Re: [PATCH v6 4/4] tests/kms_async_flips: Create subtest for overlay planes

[Alex Hung <alex.hung@amd.com>]

On 4/8/25 14:45, André Almeida wrote:
> 
> 
> Em 08/04/2025 16:26, André Almeida escreveu:
>> Em 08/04/2025 15:53, Alex Hung escreveu:
>>>
>>>
>>> On 4/8/25 10:16, Melissa Wen wrote:
>>>> On 04/07, André Almeida wrote:
>>>>> amdgpu can perform async flips in overlay planes as well, so create a
>>>>> test for that.
>>>>
>>>> Hey,
>>>>
>>>> LGTM. Thanks!
>>>>
>>>> Reviewed-by: Melissa Wen <mwen@igalia.com>
>>>>
>>>> Alex,
>>>>
>>>> Can you check if it's fine for AMD too?
>>>
>>> Hi André,
>>>
>>> I ran the subtest "overlay-atomic" on PHX APU + kernel of amd- 
>>> staging- drm-next, and they were skipped.
>>>
>>> What the kernel and ASIC did you run overlay-atomic?
>>
>> I'm using drm-misc/drm-misc-next (eff0347e7c228335e9ff64aaf02c669578) 
>> with a Steam Deck and Melissa tested in an AMD Cezanne (DCN 2.1) GPU.
>>
> 
> Also, please note that amd-staging-drm-next currently doesn't have 
> support for this, so you need to either checkout to drm-next or drm- 
> misc-next, or apply the following commits to your tree:
> 
> 41129e236f14 drm/amdgpu: Enable async flip on overlay planes
> fd40a63c63a1 drm/atomic: Let drivers decide which planes to async flip

Thanks. I was able to run overlay-atomic on either drm-misc-next or asdn 
+ these two commits.

overlay-atomic ran fine on PHX + 1080p panel, but it failed on PHX + 4K 
panel (more specifically, 3840 x 2400) with the following errors:

(kms_async_flips:8296) CRITICAL: Test assertion failure function 
test_async_flip, file ../tests/kms_async_flips.c:431:
(kms_async_flips:8296) CRITICAL: Failed assertion: (fps / 1000) > 
(data->refresh_rate * MIN_FLIPS_PER_FRAME)
(kms_async_flips:8296) CRITICAL: FPS should be significantly higher than 
the refresh rate

Where
     fps / 1000 = 121
     data->refresh_rate * MIN_FLIPS_PER_FRAME = 300

Comparing to the result of 1080p
     fps / 1000 = 1232
     data->refresh_rate * MIN_FLIPS_PER_FRAME = 300

However, this doesn't seem to be directly related to this patch and is 
something we should look into.

Reviewed-by: Alex Hung <alex.hung@amd.com>