1. 22 May, 2019 6 commits
    • Chris Wilson's avatar
      drm/i915: Load balancing across a virtual engine · 6d06779e
      Chris Wilson authored
      Having allowed the user to define a set of engines that they will want
      to only use, we go one step further and allow them to bind those engines
      into a single virtual instance. Submitting a batch to the virtual engine
      will then forward it to any one of the set in a manner as best to
      distribute load.  The virtual engine has a single timeline across all
      engines (it operates as a single queue), so it is not able to concurrently
      run batches across multiple engines by itself; that is left up to the user
      to submit multiple concurrent batches to multiple queues. Multiple users
      will be load balanced across the system.
      
      The mechanism used for load balancing in this patch is a late greedy
      balancer. When a request is ready for execution, it is added to each
      engine's queue, and when an engine is ready for its next request it
      claims it from the virtual engine. The first engine to do so, wins, i.e.
      the request is executed at the earliest opportunity (idle moment) in the
      system.
      
      As not all HW is created equal, the user is still able to skip the
      virtual engine and execute the batch on a specific engine, all within the
      same queue. It will then be executed in order on the correct engine,
      with execution on other virtual engines being moved away due to the load
      detection.
      
      A couple of areas for potential improvement left!
      
      - The virtual engine always take priority over equal-priority tasks.
      Mostly broken up by applying FQ_CODEL rules for prioritising new clients,
      and hopefully the virtual and real engines are not then congested (i.e.
      all work is via virtual engines, or all work is to the real engine).
      
      - We require the breadcrumb irq around every virtual engine request. For
      normal engines, we eliminate the need for the slow round trip via
      interrupt by using the submit fence and queueing in order. For virtual
      engines, we have to allow any job to transfer to a new ring, and cannot
      coalesce the submissions, so require the completion fence instead,
      forcing the persistent use of interrupts.
      
      - We only drip feed single requests through each virtual engine and onto
      the physical engines, even if there was enough work to fill all ELSP,
      leaving small stalls with an idle CS event at the end of every request.
      Could we be greedy and fill both slots? Being lazy is virtuous for load
      distribution on less-than-full workloads though.
      
      Other areas of improvement are more general, such as reducing lock
      contention, reducing dispatch overhead, looking at direct submission
      rather than bouncing around tasklets etc.
      
      sseu: Lift the restriction to allow sseu to be reconfigured on virtual
      engines composed of RENDER_CLASS (rcs).
      
      v2: macroize check_user_mbz()
      v3: Cancel virtual engines on wedging
      v4: Commence commenting
      v5: Replace 64b sibling_mask with a list of class:instance
      v6: Drop the one-element array in the uabi
      v7: Assert it is an virtual engine in to_virtual_engine()
      v8: Skip over holes in [class][inst] so we can selftest with (vcs0, vcs2)
      
      Link: https://github.com/intel/media-driver/pull/283Signed-off-by: default avatarChris Wilson <chris@chris-wilson.co.uk>
      Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
      Reviewed-by: default avatarTvrtko Ursulin <tvrtko.ursulin@intel.com>
      Link: https://patchwork.freedesktop.org/patch/msgid/20190521211134.16117-6-chris@chris-wilson.co.uk
      6d06779e
    • Chris Wilson's avatar
      drm/i915: Allow userspace to clone contexts on creation · b81dde71
      Chris Wilson authored
      A usecase arose out of handling context recovery in mesa, whereby they
      wish to recreate a context with fresh logical state but preserving all
      other details of the original. Currently, they create a new context and
      iterate over which bits they want to copy across, but it would much more
      convenient if they were able to just pass in a target context to clone
      during creation. This essentially extends the setparam during creation
      to pull the details from a target context instead of the user supplied
      parameters.
      
      The ideal here is that we don't expose control over anything more than
      can be obtained via CONTEXT_PARAM. That is userspace retains explicit
      control over all features, and this api is just convenience.
      
