Skip to content

K
Kernel

openeuler / Kernel
1 年多 前同步成功

通知 8
Star 0
Fork 0
K
Kernel
切换分支/标签
  1. 15 6月, 2019 2 次提交
    • C
      drm/i915: Replace engine->timeline with a plain list · 422d7df4
      由 Chris Wilson 提交于 
      To continue the onslaught of removing the assumption of a global
execution ordering, another casualty is the engine->timeline. Without an
actual timeline to track, it is overkill and we can replace it with a
much less grand plain list. We still need a list of requests inflight,
for the simple purpose of finding inflight requests (for retiring,
resetting, preemption etc).
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: NMika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190614164606.15633-3-chris@chris-wilson.co.uk
      422d7df4
    • C
      drm/i915: Keep contexts pinned until after the next kernel context switch · ce476c80
      由 Chris Wilson 提交于 
      We need to keep the context image pinned in memory until after the GPU
has finished writing into it. Since it continues to write as we signal
the final breadcrumb, we need to keep it pinned until the request after
it is complete. Currently we know the order in which requests execute on
each engine, and so to remove that presumption we need to identify a
request/context-switch we know must occur after our completion. Any
request queued after the signal must imply a context switch, for
simplicity we use a fresh request from the kernel context.

The sequence of operations for keeping the context pinned until saved is:

 - On context activation, we preallocate a node for each physical engine
   the context may operate on. This is to avoid allocations during
   unpinning, which may be from inside FS_RECLAIM context (aka the
   shrinker)

 - On context deactivation on retirement of the last active request (which
   is before we know the context has been saved), we add the
   preallocated node onto a barrier list on each engine

 - On engine idling, we emit a switch to kernel context. When this
   switch completes, we know that all previous contexts must have been
   saved, and so on retiring this request we can finally unpin all the
   contexts that were marked as deactivated prior to the switch.

We can enhance this in future by flushing all the idle contexts on a
regular heartbeat pulse of a switch to kernel context, which will also
be used to check for hung engines.

v2: intel_context_active_acquire/_release
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: NMika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190614164606.15633-1-chris@chris-wilson.co.uk
      ce476c80
  3. 11 6月, 2019 2 次提交
  5. 07 6月, 2019 2 次提交
  7. 28 5月, 2019 3 次提交
  9. 22 5月, 2019 3 次提交
    • C
      drm/i915/execlists: Virtual engine bonding · ee113690
      由 Chris Wilson 提交于 
      Some users require that when a master batch is executed on one particular
engine, a companion batch is run simultaneously on a specific slave
engine. For this purpose, we introduce virtual engine bonding, allowing
maps of master:slaves to be constructed to constrain which physical
engines a virtual engine may select given a fence on a master engine.

For the moment, we continue to ignore the issue of preemption deferring
the master request for later. Ideally, we would like to then also remove
the slave and run something else rather than have it stall the pipeline.
With load balancing, we should be able to move workload around it, but
there is a similar stall on the master pipeline while it may wait for
the slave to be executed. At the cost of more latency for the bonded
request, it may be interesting to launch both on their engines in
lockstep. (Bubbles abound.)

Opens: Also what about bonding an engine as its own master? It doesn't
break anything internally, so allow the silliness.

v2: Emancipate the bonds
v3: Couple in delayed scheduling for the selftests
v4: Handle invalid mutually exclusive bonding
v5: Mention what the uapi does
v6: s/nbond/num_bonds/
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: NTvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190521211134.16117-9-chris@chris-wilson.co.uk
      ee113690
    • C
      drm/i915: Apply an execution_mask to the virtual_engine · 78e41ddd
      由 Chris Wilson 提交于 
      Allow the user to direct which physical engines of the virtual engine
they wish to execute one, as sometimes it is necessary to override the
load balancing algorithm.

