drm/msm/dpu: tweaks for better hardware resources allocation

[PATCH 02/11] drm/msm/dpu: use the actual lm maximum width instead of a hardcoded value

Posted by Arnaud Vrac 2 years, 8 months ago

This avoids using two LMs instead of one when the display width is lower
than the maximum supported value. For example on MSM8996/MSM8998, the
actual maxwidth is 2560, so we would use two LMs for 1280x720 or
1920x1080 resolutions, while one is enough.

Signed-off-by: Arnaud Vrac <avrac@freebox.fr>
---
 drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c | 10 +++++-----
 1 file changed, 5 insertions(+), 5 deletions(-)

diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c
index 1dc5dbe585723..dd2914726c4f6 100644
--- a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c
+++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c
@@ -53,8 +53,6 @@
 
 #define IDLE_SHORT_TIMEOUT	1
 
-#define MAX_HDISPLAY_SPLIT 1080
-
 /* timeout in frames waiting for frame done */
 #define DPU_ENCODER_FRAME_DONE_TIMEOUT_FRAMES 5
 
@@ -568,10 +566,12 @@ static struct msm_display_topology dpu_encoder_get_topology(
 	 */
 	if (intf_count == 2)
 		topology.num_lm = 2;
-	else if (!dpu_kms->catalog->caps->has_3d_merge)
-		topology.num_lm = 1;
+	else if (dpu_kms->catalog->caps->has_3d_merge &&
+		 dpu_kms->catalog->mixer_count > 0 &&
+		 mode->hdisplay > dpu_kms->catalog->mixer[0].sblk->maxwidth)
+		topology.num_lm = 2;
 	else
-		topology.num_lm = (mode->hdisplay > MAX_HDISPLAY_SPLIT) ? 2 : 1;
+		topology.num_lm = 1;
 
 	if (crtc_state->ctm)
 		topology.num_dspp = topology.num_lm;

-- 
2.40.0

Re: [PATCH 02/11] drm/msm/dpu: use the actual lm maximum width instead of a hardcoded value

Posted by Dmitry Baryshkov 2 years, 8 months ago

On 19/04/2023 17:41, Arnaud Vrac wrote:
> This avoids using two LMs instead of one when the display width is lower
> than the maximum supported value. For example on MSM8996/MSM8998, the
> actual maxwidth is 2560, so we would use two LMs for 1280x720 or
> 1920x1080 resolutions, while one is enough.
> 
> Signed-off-by: Arnaud Vrac <avrac@freebox.fr>

While this looks correct (and following what we have in 4.4), later 
vendor kernels specify the topology explicitly. Probably we should check 
this with the hw guys, because it might be the following case: even 
though a single LM can supply the mode, it will spend more power 
compared to two LMs.


> ---
>   drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c | 10 +++++-----
>   1 file changed, 5 insertions(+), 5 deletions(-)
> 
> diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c
> index 1dc5dbe585723..dd2914726c4f6 100644
> --- a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c
> +++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c
> @@ -53,8 +53,6 @@
>   
>   #define IDLE_SHORT_TIMEOUT	1
>   
> -#define MAX_HDISPLAY_SPLIT 1080
> -
>   /* timeout in frames waiting for frame done */
>   #define DPU_ENCODER_FRAME_DONE_TIMEOUT_FRAMES 5
>   
> @@ -568,10 +566,12 @@ static struct msm_display_topology dpu_encoder_get_topology(
>   	 */
>   	if (intf_count == 2)
>   		topology.num_lm = 2;
> -	else if (!dpu_kms->catalog->caps->has_3d_merge)
> -		topology.num_lm = 1;
> +	else if (dpu_kms->catalog->caps->has_3d_merge &&
> +		 dpu_kms->catalog->mixer_count > 0 &&
> +		 mode->hdisplay > dpu_kms->catalog->mixer[0].sblk->maxwidth)
> +		topology.num_lm = 2;
>   	else
> -		topology.num_lm = (mode->hdisplay > MAX_HDISPLAY_SPLIT) ? 2 : 1;
> +		topology.num_lm = 1;
>   
>   	if (crtc_state->ctm)
>   		topology.num_dspp = topology.num_lm;
> 

-- 
With best wishes
Dmitry

Re: [Freedreno] [PATCH 02/11] drm/msm/dpu: use the actual lm maximum width instead of a hardcoded value