      For example, you could replace
      
      	struct context_param p = { .param = CONTEXT_PARAM_VM };
      
      	param.ctx_id = old_id;
      	gem_context_get_param(&p.param);
      
      	new_id = gem_context_create();
      
      	param.ctx_id = new_id;
      	gem_context_set_param(&p.param);
      
      	gem_vm_destroy(param.value); /* drop the ref to VM_ID handle */
      
      with
      
      	struct create_ext_param p = {
      	  { .name = CONTEXT_CREATE_CLONE },
      	  .clone_id = old_id,
      	  .flags = CLONE_FLAGS_VM
      	}
      	new_id = gem_context_create_ext(&p);
      
      and not have to worry about stray namespace pollution etc.
      Signed-off-by: default avatarChris Wilson <chris@chris-wilson.co.uk>
      Reviewed-by: default avatarTvrtko Ursulin <tvrtko.ursulin@intel.com>
      Link: https://patchwork.freedesktop.org/patch/msgid/20190521211134.16117-5-chris@chris-wilson.co.uk
      b81dde71
    • Chris Wilson's avatar
      drm/i915: Re-expose SINGLE_TIMELINE flags for context creation · 8319f44c
      Chris Wilson authored
      The SINGLE_TIMELINE flag can be used to create a context such that all
      engine instances within that context share a common timeline. This can
      be useful for mixing operations between real and virtual engines, or
      when using a composite context for a single client API context.
      Signed-off-by: default avatarChris Wilson <chris@chris-wilson.co.uk>
      Reviewed-by: default avatarTvrtko Ursulin <tvrtko.ursulin@intel.com>
      Link: https://patchwork.freedesktop.org/patch/msgid/20190521211134.16117-4-chris@chris-wilson.co.uk
      8319f44c
    • Chris Wilson's avatar
      drm/i915: Extend I915_CONTEXT_PARAM_SSEU to support local ctx->engine[] · e620f7b3
      Chris Wilson authored
      Allow the user to specify a local engine index (as opposed to
      class:index) that they can use to refer to a preset engine inside the
      ctx->engine[] array defined by an earlier I915_CONTEXT_PARAM_ENGINES.
      This will be useful for setting SSEU parameters on virtual engines that
      are local to the context and do not have a valid global class:instance
      lookup.
      
      Note that due to the ambiguity in using class:instance with
      ctx->engines[], if a user supplied engine map is active the user must
      specify the engine to alter by its index into the ctx->engines[].
      Signed-off-by: default avatarChris Wilson <chris@chris-wilson.co.uk>
      Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
      Reviewed-by: default avatarTvrtko Ursulin <tvrtko.ursulin@intel.com>
      Link: https://patchwork.freedesktop.org/patch/msgid/20190521211134.16117-3-chris@chris-wilson.co.uk
      e620f7b3
    • Chris Wilson's avatar
      drm/i915: Allow a context to define its set of engines · 976b55f0
      Chris Wilson authored
      Over the last few years, we have debated how to extend the user API to
      support an increase in the number of engines, that may be sparse and
      even be heterogeneous within a class (not all video decoders created
      equal). We settled on using (class, instance) tuples to identify a
      specific engine, with an API for the user to construct a map of engines
      to capabilities. Into this picture, we then add a challenge of virtual
      engines; one user engine that maps behind the scenes to any number of
      physical engines. To keep it general, we want the user to have full
      control over that mapping. To that end, we allow the user to constrain a
      context to define the set of engines that it can access, order fully
      controlled by the user via (class, instance). With such precise control
      in context setup, we can continue to use the existing execbuf uABI of
      specifying a single index; only now it doesn't automagically map onto
      the engines, it uses the user defined engine map from the context.
      
      v2: Fixup freeing of local on success of get_engines()
      v3: Allow empty engines[]
      v4: s/nengine/num_engines/
      v5: Replace 64 limit on num_engines with a note that execbuf is
      currently limited to only using the first 64 engines.
      v6: Actually use the engines_mutex to guard the ctx->engines.
      
      Testcase: igt/gem_ctx_engines
      Signed-off-by: default avatarChris Wilson <chris@chris-wilson.co.uk>
      Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
      Reviewed-by: default avatarTvrtko Ursulin <tvrtko.ursulin@intel.com>
      Link: https://patchwork.freedesktop.org/patch/msgid/20190521211134.16117-2-chris@chris-wilson.co.uk
      976b55f0
    • Chris Wilson's avatar
      drm/i915: Restore control over ppgtt for context creation ABI · 7f3f317a
      Chris Wilson authored
      Having hid the partially exposed new ABI from the PR, put it back again
      for completion of context recovery. A significant part of context
      recovery is the ability to reuse as much of the old context as is
      feasible (to avoid expensive reconstruction). The biggest chunk kept
      hidden at the moment is fine-control over the ctx->ppgtt (the GPU page
      tables and associated translation tables and kernel maps), so make
      control over the ctx->ppgtt explicit.
      