v2: Only kick the virtual engines on context-out if required
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: NTvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190521211134.16117-7-chris@chris-wilson.co.uk
      78e41ddd
    • C
      drm/i915: Load balancing across a virtual engine · 6d06779e
      由 Chris Wilson 提交于 
      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: NChris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: NTvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190521211134.16117-6-chris@chris-wilson.co.uk
      6d06779e
  11. 17 5月, 2019 2 次提交
    • C
      drm/i915/execlists: Drop promotion on unsubmit · 4cc79cbb
      由 Chris Wilson 提交于 
      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: NChris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: NTvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190515130052.4475-5-chris@chris-wilson.co.uk
      4cc79cbb
    • C
      drm/i915: Downgrade NEWCLIENT to non-preemptive · 68fc728b
      由 Chris Wilson 提交于 
      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: NChris 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: NTvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190515130052.4475-4-chris@chris-wilson.co.uk
      68fc728b
  13. 08 5月, 2019 3 次提交
  15. 07 5月, 2019 1 次提交
  17. 03 5月, 2019 1 次提交
  19. 01 5月, 2019 1 次提交
  21. 27 4月, 2019 1 次提交
  23. 25 4月, 2019 3 次提交
    • C
      drm/i915: Invert the GEM wakeref hierarchy · 79ffac85
      由 Chris Wilson 提交于 
      In the current scheme, on submitting a request we take a single global
GEM wakeref, which trickles down to wake up all GT power domains. This
is undesirable as we would like to be able to localise our power
management to the available power domains and to remove the global GEM
operations from the heart of the driver. (The intent there is to push
global GEM decisions to the boundary as used by the GEM user interface.)

Now during request construction, each request is responsible via its
logical context to acquire a wakeref on each power domain it intends to
utilize. Currently, each request takes a wakeref on the engine(s) and
the engines themselves take a chipset wakeref. This gives us a
transition on each engine which we can extend if we want to insert more
powermangement control (such as soft rc6). The global GEM operations
that currently require a struct_mutex are reduced to listening to pm
events from the chipset GT wakeref. As we reduce the struct_mutex
requirement, these listeners should evaporate.

Perhaps the biggest immediate change is that this removes the
struct_mutex requirement around GT power management, allowing us greater
flexibility in request construction. Another important knock-on effect,
is that by tracking engine usage, we can insert a switch back to the
kernel context on that engine immediately, avoiding any extra delay or
inserting global synchronisation barriers. This makes tracking when an
engine and its associated contexts are idle much easier -- important for
when we forgo our assumed execution ordering and need idle barriers to
unpin used contexts. In the process, it means we remove a large chunk of
code whose only purpose was to switch back to the kernel context.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Cc: Imre Deak <imre.deak@intel.com>
Reviewed-by: NTvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190424200717.1686-5-chris@chris-wilson.co.uk
      79ffac85
    • C
      drm/i915: Introduce context->enter() and context->exit() · 6eee33e8
      由 Chris Wilson 提交于 
      We wish to start segregating the power management into different control
domains, both with respect to the hardware and the user interface. The
first step is that at the lowest level flow of requests, we want to
process a context event (and not a global GEM operation). In this patch,
we introduce the context callbacks that in future patches will be
redirected to per-engine interfaces leading to global operations as
required.

The intent is that this will be guarded by the timeline->mutex, except
that retiring has not quite finished transitioning over from being
guarded by struct_mutex. So at the moment it is protected by
struct_mutex with a reminded to switch.

v2: Rename default handlers to intel_context_enter_engine.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: NTvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190424200717.1686-3-chris@chris-wilson.co.uk
      6eee33e8
    • C
      drm/i915: Move GraphicsTechnology files under gt/ · 112ed2d3
      由 Chris Wilson 提交于 
      Start partitioning off the code that talks to the hardware (GT) from the
uapi layers and move the device facing code under gt/

One casualty is s/intel_ringbuffer.h/intel_engine.h/ with the plan to
subdivide that header and body further (and split out the submission
code from the ringbuffer and logical context handling). This patch aims
to be simple motion so git can fixup inflight patches with little mess.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Acked-by: NJoonas Lahtinen <joonas.lahtinen@linux.intel.com>
Acked-by: NJani Nikula <jani.nikula@intel.com>
Acked-by: NRodrigo Vivi <rodrigo.vivi@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190424174839.7141-1-chris@chris-wilson.co.uk
      112ed2d3
  25. 24 4月, 2019 1 次提交
  27. 12 4月, 2019 3 次提交
    • C
      drm/i915: Flush the CSB pointer reset · 0edda1d6
      由 Chris Wilson 提交于 
      The HW resets it CSB tail pointer on resetting the engine. Most of the
time. In case it doesn't (and for system resume) we write the expected
value anyway. For extra paranoia, flush the write before we invalidate
the cacheline.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Acked-by: NMika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190412110159.10495-1-chris@chris-wilson.co.uk
      0edda1d6
    • C
      drm/i915/execlists: Always reset the context's RING registers · 1863e302
      由 Chris Wilson 提交于 
      During reset, we try and stop the active ring. This has the consequence
that we often clobber the RING registers within the context image. When
we find an active request, we update the context image to rerun that
request (if it was guilty, we replace the hanging user payload with
NOPs). However, we were ignoring an active context if the request had
completed, with the consequence that the next submission on that request
would start with RING_HEAD==0 and not the tail of the previous request,
causing all requests still in the ring to be rerun. Rare, but
occasionally seen within CI where we would spot that the context seqno
would reverse and complain that we were retiring an incomplete request.

