All of lore.kernel.org
 help / color / mirror / Atom feed
From: Matthew Brost <matthew.brost@intel.com>
To: Chris Wilson <chris@chris-wilson.co.uk>
Cc: intel-gfx@lists.freedesktop.org
Subject: Re: [Intel-gfx] [PATCH 36/56] drm/i915: Fair low-latency scheduling
Date: Thu, 7 Jan 2021 09:10:35 -0800	[thread overview]
Message-ID: <20210107171035.GA9023@sdutt-i7> (raw)
In-Reply-To: <161003793126.28368.13767958122371874524@build.alporthouse.com>

On Thu, Jan 07, 2021 at 04:45:31PM +0000, Chris Wilson wrote:
> Quoting Matthew Brost (2021-01-07 16:05:07)
> > On Tue, Dec 29, 2020 at 12:01:25PM +0000, Chris Wilson wrote:
> > > The first "scheduler" was a topographical sorting of requests into
> > > priority order. The execution order was deterministic, the earliest
> > > submitted, highest priority request would be executed first. Priority
> > > inheritance ensured that inversions were kept at bay, and allowed us to
> > > dynamically boost priorities (e.g. for interactive pageflips).
> > > 
> > > The minimalistic timeslicing scheme was an attempt to introduce fairness
> > > between long running requests, by evicting the active request at the end
> > > of a timeslice and moving it to the back of its priority queue (while
> > > ensuring that dependencies were kept in order). For short running
> > > requests from many clients of equal priority, the scheme is still very
> > > much FIFO submission ordering, and as unfair as before.
> > > 
> > > To impose fairness, we need an external metric that ensures that clients
> > > are interpersed, so we don't execute one long chain from client A before
> > > executing any of client B. This could be imposed by the clients
> > > themselves by using fences based on an external clock, that is they only
> > > submit work for a "frame" at frame-intervals, instead of submitting as
> > > much work as they are able to. The standard SwapBuffers approach is akin
> > > to double bufferring, where as one frame is being executed, the next is
> > > being submitted, such that there is always a maximum of two frames per
> > > client in the pipeline and so ideally maintains consistent input-output
> > > latency. Even this scheme exhibits unfairness under load as a single
> > > client will execute two frames back to back before the next, and with
> > > enough clients, deadlines will be missed.
> > > 
> > > The idea introduced by BFS/MuQSS is that fairness is introduced by
> > > metering with an external clock. Every request, when it becomes ready to
> > > execute is assigned a virtual deadline, and execution order is then
> > > determined by earliest deadline. Priority is used as a hint, rather than
> > > strict ordering, where high priority requests have earlier deadlines,
> > > but not necessarily earlier than outstanding work. Thus work is executed
> > > in order of 'readiness', with timeslicing to demote long running work.
> > > 
> > > The Achille's heel of this scheduler is its strong preference for
> > > low-latency and favouring of new queues. Whereas it was easy to dominate
> > > the old scheduler by flooding it with many requests over a short period
> > > of time, the new scheduler can be dominated by a 'synchronous' client
> > > that waits for each of its requests to complete before submitting the
> > > next. As such a client has no history, it is always considered
> > > ready-to-run and receives an earlier deadline than the long running
> > > requests. This is compensated for by refreshing the current execution's
> > > deadline and by disallowing preemption for timeslice shuffling.
> > > 
> > > To check the impact on throughput (often the downfall of latency
> > > sensitive schedulers), we used gem_wsim to simulate various transcode
> > > workloads with different load balancers, and varying the number of
> > > competing [heterogenous] clients.
> > > 
> > > +delta%------------------------------------------------------------------+
> > > |                                a                                       |
> > > |                                a                                       |
> > > |                                aa                                      |
> > > |                                aa                                      |
> > > |                                aa                                      |
> > > |                                aa                                      |
> > > |                               aaa                                      |
> > > |                              aaaa                                      |
> > > |                           a  aaaaa                                     |
> > > |                           a aaaaaa                                     |
> > > |a              aa   a      aaaaaaaaaa aa               a               a|
> > > |                                A_|                                     |
> > > +------------------------------------------------------------------------+
> > >    N          Min           Max        Median           Avg        Stddev
> > >  108   -23.982194     28.421527  -0.077474828  -0.072650418    0.16179718
> > > 
> > > The impact was on average 0.1% under contention due to the change in
> > > context execution order and number of context switches. The biggest
> > > swings are due to the execution ordering favouring one client or another,
> > > and maybe room for improvement.
> > > 
> > 
> > I haven't dug into this series deeply but it does seem plausible for
> > execlist submission. However this new scheduler seems completely broken
> > for Guc submission.
> 
> Underneath it's the same topological sort, just with a different key.
> At worst, that key can be the plain priority again. But that's pretty
> much immaterial in terms of deciding the order which requests are
> transferred from one queue to another. Execution order wants to be based
> on resource prioritisation (beyond request priorisation), ideally we
> would keep our fine grained scheduling control.
> 