      This allows userspace to create and share virtual memory address spaces
      (within the limits of a single fd) between contexts they own, along with
      the ability to query the contexts for the vm state.
      Signed-off-by: default avatarChris Wilson <chris@chris-wilson.co.uk>
      Reviewed-by: default avatarTvrtko Ursulin <tvrtko.ursulin@intel.com>
      Link: https://patchwork.freedesktop.org/patch/msgid/20190521211134.16117-1-chris@chris-wilson.co.uk
      7f3f317a
  2. 20 May, 2019 8 commits
  3. 17 May, 2019 7 commits
    • Chris Wilson's avatar
      drm/i915/execlists: Drop promotion on unsubmit · 4cc79cbb
      Chris Wilson authored
      With the disappearance of NEWCLIENT, we no longer need to provide the
      priority boost on preemption in order to prevent repeated gazumping,
      and we can remove the dead code.
      Signed-off-by: default avatarChris Wilson <chris@chris-wilson.co.uk>
      Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
      Reviewed-by: default avatarTvrtko Ursulin <tvrtko.ursulin@intel.com>
      Link: https://patchwork.freedesktop.org/patch/msgid/20190515130052.4475-5-chris@chris-wilson.co.uk
      4cc79cbb
    • Chris Wilson's avatar
      drm/i915: Downgrade NEWCLIENT to non-preemptive · 68fc728b
      Chris Wilson authored
      Commit 1413b2bc ("drm/i915: Trim NEWCLIENT boosting") had the
      intended consequence of not allowing a sequence of work that merely
      crossed into a new engine the privilege to be promoted to NEWCLIENT
      status. It also had the unintended consequence of actually making
      NEWCLIENT effective on heavily oversubscribed transcode machines and
      impacting upon their throughput.
      
      If we consider a client packet composed of (rcsA, rcsB, vcs) and 30 of
      those clients, using the NEWCLIENT boost that will be scheduled as
      
      	rcsA x 30, (rcsB, vcs) x 30
      
      where as before it would have been
      
      	(rcsA, rcsB, vcs) x 30
      
      That is with NEWCLIENT only boosting the first request of each client,
      we would execute all rcsA requests prior to running on the vcs engines;
      acruing a lot of dead time as compared to the previous case where the
      vcs engine would be started in parallel to processing the second client.
      
      The previous patch has the effect of delaying submission until it is
      required by a third party (either the user with an explicit wait, or by
      another client/engine). We reduce the NEWCLIENT bump to a mere WAIT,
      which has the effect of removing its preemptive grant and reducing it to
      the same level as any other user interaction -- that it will not be
      promoted above the interengine dependencies, and so preventing NEWCLIENTS
      from starving other engines. This a large nerf to the rrul properties of
      the current NEWCLIENT, but it still does give prioritised submission to
      new requests from light workloads.
      
      References: b16c7651 ("drm/i915: Priority boost for new clients")
      Fixes: 1413b2bc ("drm/i915: Trim NEWCLIENT boosting") # customer impact
      Signed-off-by: default avatarChris Wilson <chris@chris-wilson.co.uk>
      Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
      Cc: Dmitry Rogozhkin <dmitry.v.rogozhkin@intel.com>
      Cc: Dmitry Ermilov <dmitry.ermilov@intel.com>
      Reviewed-by: default avatarTvrtko Ursulin <tvrtko.ursulin@intel.com>
      Link: https://patchwork.freedesktop.org/patch/msgid/20190515130052.4475-4-chris@chris-wilson.co.uk
      68fc728b
    • Chris Wilson's avatar
      drm/i915: Bump signaler priority on adding a waiter · 6e7eb7a8
      Chris Wilson authored
      The handling of the no-preemption priority level imposes the restriction
      that we need to maintain the implied ordering even though preemption is
      disabled. Otherwise we may end up with an AB-BA deadlock across multiple
      engine due to a real preemption event reordering the no-preemption
      WAITs. To resolve this issue we currently promote all requests to WAIT
      on unsubmission, however this interferes with the timeslicing
      requirement that we do not apply any implicit promotion that will defeat
      the round-robin timeslice list. (If we automatically promote the active
      request it will go back to the head of the queue and not the tail!)
      