Posted by Jeykumar Sankaran 2 years, 8 months ago


On 4/19/2023 3:23 PM, Dmitry Baryshkov wrote:
> On 19/04/2023 17:41, Arnaud Vrac wrote:
>> This avoids using two LMs instead of one when the display width is lower
>> than the maximum supported value. For example on MSM8996/MSM8998, the
>> actual maxwidth is 2560, so we would use two LMs for 1280x720 or
>> 1920x1080 resolutions, while one is enough.
>>
>> Signed-off-by: Arnaud Vrac <avrac@freebox.fr>
> 
> While this looks correct (and following what we have in 4.4), later 
> vendor kernels specify the topology explicitly. Probably we should check 
> this with the hw guys, because it might be the following case: even 
> though a single LM can supply the mode, it will spend more power 
> compared to two LMs.
> 
> 
Yes. 2 LM split will allow the HW to run in lower mdp core clock. Can 
you maintain the split_threshold in the hw catalog until per mode 
topology is available?

Jeykumar S
>> ---
>>   drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c | 10 +++++-----
>>   1 file changed, 5 insertions(+), 5 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c 
>> b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c
>> index 1dc5dbe585723..dd2914726c4f6 100644
>> --- a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c
>> +++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c
>> @@ -53,8 +53,6 @@
>>   #define IDLE_SHORT_TIMEOUT    1
>> -#define MAX_HDISPLAY_SPLIT 1080
>> -
>>   /* timeout in frames waiting for frame done */
>>   #define DPU_ENCODER_FRAME_DONE_TIMEOUT_FRAMES 5
>> @@ -568,10 +566,12 @@ static struct msm_display_topology 
>> dpu_encoder_get_topology(
>>        */
>>       if (intf_count == 2)
>>           topology.num_lm = 2;
>> -    else if (!dpu_kms->catalog->caps->has_3d_merge)
>> -        topology.num_lm = 1;
>> +    else if (dpu_kms->catalog->caps->has_3d_merge &&
>> +         dpu_kms->catalog->mixer_count > 0 &&
>> +         mode->hdisplay > dpu_kms->catalog->mixer[0].sblk->maxwidth)
>> +        topology.num_lm = 2;
>>       else
>> -        topology.num_lm = (mode->hdisplay > MAX_HDISPLAY_SPLIT) ? 2 : 1;
>> +        topology.num_lm = 1;
>>       if (crtc_state->ctm)
>>           topology.num_dspp = topology.num_lm;
>>
>

Re: [Freedreno] [PATCH 02/11] drm/msm/dpu: use the actual lm maximum width instead of a hardcoded value

Posted by Dmitry Baryshkov 2 years, 7 months ago

On 20/04/2023 20:47, Jeykumar Sankaran wrote:
> 
> 
> On 4/19/2023 3:23 PM, Dmitry Baryshkov wrote:
>> On 19/04/2023 17:41, Arnaud Vrac wrote:
>>> This avoids using two LMs instead of one when the display width is lower
>>> than the maximum supported value. For example on MSM8996/MSM8998, the
>>> actual maxwidth is 2560, so we would use two LMs for 1280x720 or
>>> 1920x1080 resolutions, while one is enough.
>>>
>>> Signed-off-by: Arnaud Vrac <avrac@freebox.fr>
>>
>> While this looks correct (and following what we have in 4.4), later 
>> vendor kernels specify the topology explicitly. Probably we should 
>> check this with the hw guys, because it might be the following case: 
>> even though a single LM can supply the mode, it will spend more power 
>> compared to two LMs.
>>
>>
> Yes. 2 LM split will allow the HW to run in lower mdp core clock. Can 
> you maintain the split_threshold in the hw catalog until per mode 
> topology is available?

I don't think it warrants the trouble, unless we have a real usecase 
when the device is short of LMs.

Arnaud, I'll mark this patch as Rejected for now, unless it fixes an LM 
shortage for your platform.