    <0> [412.390350]   <idle>-0       3d.s2 408373352us : __i915_request_submit: rcs0 fence 1e95b:3640 -> current 3638
    <0> [412.390350]   <idle>-0       3d.s2 408373353us : __i915_request_submit: rcs0 fence 1e95b:3642 -> current 3638
    <0> [412.390350]   <idle>-0       3d.s2 408373354us : __i915_request_submit: rcs0 fence 1e95b:3644 -> current 3638
    <0> [412.390350]   <idle>-0       3d.s2 408373354us : __i915_request_submit: rcs0 fence 1e95b:3646 -> current 3638
    <0> [412.390350]   <idle>-0       3d.s2 408373356us : __execlists_submission_tasklet: rcs0 in[0]:  ctx=2.1, fence 1e95b:3646 (current 3638), prio=4
    <0> [412.390350] i915_sel-4613    0.... 408373374us : __i915_request_commit: rcs0 fence 1e95b:3648
    <0> [412.390350] i915_sel-4613    0d..1 408373377us : process_csb: rcs0 cs-irq head=2, tail=3
    <0> [412.390350] i915_sel-4613    0d..1 408373377us : process_csb: rcs0 csb[3]: status=0x00000001:0x00000000, active=0x1
    <0> [412.390350] i915_sel-4613    0d..1 408373378us : __i915_request_submit: rcs0 fence 1e95b:3648 -> current 3638
    <0> [412.390350]   <idle>-0       3..s1 408373378us : execlists_submission_tasklet: rcs0 awake?=1, active=5
    <0> [412.390350] i915_sel-4613    0d..1 408373379us : __execlists_submission_tasklet: rcs0 in[0]:  ctx=2.2, fence 1e95b:3648 (current 3638), prio=4
    <0> [412.390350] i915_sel-4613    0.... 408373381us : i915_reset_engine: rcs0 flags=4
    <0> [412.390350] i915_sel-4613    0.... 408373382us : execlists_reset_prepare: rcs0: depth<-0
    <0> [412.390350]   <idle>-0       3d.s2 408373390us : process_csb: rcs0 cs-irq head=3, tail=4
    <0> [412.390350]   <idle>-0       3d.s2 408373390us : process_csb: rcs0 csb[4]: status=0x00008002:0x00000002, active=0x1
    <0> [412.390350]   <idle>-0       3d.s2 408373390us : process_csb: rcs0 out[0]: ctx=2.2, fence 1e95b:3648 (current 3640), prio=4
    <0> [412.390350] i915_sel-4613    0.... 408373401us : intel_engine_stop_cs: rcs0
    <0> [412.390350] i915_sel-4613    0d..1 408373402us : process_csb: rcs0 cs-irq head=4, tail=4
    <0> [412.390350] i915_sel-4613    0.... 408373403us : intel_gpu_reset: engine_mask=1
    <0> [412.390350] i915_sel-4613    0d..1 408373408us : execlists_cancel_port_requests: rcs0:port0 fence 1e95b:3648, (current 3648)
    <0> [412.390350] i915_sel-4613    0.... 408373442us : intel_engine_cancel_stop_cs: rcs0
    <0> [412.390350] i915_sel-4613    0.... 408373442us : execlists_reset_finish: rcs0: depth->0
    <0> [412.390350] ksoftirq-26      3..s. 408373442us : execlists_submission_tasklet: rcs0 awake?=1, active=0
    <0> [412.390350] ksoftirq-26      3d.s1 408373443us : process_csb: rcs0 cs-irq head=5, tail=5
    <0> [412.390350] i915_sel-4613    0.... 408373475us : i915_request_retire: rcs0 fence 1e95b:3640, current 3648
    <0> [412.390350] i915_sel-4613    0.... 408373476us : i915_request_retire: __retire_engine_request(rcs0) fence 1e95b:3640, current 3648
    <0> [412.390350] i915_sel-4613    0.... 408373494us : __i915_request_commit: rcs0 fence 1e95b:3650
    <0> [412.390350] i915_sel-4613    0d..1 408373496us : process_csb: rcs0 cs-irq head=5, tail=5
    <0> [412.390350] i915_sel-4613    0d..1 408373496us : __i915_request_submit: rcs0 fence 1e95b:3650 -> current 3648
    <0> [412.390350] i915_sel-4613    0d..1 408373498us : __execlists_submission_tasklet: rcs0 in[0]:  ctx=2.1, fence 1e95b:3650 (current 3648), prio=6
    <0> [412.390350] i915_sel-4613    0.... 408373500us : i915_request_retire_upto: rcs0 fence 1e95b:3648, current 3648
    <0> [412.390350] i915_sel-4613    0.... 408373500us : i915_request_retire: rcs0 fence 1e95b:3642, current 3648
    <0> [412.390350] i915_sel-4613    0.... 408373501us : i915_request_retire: __retire_engine_request(rcs0) fence 1e95b:3642, current 3648
    <0> [412.390350] i915_sel-4613    0.... 408373514us : i915_request_retire: rcs0 fence 1e95b:3644, current 3648
    <0> [412.390350] i915_sel-4613    0.... 408373515us : i915_request_retire: __retire_engine_request(rcs0) fence 1e95b:3644, current 3648
    <0> [412.390350] i915_sel-4613    0.... 408373527us : i915_request_retire: rcs0 fence 1e95b:3646, current 3640
    <0> [412.390350]   <idle>-0       3..s1 408373569us : execlists_submission_tasklet: rcs0 awake?=1, active=1
    <0> [412.390350]   <idle>-0       3d.s2 408373569us : process_csb: rcs0 cs-irq head=5, tail=1
    <0> [412.390350]   <idle>-0       3d.s2 408373570us : process_csb: rcs0 csb[0]: status=0x00000001:0x00000000, active=0x1
    <0> [412.390350]   <idle>-0       3d.s2 408373570us : process_csb: rcs0 csb[1]: status=0x00000018:0x00000002, active=0x5
    <0> [412.390350]   <idle>-0       3d.s2 408373570us : process_csb: rcs0 out[0]: ctx=2.1, fence 1e95b:3650 (current 3650), prio=6
    <0> [412.390350]   <idle>-0       3d.s2 408373571us : process_csb: rcs0 completed ctx=2
    <0> [412.390350] i915_sel-4613    0.... 408373621us : i915_request_retire: i915_request_retire:253 GEM_BUG_ON(!i915_request_completed(request))