      So we need implicit promotion to prevent reordering around semaphores
      where we are not allowed to preempt, and we must avoid implicit
      promotion on unsubmission. So instead of at unsubmit, if we apply that
      implicit promotion on adding the dependency, we avoid the semaphore
      deadlock and we also reduce the gains made by the promotion for user
      space waiting. Furthermore, by keeping the earlier dependencies at a
      higher level, we reduce the search space for timeslicing without
      altering runtime scheduling too badly (no dependencies at all will be
      assigned a higher priority for rrul).
      
      v2: Limit the bump to external edges (as originally intended) i.e.
      between contexts and out to the user.
      
      Testcase: igt/gem_concurrent_blit
      Signed-off-by: default avatarChris Wilson <chris@chris-wilson.co.uk>
      Reviewed-by: default avatarTvrtko Ursulin <tvrtko.ursulin@intel.com>
      Link: https://patchwork.freedesktop.org/patch/msgid/20190515130052.4475-3-chris@chris-wilson.co.uk
      6e7eb7a8
    • Chris Wilson's avatar
      drm/i915/hdcp: Use both bits for device_count · af461ff3
      Chris Wilson authored
      Smatch spotted:
      drivers/gpu/drm/i915//intel_hdcp.c:1406 hdcp2_authenticate_repeater_topology() warn: should this be a bitwise op?
      
      and indeed looks to be suspect that we do need to use a bitwise or to
      combine the two register fields into one counter.
      Signed-off-by: default avatarChris Wilson <chris@chris-wilson.co.uk>
      Cc: Ramalingam C <ramalingam.c@intel.com>
      Reviewed-by: default avatarRamalingam C <ramalingam.c@intel.com>
      Link: https://patchwork.freedesktop.org/patch/msgid/20190517102225.3069-3-chris@chris-wilson.co.uk
      af461ff3
    • Chris Wilson's avatar
      drm/i915/dp: Initialise locals for static analysis · 96ac0813
      Chris Wilson authored
      Just to squelch an smatch warning that doesn't see the with_() being
      taken unconditionally:
      drivers/gpu/drm/i915//intel_dp.c:230 intel_dp_get_fia_supported_lane_count() error: uninitialized symbol 'lane_info'.
      drivers/gpu/drm/i915//intel_dp.c:5338 intel_digital_port_connected() error: uninitialized symbol 'is_connected'.
      Signed-off-by: default avatarChris Wilson <chris@chris-wilson.co.uk>
      Cc: Imre Deak <imre.deak@intel.com>
      Reviewed-by: default avatarImre Deak <imre.deak@intel.com>
      Link: https://patchwork.freedesktop.org/patch/msgid/20190517102225.3069-2-chris@chris-wilson.co.uk
      96ac0813
    • Chris Wilson's avatar
      drm/i915: Truly bump ready tasks ahead of busywaits · 17db337f
      Chris Wilson authored
      In commit b7404c7e ("drm/i915: Bump ready tasks ahead of
      busywaits"), I tried cutting a corner in order to not install a signal
      for each of our dependencies, and only listened to requests on which we
      were intending to busywait. The compromise that was made was that
      instead of then being able to promote the request with a full
      NOSEMAPHORE like its non-busywaiting brethren, as we had not ensured we
      had cleared the semaphore chain, we settled for only using the NEWCLIENT
      boost. With an over saturated system with multiple NEWCLIENTS in flight
      at any time, this was found to be an inadequate promotion and left us
      with a much poorer scheduling order than prior to using semaphores.
      
      The outcome of this patch, is that all requests have NOSEMAPHORE
      priority when they have no dependencies and are ready to run and not
      busywait, restoring the pre-semaphore ordering on saturated systems.
      