> 
> Jeykumar S
>>> ---
>>>   drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c | 10 +++++-----
>>>   1 file changed, 5 insertions(+), 5 deletions(-)
>>>
>>> diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c 
>>> b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c
>>> index 1dc5dbe585723..dd2914726c4f6 100644
>>> --- a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c
>>> +++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c
>>> @@ -53,8 +53,6 @@
>>>   #define IDLE_SHORT_TIMEOUT    1
>>> -#define MAX_HDISPLAY_SPLIT 1080
>>> -
>>>   /* timeout in frames waiting for frame done */
>>>   #define DPU_ENCODER_FRAME_DONE_TIMEOUT_FRAMES 5
>>> @@ -568,10 +566,12 @@ static struct msm_display_topology 
>>> dpu_encoder_get_topology(
>>>        */
>>>       if (intf_count == 2)
>>>           topology.num_lm = 2;
>>> -    else if (!dpu_kms->catalog->caps->has_3d_merge)
>>> -        topology.num_lm = 1;
>>> +    else if (dpu_kms->catalog->caps->has_3d_merge &&
>>> +         dpu_kms->catalog->mixer_count > 0 &&
>>> +         mode->hdisplay > dpu_kms->catalog->mixer[0].sblk->maxwidth)
>>> +        topology.num_lm = 2;
>>>       else
>>> -        topology.num_lm = (mode->hdisplay > MAX_HDISPLAY_SPLIT) ? 2 
>>> : 1;
>>> +        topology.num_lm = 1;
>>>       if (crtc_state->ctm)
>>>           topology.num_dspp = topology.num_lm;
>>>
>>

-- 
With best wishes
Dmitry

Re: [Freedreno] [PATCH 02/11] drm/msm/dpu: use the actual lm maximum width instead of a hardcoded value

Posted by Arnaud Vrac 2 years, 6 months ago

Le sam. 20 mai 2023 à 22:49, Dmitry Baryshkov
<dmitry.baryshkov@linaro.org> a écrit :
>
> On 20/04/2023 20:47, Jeykumar Sankaran wrote:
> >
> >
> > On 4/19/2023 3:23 PM, Dmitry Baryshkov wrote:
> >> On 19/04/2023 17:41, Arnaud Vrac wrote:
> >>> This avoids using two LMs instead of one when the display width is lower
> >>> than the maximum supported value. For example on MSM8996/MSM8998, the
> >>> actual maxwidth is 2560, so we would use two LMs for 1280x720 or
> >>> 1920x1080 resolutions, while one is enough.
> >>>
> >>> Signed-off-by: Arnaud Vrac <avrac@freebox.fr>
> >>
> >> While this looks correct (and following what we have in 4.4), later
> >> vendor kernels specify the topology explicitly. Probably we should
> >> check this with the hw guys, because it might be the following case:
> >> even though a single LM can supply the mode, it will spend more power
> >> compared to two LMs.
> >>
> >>
> > Yes. 2 LM split will allow the HW to run in lower mdp core clock. Can
> > you maintain the split_threshold in the hw catalog until per mode
> > topology is available?
>
> I don't think it warrants the trouble, unless we have a real usecase
> when the device is short of LMs.
>
> Arnaud, I'll mark this patch as Rejected for now, unless it fixes an LM
> shortage for your platform.

It's fine, if I remember correctly I wrote this patch because display
wouldn't work before I fixed the LM pairings on msm8998, but now it's
not a requirement anymore.

>
> >
> > Jeykumar S
> >>> ---
> >>>   drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c | 10 +++++-----
> >>>   1 file changed, 5 insertions(+), 5 deletions(-)
> >>>
> >>> diff --git a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c
> >>> b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c
> >>> index 1dc5dbe585723..dd2914726c4f6 100644
> >>> --- a/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c
> >>> +++ b/drivers/gpu/drm/msm/disp/dpu1/dpu_encoder.c
> >>> @@ -53,8 +53,6 @@
> >>>   #define IDLE_SHORT_TIMEOUT    1
> >>> -#define MAX_HDISPLAY_SPLIT 1080
> >>> -
> >>>   /* timeout in frames waiting for frame done */
> >>>   #define DPU_ENCODER_FRAME_DONE_TIMEOUT_FRAMES 5
> >>> @@ -568,10 +566,12 @@ static struct msm_display_topology
> >>> dpu_encoder_get_topology(
> >>>        */
> >>>       if (intf_count == 2)
> >>>           topology.num_lm = 2;
> >>> -    else if (!dpu_kms->catalog->caps->has_3d_merge)
> >>> -        topology.num_lm = 1;
> >>> +    else if (dpu_kms->catalog->caps->has_3d_merge &&
> >>> +         dpu_kms->catalog->mixer_count > 0 &&
> >>> +         mode->hdisplay > dpu_kms->catalog->mixer[0].sblk->maxwidth)
> >>> +        topology.num_lm = 2;
> >>>       else
> >>> -        topology.num_lm = (mode->hdisplay > MAX_HDISPLAY_SPLIT) ? 2
> >>> : 1;
> >>> +        topology.num_lm = 1;
> >>>       if (crtc_state->ctm)
> >>>           topology.num_dspp = topology.num_lm;
> >>>
> >>
>
> --
> With best wishes
> Dmitry
>