v2: Fixup the cancellation path to drain the CSB and reset the pointers.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
Reviewed-by: NMika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190411130515.20716-2-chris@chris-wilson.co.uk
      1863e302
    • C
      drm/i915/guc: Implement reset locally · 292ad25c
      由 Chris Wilson 提交于 
      Before causing guc and execlists to diverge further (breaking guc in the
process), take a copy of the current reset procedure and make it local to
the guc submission backend
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: NMika Kuoppala <mika.kuoppala@linux.intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190411130515.20716-1-chris@chris-wilson.co.uk
      292ad25c
  29. 11 4月, 2019 3 次提交
  31. 05 4月, 2019 2 次提交
    • C
      drm/i915: Make RING_PDP relative to engine->mmio_base · 6d425728
      由 Chris Wilson 提交于 
      The PDP registers are an oddity inside the set of context saved
registers in that they take the engine as a parameter to the macro and
not the mmio_base as the others do. Make it accept the engine->mmio_base
for consistency in programming the context registers.

add/remove: 0/0 grow/shrink: 2/1 up/down: 3/-32 (-29)
Function                                     old     new   delta
emit_ppgtt_update                            324     326      +2
capture                                     5102    5103      +1
execlists_init_reg_state.isra               1128    1096     -32

And similar savings later!
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: NTvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190405123831.9724-1-chris@chris-wilson.co.uk
      6d425728
    • C
      drm/i915/execlists: Enable coarse preemption boundaries for gen8 · bac24f59
      由 Chris Wilson 提交于 
      When we introduced preemption, we chose to keep it disabled for gen8 as
supporting preemption inside GPGPU user batches required various w/a in
userspace. Since then, the desire to preempt long queues of requests
between batches (e.g. within busywaiting semaphores) has grown. So allow
arbitration within the busywaits and between requests, but disable
arbitration within user batches so that we can preempt between requests
and not risk breaking GPGPU.

However, since this preemption is much coarser and doesn't interfere
with userspace, we decline to include it amongst the scheduler
capabilities. (This is also required for us to skip over the preemption
selftests that expect to be able to preempt user batches.)

Michal suggested that we could perhaps allow preemption inside gen8
userspace batches if we can satisfy ourselves that the default
preemption settings are viable with existing userspace (principally
OpenCL which already should carry any known workaround). We could then
merge the two code paths back into one, even dropping the artifical
has-preemption device feature flag.

