Re: [i-g-t 1/2] tests/amdgpu/amd_dp_dsc: Do not disable DSC for high bw panels

[George Zhang <george.zhang@amd.com>]

On 9/3/26 14:29, Alex Hung wrote:
> 
> 
> On 9/2/26 09:57, George Zhang wrote:
>> The test previously attempted to disable DSC even on high bandwidth
>> panels. This would erroneously fail the dsc_enable test even though DSC
>> is on.
>>
>> Fix this by checking what mode the panel defaults to when set to
>> automatic. If DSC is on, then do not attempt to force it off later.
> 
> Doesn't high bandwidth panel need DSC off? How is a high bandwidth panel 
> expected to behave?

For a high bandwidth panel (4k240 for example), DSC cannot be off, 
otherwise the panel will be starved of data and go dark.

I'll follow up with a v2 to fix the issue in this series

Thanks,
George

> 
>>
>> Signed-off-by: George Zhang <george.zhang@amd.com>
>> ---
>>   tests/amdgpu/amd_dp_dsc.c | 11 +++++++++--
>>   1 file changed, 9 insertions(+), 2 deletions(-)
>>
>> diff --git a/tests/amdgpu/amd_dp_dsc.c b/tests/amdgpu/amd_dp_dsc.c
>> index dc71e903e..ccaf64c66 100644
>> --- a/tests/amdgpu/amd_dp_dsc.c
>> +++ b/tests/amdgpu/amd_dp_dsc.c
>> @@ -137,6 +137,9 @@ static void test_dsc_enable(data_t *data)
>>           igt_output_set_crtc(output,
>>                       crtc);
>>           igt_plane_set_fb(data->primary[crtc->crtc_index], &ref_fb);
>> +
>> +        /* Make sure we are in automatic mode before we start*/
> Add a space after start, i.e., "start */"
> 
>> +        igt_amd_write_dsc_clock_en(data->fd, output->name, 
>> DSC_AUTOMATIC);
>>           igt_display_commit_atomic(display, 
>> DRM_MODE_ATOMIC_ALLOW_MODESET, 0);
>>
>>           test_conn_cnt++;
>> @@ -154,7 +157,9 @@ static void test_dsc_enable(data_t *data)
>>           dsc_on = igt_amd_read_dsc_clock_status(data->fd, output- 
>> >name) == 1;
>>
>>           /* Revert DSC to automatic state */
>> -        igt_amd_write_dsc_clock_en(data->fd, output->name, 
>> DSC_FORCE_OFF);
>> +        if (!dsc_before)
>> +            /* We can only force off if the panel allows it. If DSC 
>> was on in automatic mode, we cannot force it off. */
> Break this one-line comment to multiple lines, and place it before "if 
> (!dsc_before)"
> 
>> +            igt_amd_write_dsc_clock_en(data->fd, output->name, 
>> DSC_FORCE_OFF);
>>
>>           igt_plane_set_fb(data->primary[crtc->crtc_index], &ref_fb);
>>           igt_display_commit_atomic(display, 
>> DRM_MODE_ATOMIC_ALLOW_MODESET, NULL);
>> @@ -286,6 +291,7 @@ static void 
>> test_dsc_slice_dimensions_change(data_t *data)
>>           igt_output_set_crtc(output,
>>                       crtc);
>>           igt_plane_set_fb(data->primary[crtc->crtc_index], &ref_fb);
>> +        igt_amd_write_dsc_clock_en(data->fd, output->name, 
>> DSC_AUTOMATIC);
>>           igt_display_commit_atomic(display, 
>> DRM_MODE_ATOMIC_ALLOW_MODESET, 0);
>>
>>           test_conn_cnt++;
>> @@ -317,7 +323,8 @@ static void 
>> test_dsc_slice_dimensions_change(data_t *data)
>>           }
>>
>>           /* Force disable DSC */
> Please update this comment too.
> 
>> -        igt_amd_write_dsc_clock_en(data->fd, output->name, 
>> DSC_FORCE_OFF);
>> +        if (!dsc_before)
>> +            igt_amd_write_dsc_clock_en(data->fd, output->name, 
>> DSC_FORCE_OFF);
>>
>>           igt_plane_set_fb(data->primary[crtc->crtc_index], &ref_fb);
>>           igt_display_commit_atomic(display, 
>> DRM_MODE_ATOMIC_ALLOW_MODESET, NULL);
>> -- 
>> 2.55.0
>