So maybe just hook for the key?

> I'm not convinced by the state of the fw's scheduler...
>

Can't really argue here, but I think it would be really helpful if you
raised your concerns with the fw team so that can be addressed.
 
> > The scheduler really isn't used that much with GuC
> > submission but when it is the old one works perfectly. The vfunc for
> > the scheduler is deleted in patch #25. I don't think that is good idea
> > as it appears we are trending to 2 separate schedulers.
> 
> You still use the i915_request_set_priority() entry point as the
> caller into the backend bypass.
>

I thought we want to avoid backend specific branches, hence vfuncs.

> Isn't the current guc.fw a second closed driver for the same HW anyway?
>

Not sure I follow this.
 
> > Lastly this series seems to be on the wrong list, I believe it should be
> > posted elsewhere first.
> 
> It's the right place for upstreaming bug fixes on stable platforms. And
> after a year of being here, that's a moot point.

Wow, I feel your pain if this has been out here for a year. But still it
is wrong place. I can help and will help with the reviews if it is in
the correct spot.

Matt

> -Chris
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx

  reply	other threads:[~2021-01-07 17:15 UTC|newest]

Thread overview: 65+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2020-12-29 12:00 [Intel-gfx] [PATCH 01/56] drm/i915/gt: Restore ce->signal flush before releasing virtual engine Chris Wilson
2020-12-29 12:00 ` [Intel-gfx] [PATCH 02/56] drm/i915/gt: Only retire on the last breadcrumb if the last request Chris Wilson
2020-12-29 12:00 ` [Intel-gfx] [PATCH 03/56] drm/i915/gt: Cancel submitted requests upon context reset Chris Wilson
2020-12-30 21:07   ` Mika Kuoppala
2020-12-29 12:00 ` [Intel-gfx] [PATCH 04/56] drm/i915/gt: Pull context closure check from request submit to schedule-in Chris Wilson
2020-12-29 12:00 ` [Intel-gfx] [PATCH 05/56] drm/i915/gem: Peek at the inflight context Chris Wilson
2020-12-29 12:00 ` [Intel-gfx] [PATCH 06/56] drm/i915: Mark up protected uses of 'i915_request_completed' Chris Wilson
2020-12-29 12:00 ` [Intel-gfx] [PATCH 07/56] drm/i915: Drop i915_request.lock serialisation around await_start Chris Wilson
2020-12-29 12:00 ` [Intel-gfx] [PATCH 08/56] drm/i915: Drop i915_request.lock requirement for intel_rps_boost() Chris Wilson
2020-12-29 12:00 ` [Intel-gfx] [PATCH 09/56] drm/i915/gem: Reduce ctx->engine_mutex for reading the clone source Chris Wilson
2020-12-29 12:00 ` [Intel-gfx] [PATCH 10/56] drm/i915/gem: Reduce ctx->engines_mutex for get_engines() Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 11/56] drm/i915: Reduce test_and_set_bit to set_bit in i915_request_submit() Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 12/56] drm/i915/gt: Drop atomic for engine->fw_active tracking Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 13/56] drm/i915/gt: Extract busy-stats for ring-scheduler Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 14/56] drm/i915/gt: Convert stats.active to plain unsigned int Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 15/56] drm/i915/gt: Do not suspend bonded requests if one hangs Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 16/56] drm/i915/gt: Remove timeslice suppression Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 17/56] drm/i915/gt: Skip over completed active execlists, again Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 18/56] drm/i915: Strip out internal priorities Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 19/56] drm/i915: Remove I915_USER_PRIORITY_SHIFT Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 20/56] drm/i915: Replace engine->schedule() with a known request operation Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 21/56] drm/i915: Teach the i915_dependency to use a double-lock Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 22/56] drm/i915: Restructure priority inheritance Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 23/56] drm/i915/selftests: Measure set-priority duration Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 24/56] drm/i915/selftests: Exercise priority inheritance around an engine loop Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 25/56] drm/i915: Improve DFS