[PATCH V9] drm/xe/uapi: Use hint for guc to set GT frequency

[Lucas De Marchi]

On Tue, Feb 25, 2025 at 04:39:25AM +0000, Upadhyay, Tejas wrote:
>
>
>> -----Original Message-----
>> From: De Marchi, Lucas <lucas.demarchi at intel.com>
>> Sent: Tuesday, February 25, 2025 12:42 AM
>> To: Upadhyay, Tejas <tejas.upadhyay at intel.com>
>> Cc: intel-xe at lists.freedesktop.org; dri-devel at lists.freedesktop.org;
>> Belgaumkar, Vinay <vinay.belgaumkar at intel.com>
>> Subject: Re: [PATCH V9] drm/xe/uapi: Use hint for guc to set GT frequency
>>
>> On Mon, Feb 24, 2025 at 09:43:09AM -0600, Lucas De Marchi wrote:
>> >On Wed, Feb 12, 2025 at 05:08:30PM +0530, Tejas Upadhyay wrote:
>> >>Allow user to provide a low latency hint. When set, KMD sends a hint
>> >>to GuC which results in special handling for that process. SLPC will
>> >>ramp the GT frequency aggressively every time it switches to this
>> >>process.
>> >>
>> >>We need to enable the use of SLPC Compute strategy during init, but it
>> >>will apply only to processes that set this bit during process
>> >>creation.
>> >>
>> >>Improvement with this approach as below:
>> >>
>> >>Before,
>> >>
>> >>:~$ NEOReadDebugKeys=1 EnableDirectSubmission=0 clpeak
>> >>--kernel-latency
>> >>Platform: Intel(R) OpenCL Graphics
>> >> Device: Intel(R) Graphics [0xe20b]
>> >>   Driver version  : 24.52.0 (Linux x64)
>> >>   Compute units   : 160
>> >>   Clock frequency : 2850 MHz
>> >>   Kernel launch latency : 283.16 us
>> >>
>> >>After,
>> >>
>> >>:~$ NEOReadDebugKeys=1 EnableDirectSubmission=0 clpeak
>> >>--kernel-latency
>> >>Platform: Intel(R) OpenCL Graphics
>> >> Device: Intel(R) Graphics [0xe20b]
>> >>   Driver version  : 24.52.0 (Linux x64)
>> >>   Compute units   : 160
>> >>   Clock frequency : 2850 MHz
>> >>
>> >>   Kernel launch latency : 63.38 us
>> >>
>> >>UMD Compute PR : https://github.com/intel/compute-runtime/pull/794
>> >>UMD Mesa PR :
>> >>https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/33214
>> >>
>> >>v9(Vinay):
>> >> - remove extra line, align commit message
>> >>v8(Vinay):
>> >> - Add separate example for using low latency hint
>> >>v7(Jose):
>> >> - Update UMD PR
>> >> - applicable to all gpus
>> >>V6:
>> >> - init flags, remove redundant flags check (MAuld)
>> >>V5:
>> >> - Move uapi doc to documentation and GuC ABI specific change
>> >>(Rodrigo)
>> >
>> >hmn... that doesn't look right.
>> >
>> >...
>> >
>> >>diff --git a/Documentation/gpu/drm-uapi.rst
>> >>b/Documentation/gpu/drm-uapi.rst index b75cc9a70d1f..7337d1be45ef
>> >>100644
>> >>--- a/Documentation/gpu/drm-uapi.rst
>> >>+++ b/Documentation/gpu/drm-uapi.rst
>> >>@@ -583,3 +583,21 @@ dma-buf interoperability
>> >>
>> >>Please see Documentation/userspace-api/dma-buf-alloc-exchange.rst for
>> >>information on how dma-buf is integrated and exposed within DRM.
>> >>+
>> >>+Low latency hint by user
>> >>+========================
>> >>+
>> >>+Allow users to provide a hint to kernel for cases demanding low
>> >>+latency profile. Please note it will have impact on power
>> >>+consumption. User can indicate low latency hint with flag while
>> >>+creating exec queue as mentioned below,
>> >>+
>> >>+     struct drm_xe_exec_queue_create exec_queue_create = {
>> >>+          .flags = DRM_XE_EXEC_QUEUE_LOW_LATENCY_HINT,
>> >>+          .extensions = 0,
>> >>+          .vm_id = vm,
>> >>+          .num_bb_per_exec = 1,
>> >>+          .num_eng_per_bb = 1,
>> >>+          .instances = to_user_pointer(&instance),
>> >>+     };
>> >>+     ioctl(fd, DRM_IOCTL_XE_EXEC_QUEUE_CREATE, &exec_queue_create);
>> >
>> >how does this driver-specific doc make sense in this file?
>
>Thanks for looking at this.
>
>There was suggestion to add this in documentation during review, @Vivi, Rodrigo Can we remove this? And if we remove this whats point in keeping documentation here for 4 line paragraph?
>@De Marchi, Lucas Do you mean completely remove from here, just keep in xe_drm.h?

AFAICS in v4 the request was to add it to the documentation, not to add
it to a random documentation file. In v4 this example was only in the
commit message rather than in the documentation.

If you add it as kernel-doc you are adding it to the documentation. See
https://docs.kernel.org/gpu/driver-uapi.html#c.drm_xe_exec_queue_create
			    ^^^^^^^^^^^

The xe uapi shows up here:
https://docs.kernel.org/gpu/driver-uapi.html#drm-xe-uapi because:

Documentation/gpu/index.rst:
	...
	driver-uapi
	...

Documentation/gpu/driver-uapi:
	
	drm/xe uAPI
	===========

	.. kernel-doc:: include/uapi/drm/xe_drm.h

In v9 this example is duplicated in Documentation/gpu/drm-uapi.rst
(wrong file) and xe_drm.h (right one).

I hope that clarifies.

Lucas De Marchi

>
>Tejas
>
>>
>> to avoid needless noise in the log, I zapped this commit from
>> drm-xe-next:
>>
>>   + 8f78c0dfbe9d...8b4b3af869e9 drm-xe-next -> drm-xe-next (forced
>> update)
>>
>> 8b4b3af o [drm-xe-next] {drm-xe/drm-xe-next} {drm-xe/HEAD}
>> drm/xe/userptr: remove tmp_evict list
>> 6b93cb9 o drm/xe/userptr: fix EFAULT handling
>> 4e37e92 o drm/xe/userptr: restore invalidation list on error 8f78c0d │ o
>> drm/xe/userptr: remove tmp_evict list f2211aa │ o drm/xe/userptr: fix EFAULT
>> handling b9aaabb │ o drm/xe/userptr: restore invalidation list on error
>> bee37a8 │ o drm/xe/uapi: Use hint for guc to set GT frequency
>>
>> this documentation is also in include/uapi/drm/xe_drm.h, so maybe the best
>> thing to do is probably to just drop this part that got included in the wrong
>> part. Once this is fixed we can add it back.
>>
>> Lucas De Marchi