      We can demonstrate the effect of poor scheduling order by oversaturating
      the system using gem_wsim on a system with multiple vcs engines
      (i.e running the same workloads across more clients than required for
      peak throughput, e.g. media_load_balance_17i7.wsim -c4 -b context):
      
      x v5.1 (normalized)
      + tip
      * fix
      +------------------------------------------------------------------------+
      |                                                                    x   |
      |                                                                    x   |
      |                                                                    x   |
      |                                                                    x   |
      |                                                                   %x   |
      |                                                                  %%x   |
      |                                                                  %%x   |
      |                                                                  %%x   |
      |                                                                  %%x   |
      |                                                                  %%x   |
      |                                                                  %%x   |
      |                                                                  %%x   |
      |                                                                  %%x   |
      |                                                                  %%x   |
      |                                                                  %%x   |
      |                                                                  %#x   |
      |                                                                  %#x   |
      |                                                                  %#x   |
      |                                                                  %#x   |
      |                                                                  %#x   |
      |         +                                                        %#xx  |
      |         +                                                        %#xx  |
      |         +                                                       %%#xx  |
      |         +                                                       %%#xx  |
      |         +                                                       %%#xx  |
      |         +                                                       %%#xx  |
      |         +                                                       %%##x  |
      |         +++                                                     %%##x  |
      |         +++                                                     %%##x  |
      |         +++                                                     %%##x  |
      |        ++++                                                     %%##x  |
      |        ++++                                                     %%##x  |
      |        ++++                                                     %%##xx |
      |        ++++                                                     %###xx |
      |        ++++                                                     %###xx |
      |        ++++                                                     %###xx |
      |        ++++                                                     %###xx |
      |        ++++ +                                                   %#O#xx |
      |        ++++ +                                                   %#O#xx |
      |        ++++++ +                                                 %#O#xx |
      |       ++++++++++                                                %OOOxxx|
      |       ++++++++++       +                                       %#OOO#xx|
      |     + ++++++++++++ ++ +++++    +                        ++    @@OOOO#xx|
      |                                                                   |A_| |
      ||__________M_______A____________________|                               |
      |                                                                 |A_|   |
      +------------------------------------------------------------------------+
          N           Min           Max        Median           Avg        Stddev
      x 120       0.99456       1.00628      0.999985     1.0001545  0.0024387139
      + 120      0.873021       1.00037      0.884134    0.90148752   0.039190862
      Difference at 99.5% confidence
      	-0.098667 +/- 0.0110762
      	-9.86517% +/- 1.10745%
      	(Student's t, pooled s = 0.0277657)
      % 120      0.990207       1.00165     0.9970265    0.99699748     0.0021024
      Difference at 99.5% confidence
      	-0.003157 +/- 0.000908245
      	-0.315651% +/- 0.0908105%
      	(Student's t, pooled s = 0.00227678)
      
      Fixes: b7404c7e ("drm/i915: Bump ready tasks ahead of busywaits")
      Signed-off-by: default avatarChris Wilson <chris@chris-wilson.co.uk>
      Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
      Cc: Dmitry Rogozhkin <dmitry.v.rogozhkin@intel.com>
      Cc: Dmitry Ermilov <dmitry.ermilov@intel.com>
      Reviewed-by: default avatarTvrtko Ursulin <tvrtko.ursulin@intel.com>
      Link: https://patchwork.freedesktop.org/patch/msgid/20190515130052.4475-2-chris@chris-wilson.co.uk
      17db337f
    • Chris Wilson's avatar
      drm/i915: Mark semaphores as complete on unsubmit out if payload was started · dba5a7f3
      Chris Wilson authored
      Avoid charging us for the presumed busywait if the request was preempted
      after successfully using semaphores to reduce inter-engine latency.
      
      v2: Bump the priority to reflect the lack of semaphores now required.
      
      References: ca6e56f6 ("drm/i915: Disable semaphore busywaits on saturated systems")
      Signed-off-by: default avatarChris Wilson <chris@chris-wilson.co.uk>
      Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
      Reviewed-by: default avatarTvrtko Ursulin <tvrtko.ursulin@intel.com>
      Link: https://patchwork.freedesktop.org/patch/msgid/20190515130052.4475-1-chris@chris-wilson.co.uk
      dba5a7f3
  4. 14 May, 2019 16 commits
  5. 13 May, 2019 3 commits