| Message ID | 20191216120704.958-1-tvrtko.ursulin@linux.intel.com (mailing list archive) |
|---|---|
| Headers | show |
| Series | Per client engine busyness | expand |
Quoting Tvrtko Ursulin (2019-12-16 12:06:59) > Implementation wise we add a a bunch of files in sysfs like: > > # cd /sys/class/drm/card0/clients/ > # tree > . > ├── 7 > │ ├── busy > │ │ ├── 0 Prefer '0' over rcs? > I will post the corresponding patch to intel_gpu_top for reference as well. The other requirement is that we need to at least prove the sysfs interface exists in gt. perf_sysfs? Quick list, - check igt_spin_t responses (pretty much verbatim of perf_pmu.c) - check the client name is correct around fd passing - check interactions with ctx->engines[] -Chris
On 16/12/2019 13:09, Chris Wilson wrote: > Quoting Tvrtko Ursulin (2019-12-16 12:06:59) >> Implementation wise we add a a bunch of files in sysfs like: >> >> # cd /sys/class/drm/card0/clients/ >> # tree >> . >> ├── 7 >> │ ├── busy >> │ │ ├── 0 > > Prefer '0' over rcs? I think so, saves userspace keeping a map of names to class enum. Or maybe it doesn't, depends. Saves us having to come up with ABI names. But I think I could be easily convinced either way. >> I will post the corresponding patch to intel_gpu_top for reference as well. > > The other requirement is that we need to at least prove the sysfs > interface exists in gt. perf_sysfs? > > Quick list, > - check igt_spin_t responses (pretty much verbatim of perf_pmu.c) > - check the client name is correct around fd passing > - check interactions with ctx->engines[] Yep, I know it will be needed. But haven't been bothering yet since the series has been in a hopeless mode for what, two years or so. I forgot to name it RFC this time round.. :) Regards, Tvrtko
From: Tvrtko Ursulin <tvrtko.ursulin@intel.com> Another re-spin of the per-client engine busyness series. Review feedback from last round has been addressed* and the tracking simplified. (*Apart from re-using the ctx->idr_lock for the global toggle, I kept using struct mutext for that.) Internally we track time spent on engines for each struct intel_context. This can serve as a building block for several features from the want list: smarter scheduler decisions, getrusage(2)-like per-GEM-context functionality wanted by some customers, cgroups controller, dynamic SSEU tuning,... Externally, in sysfs, we expose time spent on GPU per client and per engine class. There is also a global toggle to enable this extra tracking although it is open whether it is warranted and we should not just always track. Sysfs interface enables us to implement a "top-like" tool for GPU tasks. Or with a "screenshot": ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ intel-gpu-top - 906/ 955 MHz; 0% RC6; 5.30 Watts; 933 irqs/s IMC reads: 4414 MiB/s IMC writes: 3805 MiB/s ENGINE BUSY MI_SEMA MI_WAIT Render/3D/0 93.46% |████████████████████████████████▋ | 0% 0% Blitter/0 0.00% | | 0% 0% Video/0 0.00% | | 0% 0% VideoEnhance/0 0.00% | | 0% 0% PID NAME Render/3D Blitter Video VideoEnhance 2733 neverball |██████▌ || || || | 2047 Xorg |███▊ || || || | 2737 glxgears |█▍ || || || | 2128 xfwm4 | || || || | 2047 Xorg | || || || | ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ Implementation wise we add a a bunch of 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 Files in 'busy' directories are numbered using the engine class ABI values and they contain accumulated nanoseconds each client spent on engines of a respective class. 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 | 24 +- drivers/gpu/drm/i915/gt/intel_context.c | 20 ++ drivers/gpu/drm/i915/gt/intel_context.h | 11 + 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 | 47 +++- drivers/gpu/drm/i915/i915_drv.h | 41 +++ drivers/gpu/drm/i915/i915_gem.c | 234 +++++++++++++++++- drivers/gpu/drm/i915/i915_sysfs.c | 84 +++++++ 9 files changed, 465 insertions(+), 21 deletions(-)