to help and would be incorrect.

Pls re-check.

Thanks,
Dikshita
> I guess this in-firmware calculations are not exactly the same the DRM
> driver expects.
> 
> Honestly I think it's preferable to sync the stride/scanlines calculations
> between the
> driver and the firmware.
> For example I used the same "ALIGN(width, 192)" for the v2 iris patchset
> which was wrong
> because ALIGN only works on Power Of Two number but it still worked because
> this HFI
> HFI_PROP_UBWC_STRIDE_SCANLINE command was called with the wrong aligned width.
> 
> Neil
> 
>>
>> Thanks,
>> Dikshita
>