[PATCH v2 0/2] Fix serialization on burst of unbinds - v2

[Matthew Brost <matthew.brost@intel.com>]

Attempting to resolve part of [1]; this solution differs than v1 [2] by
changing last fence semantics detailed in the patches.

Overview of issue in [1]:

When a burst of unbind jobs is issued, a dependency chain can form
between the TLB invalidation of a previous unbind job and the current
one. This leads to undesirable serialization, causing current jobs to
wait unnecessarily for prior TLB invalidations, execute on the GPU when
not needed, and significantly slow down the unbind burst—resulting in up
to a 4× slowdown.
 
Matt

[1] https://gitlab.freedesktop.org/drm/xe/kernel/-/issues/6047
[2] https://patchwork.freedesktop.org/series/156144/

Matthew Brost (2):
  drm/xe: Add last fence attachment to TLB invalidation job queues
  drm/xe: Decouple bind queue last fence from TLB invalidations

 drivers/gpu/drm/xe/xe_exec_queue.c       | 100 +++++++++++++++++++++++
 drivers/gpu/drm/xe/xe_exec_queue.h       |  18 ++++
 drivers/gpu/drm/xe/xe_exec_queue_types.h |   5 ++
 drivers/gpu/drm/xe/xe_pt.c               |  64 +++++----------
 drivers/gpu/drm/xe/xe_sync.c             |  62 ++++++++++++--
 drivers/gpu/drm/xe/xe_vm.c               |  60 ++++++++------
 6 files changed, 232 insertions(+), 77 deletions(-)

-- 
2.34.1