1.. SPDX-License-Identifier: GPL-2.0+ 2 3========================= 4 drm/Panthor CSF driver 5========================= 6 7.. _panthor-usage-stats: 8 9Panthor DRM client usage stats implementation 10============================================== 11 12The drm/Panthor driver implements the DRM client usage stats specification as 13documented in :ref:`drm-client-usage-stats`. 14 15Example of the output showing the implemented key value pairs and entirety of 16the currently possible format options: 17 18:: 19 pos: 0 20 flags: 02400002 21 mnt_id: 29 22 ino: 491 23 drm-driver: panthor 24 drm-client-id: 10 25 drm-engine-panthor: 111110952750 ns 26 drm-cycles-panthor: 94439687187 27 drm-maxfreq-panthor: 1000000000 Hz 28 drm-curfreq-panthor: 1000000000 Hz 29 panthor-resident-memory: 10396 KiB 30 panthor-active-memory: 10396 KiB 31 drm-total-memory: 16480 KiB 32 drm-shared-memory: 0 33 drm-active-memory: 16200 KiB 34 drm-resident-memory: 16480 KiB 35 drm-purgeable-memory: 0 36 37Possible `drm-engine-` key names are: `panthor`. 38`drm-curfreq-` values convey the current operating frequency for that engine. 39 40Users must bear in mind that engine and cycle sampling are disabled by default, 41because of power saving concerns. `fdinfo` users and benchmark applications which 42query the fdinfo file must make sure to toggle the job profiling status of the 43driver by writing into the appropriate sysfs node:: 44 45 echo <N> > /sys/bus/platform/drivers/panthor/[a-f0-9]*.gpu/profiling 46 47Where `N` is a bit mask where cycle and timestamp sampling are respectively 48enabled by the first and second bits. 49 50Possible `panthor-*-memory` keys are: `active` and `resident`. 51These values convey the sizes of the internal driver-owned shmem BO's that 52aren't exposed to user-space through a DRM handle, like queue ring buffers, 53sync object arrays and heap chunks. Because they are all allocated and pinned 54at creation time, only `panthor-resident-memory` is necessary to tell us their 55size. `panthor-active-memory` shows the size of kernel BO's associated with 56VM's and groups currently being scheduled for execution by the GPU. 57