[PATCH 1/5] drm/sched:add new priority level

Das, Nirmoy nirmoy.das at amd.com
Tue Aug 24 13:23:59 UTC 2021


On 8/24/2021 3:18 PM, Christian König wrote:
> Am 24.08.21 um 14:39 schrieb Das, Nirmoy:
>>
>> On 8/24/2021 2:07 PM, Christian König wrote:
>>> Am 24.08.21 um 13:57 schrieb Das, Nirmoy:
>>>> Hi Christian,
>>>>
>>>> On 8/24/2021 8:10 AM, Christian König wrote:
>>>>> I haven't followed the previous discussion, but that looks like 
>>>>> this change is based on a misunderstanding.
>>>>
>>>>
>>>> In previous discussion I sort of suggested to have new DRM prio as 
>>>> I didn't see any other way to map priority provided by the 
>>>> userspace to this new 3rd hw priority.
>>>>
>>>> Do you think we should use other information from userspace like 
>>>> queue id to determine hardware priority ?
>>>
>>> If I'm not completely mistaken we have dropped the concept of 
>>> exposing multiple queues/instances completely.
>>
>>
>> Yes, that is my understanding too.
>>
>>>
>>> What we should probably do is to use the (cleaned up) UAPI enum for 
>>> init_priority and override_priority instead of the drm scheduler enums.
>>
>>
>> I went through the drm code, now I see what you mean. So what we are 
>> now doing is:  mapping  AMDGPU_CTX_PRIORITY_*  to 
>> DRM_SCHED_PRIORITY_*  and then to hw priority which is not nice.
>>
>> We should rather map AMDGPU_CTX_PRIORITY_* to hw priority directly.
>
> Exactly that's my idea, yes.
>
> If you want feel free to put this on your TODO for a potential cleanup.


Yes Sure, I will try have a patch ready asap so that Satyajit can work 
on top of that.


Regards,

Nirmoy

>
> Christian.
>
>>
>>
>> Regards,
>>
>> Nirmoy
>>
>>
>>>
>>> Regards,
>>> Christian.
>>>
>>>>
>>>>
>>>> Regards,
>>>>
>>>> Nirmoy
>>>>
>>>>>
>>>>> Those here are the software priorities used in the scheduler, but 
>>>>> what you are working on are the hardware priorities.
>>>>>
>>>>> That are two completely different things which we shouldn't mix up.
>>>>>
>>>>> Regards,
>>>>> Christian.
>>>>>
>>>>> Am 24.08.21 um 07:55 schrieb Satyajit Sahu:
>>>>>> Adding a new priority level DRM_SCHED_PRIORITY_VERY_HIGH
>>>>>>
>>>>>> Signed-off-by: Satyajit Sahu <satyajit.sahu at amd.com>
>>>>>> ---
>>>>>>   include/drm/gpu_scheduler.h | 1 +
>>>>>>   1 file changed, 1 insertion(+)
>>>>>>
>>>>>> diff --git a/include/drm/gpu_scheduler.h 
>>>>>> b/include/drm/gpu_scheduler.h
>>>>>> index d18af49fd009..d0e5e234da5f 100644
>>>>>> --- a/include/drm/gpu_scheduler.h
>>>>>> +++ b/include/drm/gpu_scheduler.h
>>>>>> @@ -40,6 +40,7 @@ enum drm_sched_priority {
>>>>>>       DRM_SCHED_PRIORITY_MIN,
>>>>>>       DRM_SCHED_PRIORITY_NORMAL,
>>>>>>       DRM_SCHED_PRIORITY_HIGH,
>>>>>> +    DRM_SCHED_PRIORITY_VERY_HIGH,
>>>>>>       DRM_SCHED_PRIORITY_KERNEL,
>>>>>>         DRM_SCHED_PRIORITY_COUNT,
>>>>>
>>>
>


More information about the amd-gfx mailing list