Testcase: igt/gem_exec_scheduler/semaphore-user
References: beecec90 ("drm/i915/execlists: Preemption!")
Fixes: e8861964 ("drm/i915: Use HW semaphores for inter-engine synchronisation on gen8+")
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Cc: Michal Winiarski <michal.winiarski@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@intel.com>
Reviewed-by: Michal Winiarski <michal.winiarski@intel.com> #irc
Reviewed-by: NTvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190329134024.5254-1-chris@chris-wilson.co.uk
      bac24f59
  33. 27 3月, 2019 2 次提交
  35. 22 3月, 2019 2 次提交
    • C
      drm/i915: Allow contexts to share a single timeline across all engines · ea593dbb
      由 Chris Wilson 提交于 
      Previously, our view has been always to run the engines independently
within a context. (Multiple engines happened before we had contexts and
timelines, so they always operated independently and that behaviour
persisted into contexts.) However, at the user level the context often
represents a single timeline (e.g. GL contexts) and userspace must
ensure that the individual engines are serialised to present that
ordering to the client (or forgot about this detail entirely and hope no
one notices - a fair ploy if the client can only directly control one
engine themselves ;)

In the next patch, we will want to construct a set of engines that
operate as one, that have a single timeline interwoven between them, to
present a single virtual engine to the user. (They submit to the virtual
engine, then we decide which engine to execute on based.)

To that end, we want to be able to create contexts which have a single
timeline (fence context) shared between all engines, rather than multiple
timelines.

v2: Move the specialised timeline ordering to its own function.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Reviewed-by: NTvrtko Ursulin <tvrtko.ursulin@intel.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190322092325.5883-4-chris@chris-wilson.co.uk
      ea593dbb
    • C
      drm/i915: Flush pages on acquisition · a679f58d
      由 Chris Wilson 提交于 
      When we return pages to the system, we ensure that they are marked as
being in the CPU domain since any external access is uncontrolled and we
must assume the worst. This means that we need to always flush the pages
on acquisition if we need to use them on the GPU, and from the beginning
have used set-domain. Set-domain is overkill for the purpose as it is a
general synchronisation barrier, but our intent is to only flush the
pages being swapped in. If we move that flush into the pages acquisition
phase, we know then that when we have obj->mm.pages, they are coherent
with the GPU and need only maintain that status without resorting to
heavy handed use of set-domain.

The principle knock-on effect for userspace is through mmap-gtt
pagefaulting. Our uAPI has always implied that the GTT mmap was async
(especially as when any pagefault occurs is unpredicatable to userspace)
and so userspace had to apply explicit domain control itself
(set-domain). However, swapping is transparent to the kernel, and so on
first fault we need to acquire the pages and make them coherent for
access through the GTT. Our use of set-domain here leaks into the uABI
that the first pagefault was synchronous. This is unintentional and
baring a few igt should be unoticed, nevertheless we bump the uABI
version for mmap-gtt to reflect the change in behaviour.

Another implication of the change is that gem_create() is presumed to
create an object that is coherent with the CPU and is in the CPU write
domain, so a set-domain(CPU) following a gem_create() would be a minor
operation that merely checked whether we could allocate all pages for
the object. On applying this change, a set-domain(CPU) causes a clflush
as we acquire the pages. This will have a small impact on mesa as we move
the clflush here on !llc from execbuf time to create, but that should
have minimal performance impact as the same clflush exists but is now
done early and because of the clflush issue, userspace recycles bo and
so should resist allocating fresh objects.

Internally, the presumption that objects are created in the CPU
write-domain and remain so through writes to obj->mm.mapping is more
prevalent than I expected; but easy enough to catch and apply a manual
flush.

For the future, we should push the page flush from the central
set_pages() into the callers so that we can more finely control when it
is applied, but for now doing it one location is easier to validate, at
the cost of sometimes flushing when there is no need.
Signed-off-by: NChris Wilson <chris@chris-wilson.co.uk>
Cc: Matthew Auld <matthew.william.auld@gmail.com>
Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Antonio Argenziano <antonio.argenziano@intel.com>
Cc: Joonas Lahtinen <joonas.lahtinen@linux.intel.com>
Reviewed-by: NMatthew Auld <matthew.william.auld@gmail.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20190321161908.8007-1-chris@chris-wilson.co.uk
      a679f58d
  37. 21 3月, 2019 2 次提交
  39. 19 3月, 2019 1 次提交