Date:   Sat, 19 Feb 2022 00:36:45 +0300
From:   Dmitry Baryshkov <dmitry.baryshkov@...aro.org>
To:     Abhinav Kumar <quic_abhinavk@...cinc.com>,
        Vinod Koul <vkoul@...nel.org>
Cc:     Jonathan Marek <jonathan@...ek.ca>,
        David Airlie <airlied@...ux.ie>, linux-arm-msm@...r.kernel.org,
        linux-kernel@...r.kernel.org,
        Abhinav Kumar <abhinavk@...eaurora.org>,
        Bjorn Andersson <bjorn.andersson@...aro.org>,
        Rob Clark <robdclark@...il.com>,
        dri-devel@...ts.freedesktop.org, Daniel Vetter <daniel@...ll.ch>,
        freedreno@...ts.freedesktop.org
Subject: Re: [Freedreno] [REPOST PATCH v4 08/13] drm/msm/disp/dpu1: Don't use
 DSC with mode_3d

On 19/02/2022 00:29, Abhinav Kumar wrote:
> 
> 
> On 2/18/2022 1:21 PM, Dmitry Baryshkov wrote:
>> On 18/02/2022 23:46, Abhinav Kumar wrote:
>>>
>>>
>>> On 2/16/2022 11:12 PM, Dmitry Baryshkov wrote:
>>>> On 17/02/2022 09:33, Abhinav Kumar wrote:
>>>>>
>>>>>
>>>>> On 2/16/2022 10:10 PM, Vinod Koul wrote:
>>>>>> On 16-02-22, 19:11, Abhinav Kumar wrote:
>>>>>>>
>>>>>>>
>>>>>>> On 2/10/2022 2:34 AM, Vinod Koul wrote:
>>>>>>>> We cannot enable mode_3d when we are using the DSC. So pass
>>>>>>>> configuration to detect DSC is enabled and not enable mode_3d
>>>>>>>> when we are using DSC
>>>>>>>>
>>>>>>>> We add a helper dpu_encoder_helper_get_dsc() to detect dsc
>>>>>>>> enabled and pass this to .setup_intf_cfg()
>>>>>>>>
>>>>>>>> Signed-off-by: Vinod Koul <vkoul@...nel.org>
>>>>>>>
>>>>>>> We should not use 3D mux only when we use DSC merge topology.
>>>>>>> I agree that today we use only 2-2-1 topology for DSC which means 
>>>>>>> its using
>>>>>>> DSC merge.
>>>>>>>
>>>>>>> But generalizing that 3D mux should not be used for DSC is not 
>>>>>>> right.
>>>>>>>
>>>>>>> You can detect DSC merge by checking if there are two encoders 
>>>>>>> and one
>>>>>>> interface in the topology and if so, you can disable 3D mux.
>>>>>>
>>>>>> Right now with DSC we disable that as suggested by Dmitry last time.
>>>>>> Whenever we introduce merge we should revisit this, for now this 
>>>>>> should
>>>>>> suffice
>>>>>>
>>>>>
>>>>> Sorry I didnt follow.
>>>>>
>>>>> The topology which you are supporting today IS DSC merge 2-2-1. I 
>>>>> didnt get what you mean by "whenever we introduce".
>>>>>
>>>>> I didnt follow Dmitry's comment either.
>>>>>
>>>>> "anybody adding support for SDE_RM_TOPOLOGY_DUALPIPE_3DMERGE_DSC 
>>>>> handle this."
>>>>>
>>>>> 3D mux shouldnt be used when DSC merge is used.
>>>>>
>>>>> The topology Dmitry is referring to will not use DSC merge but you 
>>>>> are using it here and thats why you had to make this patch in the 
>>>>> first place. So I am not sure why would someone who uses 3D merge 
>>>>> topology worry about DSC merge. Your patch is the one which deals 
>>>>> with the topology in question.
>>>>>
>>>>> What I am suggesting is a small but necessary improvement to this 
>>>>> patch.
>>>>
>>>> It seems that we can replace this patch by changing 
>>>> dpu_encoder_helper_get_3d_blend_mode() to contain the following 
>>>> condition (instead of the one present there). Does the following 
>>>> seem correct to you:
>>>>
>>>> static inline enum dpu_3d_blend_mode 
>>>> dpu_encoder_helper_get_3d_blend_mode(
>>>>                  struct dpu_encoder_phys *phys_enc)
>>>> {
>>>>          struct dpu_crtc_state *dpu_cstate;
>>>>
>>>>          if (!phys_enc || phys_enc->enable_state == DPU_ENC_DISABLING)
>>>>                  return BLEND_3D_NONE;
>>>>
>>>>          dpu_cstate = to_dpu_crtc_state(phys_enc->parent->crtc->state);
>>>>
>>>> +    /* Use merge_3d unless DSCMERGE topology is used */
>>>>          if (phys_enc->split_role == ENC_ROLE_SOLO &&
>>>> +           hweight(dpu_encoder_helper_get_dsc(phys_enc)) != 1 &&
>>
>> Yes, the correct should be:
>> hweight(...) == 2
>>
>>>>              dpu_cstate->num_mixers == CRTC_DUAL_MIXERS)
>>>>                  return BLEND_3D_H_ROW_INT;
>>>>
>>>>          return BLEND_3D_NONE;
>>>> }
>>>
>>> This will not be enough. To detect whether DSC merge is enabled you 
>>> need to query the topology. The above condition only checks if DSC is 
>>> enabled not DSC merge.
>>>
>>> So the above function can be modified to use a helper like below 
>>> instead of the hweight.
>>>
>>> bool dpu_encoder_get_dsc_merge_info(struct dpu_encoder_virt *dpu_enc)
>>> {
>>>      struct msm_display_topology topology = {0};
>>>
>>>      topology = dpu_encoder_get_topology(...);
>>>
>>>      if (topology.num_dsc > topology.num_intf)
>>
>> num_intf is 1 or 2. If it's one, the split_role is SOLO
>> hweight would return a num of bits in the DSC mask. It's 0, 1 or 2.
>> So, if the split_role is SOLO and hweight is 2, we get exactly your 
>> condition.
>>
> num_intf is 1 in this case as only single interface is used. But even 4 
> dsc encoders going to 2 interfaces its DSC merge. So assuming num_intf 
> as 1 always is not right. Thats why I suggested a generalized confition 
> like above.

Ah, quadpipe. Yes, then I was incorrect.

> 
>> Does that sound correct?
>>
>>>          return true;
>>>      else
>>>          return false;
>>> }
>>>
>>> if (!dpu_encoder_get_dsc_merge_info() && other conditions listed above)
>>>      return BLEND_3D_H_ROW_INT;
>>> else
>>>      BLEND_3D_NONE;
>>>>
>>>>
>>>>>
>>>>> All that you have to do is in query whether DSC merge is used from 
>>>>> the topology. You can do it in multiple ways:
>>>>>
>>>>> 1) Either query this from the encoder
>>>>> 2) Store a bool "dsc_merge" in the intf_cfg
>>>>>
>>>>
>>>>
>>
>>


-- 
With best wishes
Dmitry

Powered by blists - more mailing lists