for priority inheritance Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 26/56] drm/i915: Extract request submission from execlists Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 27/56] drm/i915: Extract request rewinding " Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 28/56] drm/i915: Extract request suspension from the execlists backend Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 29/56] drm/i915: Extract the ability to defer and rerun a request later Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 30/56] drm/i915: Fix the iterative dfs for defering requests Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 31/56] drm/i915: Move common active lists from engine to i915_scheduler Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 32/56] drm/i915: Move scheduler queue Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 33/56] drm/i915: Move tasklet from execlists to sched Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 34/56] drm/i915: Replace priolist rbtree with a skiplist Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 35/56] drm/i915: Wrap cmpxchg64 with try_cmpxchg64() helper Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 36/56] drm/i915: Fair low-latency scheduling Chris Wilson
2021-01-07 16:05   ` Matthew Brost
2021-01-07 16:45     ` Chris Wilson
2021-01-07 17:10       ` Matthew Brost [this message]
2020-12-29 12:01 ` [Intel-gfx] [PATCH 37/56] drm/i915/gt: Specify a deadline for the heartbeat Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 38/56] drm/i915: Extend the priority boosting for the display with a deadline Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 39/56] drm/i915/gt: Support virtual engine queues Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 40/56] drm/i915: Move saturated workload detection back to the context Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 41/56] drm/i915: Bump default timeslicing quantum to 5ms Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 42/56] drm/i915/gt: Wrap intel_timeline.has_initial_breadcrumb Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 43/56] drm/i915/gt: Track timeline GGTT offset separately from subpage offset Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 44/56] drm/i915/gt: Add timeline "mode" Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 45/56] drm/i915/gt: Use indices for writing into relative timelines Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 46/56] drm/i915/selftests: Exercise relative timeline modes Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 47/56] drm/i915/gt: Use ppHWSP for unshared non-semaphore related timelines Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 48/56] Restore "drm/i915: drop engine_pin/unpin_breadcrumbs_irq" Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 49/56] drm/i915/gt: Couple tasklet scheduling for all CS interrupts Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 50/56] drm/i915/gt: Support creation of 'internal' rings Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 51/56] drm/i915/gt: Use client timeline address for seqno writes Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 52/56] drm/i915/gt: Infrastructure for ring scheduling Chris Wilson
2020-12-29 18:34   ` kernel test robot
2020-12-29 18:34     ` kernel test robot
2020-12-29 12:01 ` [Intel-gfx] [PATCH 53/56] drm/i915/gt: Enable busy-stats for ring-scheduler Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 54/56] drm/i915/gt: Implement ring scheduler for gen6/7 Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 55/56] drm/i915/gt: Enable ring scheduling " Chris Wilson
2020-12-29 12:01 ` [Intel-gfx] [PATCH 56/56] drm/i915/gt: Limit C-states while waiting for requests Chris Wilson
2020-12-29 12:20 ` [Intel-gfx] ✗ Fi.CI.CHECKPATCH: warning for series starting with [01/56] drm/i915/gt: Restore ce->signal flush before releasing virtual engine Patchwork
2020-12-29 12:22 ` [Intel-gfx] ✗ Fi.CI.SPARSE: " Patchwork
2020-12-29 12:49 ` [Intel-gfx] ✗ Fi.CI.BAT: failure " Patchwork

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20210107171035.GA9023@sdutt-i7 \
    --to=matthew.brost@intel.com \
    --cc=chris@chris-wilson.co.uk \
    --cc=intel-gfx@lists.freedesktop.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.