[Intel-gfx] [RFC 0/5] Per client engine busyness (all aboard the sysfs train!)

Tvrtko Ursulin tvrtko.ursulin at linux.intel.com
Fri Oct 25 14:21:26 UTC 2019


From: Tvrtko Ursulin <tvrtko.ursulin at intel.com>

It was quite some time since I last posted this RFC, but recently there has been
some new interest, this time from OpenCL and related customers, so I decided to
give it a quick respin and test the waters.

This time round it has been really hastily rebased since the upstream changed
quite a lot and I have very little confidence it is technically correct. But it
is enough to illustrate a point of what this feature could provide:

In short it enables a "top-like" display for GPU tasks. Or with a screenshot:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
intel-gpu-top -  948/ 999 MHz;    0% RC6;  3.65 Watts;     2165 irqs/s

      IMC reads:     5015 MiB/s
     IMC writes:      143 MiB/s

          ENGINE      BUSY                                                    MI_SEMA MI_WAIT
     Render/3D/0   56.60% |███████████████████████████▋                     |      0%      0%
       Blitter/0   95.65% |██████████████████████████████████████████████▊  |      0%      0%
         Video/0   40.92% |████████████████████                             |      0%      0%
  VideoEnhance/0    0.00% |                                                 |      0%      0%

  PID            NAME       RCS               BCS               VCS               VECS
 5347        gem_wsim |███████▍        ||███████████████▏||██████▌         ||                |
 4929            Xorg |▎               ||                ||                ||                |
 5305        glxgears |                ||                ||                ||                |
 5303        glxgears |                ||                ||                ||                |
 5024           xfwm4 |                ||                ||                ||                |
 4929            Xorg |                ||                ||                ||                |
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Implementation wise we would get a a bunch of per-drm-client-per-engine-class
files in sysfs like:

	# cd /sys/class/drm/card0/clients/
	# tree
	.
	├── 7
	│   ├── busy
	│   │   ├── 0
	│   │   ├── 1
	│   │   ├── 2
	│   │   └── 3
	│   ├── name
	│   └── pid
	├── 8
	│   ├── busy
	│   │   ├── 0
	│   │   ├── 1
	│   │   ├── 2
	│   │   └── 3
	│   ├── name
	│   └── pid
	├── 9
	│   ├── busy
	│   │   ├── 0
	│   │   ├── 1
	│   │   ├── 2
	│   │   └── 3
	│   ├── name
	│   └── pid
	└── enable_stats

I will post the corresponding patch to intel_gpu_top for reference as well.

Tvrtko Ursulin (5):
  drm/i915: Track per-context engine busyness
  drm/i915: Expose list of clients in sysfs
  drm/i915: Update client name on context create
  drm/i915: Expose per-engine client busyness
  drm/i915: Add sysfs toggle to enable per-client engine stats

 drivers/gpu/drm/i915/gem/i915_gem_context.c   |  17 +-
 drivers/gpu/drm/i915/gt/intel_context.c       |  20 ++
 drivers/gpu/drm/i915/gt/intel_context.h       |   9 +
 drivers/gpu/drm/i915/gt/intel_context_types.h |   9 +
 drivers/gpu/drm/i915/gt/intel_engine_cs.c     |  16 +-
 drivers/gpu/drm/i915/gt/intel_lrc.c           |  66 +++++-
 drivers/gpu/drm/i915/i915_drv.h               |  38 +++
 drivers/gpu/drm/i915/i915_gem.c               | 218 +++++++++++++++++-
 drivers/gpu/drm/i915/i915_sysfs.c             |  81 +++++++
 9 files changed, 451 insertions(+), 23 deletions(-)

-- 
2.20.1



More information about the Intel-gfx mailing list