[PATCH 0/6] Prefer wbinvd() where appropriate

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



A while back we had a workload which severely suffered from excessive
clflushing. I sent out a patch series which solved this generically in DRM core,
but that had the unfortunate side effect of possibly regressing non-intel
platforms.

I've re-spun the series to only take this shortcut when on an Intel platform.
This series is untested by me since we've now worked around the problem in a
different way within mesa. Jesse asked me to send this out, though I don't have
time to see if it's still useful (until my mesa patches are merged, it would be
useful, but afterwards, I do not know). I also didn't check that I rebased
things properly.

Do what you want with them...


Ben Widawsky (6):
  drm/i915: Remove the useless flush_chipset
  drm/i915: Pass eb_vmas to execbuffer implementations
  drm/i915: Opportunistically reduce flushing at execbuf
  drm/i915: Add debugfs knobs for wbinvd threshold
  drm/i915: Extract checking the necessity of flush
  drm/i915: obey wbinvd threshold in more places

 drivers/gpu/drm/i915/i915_debugfs.c        | 34 +++++++++++
 drivers/gpu/drm/i915/i915_drv.h            | 25 +++++++-
 drivers/gpu/drm/i915/i915_gem.c            | 96 ++++++++++++++++++++----------
 drivers/gpu/drm/i915/i915_gem_execbuffer.c | 46 +++++++-------
 drivers/gpu/drm/i915/i915_gem_gtt.c        | 13 +++-
 drivers/gpu/drm/i915/intel_lrc.c           | 21 ++++---
 drivers/gpu/drm/i915/intel_lrc.h           |  3 +-
 7 files changed, 172 insertions(+), 66 deletions(-)

-- 
2.3.0

_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
http://lists.freedesktop.org/mailman/listinfo/intel-gfx





[Index of Archives]     [Linux USB Devel]     [Linux Audio Users]     [Yosemite News]     [Linux Kernel]     [Linux SCSI]
  Powered by Linux