From: Mika Kuoppala <mika.kuoppala@linux.intel.com>
To: Chris Wilson <chris@chris-wilson.co.uk>, intel-gfx@lists.freedesktop.org
Subject: Re: [PATCH 04/12] drm/i915/execlists: Refactor CSB state machine
Date: Mon, 01 Jul 2019 14:49:48 +0300 [thread overview]
Message-ID: <875zom3rfn.fsf@gaia.fi.intel.com> (raw)
In-Reply-To: <20190701100502.15639-4-chris@chris-wilson.co.uk>
Chris Wilson <chris@chris-wilson.co.uk> writes:
> Daniele pointed out that the CSB status information will change with
> Tigerlake and suggested that we could rearrange our state machine to
> hide the differences in generation. gcc also prefers the explicit state
> machine, so make it so:
>
> process_csb 1980 1967 -13
>
> Suggested-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
> Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
> Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
> Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
> Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
> ---
> drivers/gpu/drm/i915/gt/intel_lrc.c | 64 ++++++++++++++++++++---------
> 1 file changed, 44 insertions(+), 20 deletions(-)
>
> diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
> index 471e134de186..953b3938a85f 100644
> --- a/drivers/gpu/drm/i915/gt/intel_lrc.c
> +++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
> @@ -1279,6 +1279,30 @@ reset_in_progress(const struct intel_engine_execlists *execlists)
> return unlikely(!__tasklet_is_enabled(&execlists->tasklet));
> }
>
> +enum csb_step {
> + CSB_NOP,
> + CSB_PROMOTE,
> + CSB_PREEMPT,
> + CSB_COMPLETE,
> +};
> +
> +static inline enum csb_step
> +csb_parse(const struct intel_engine_execlists *execlists, const u32 *csb)
> +{
> + unsigned int status = *csb;
Could be const u32 aswell (stylistic).
Just makes me ponder why you want to read csb in here
and not in the callsite.
Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>
> +
> + if (status & GEN8_CTX_STATUS_IDLE_ACTIVE)
> + return CSB_PROMOTE;
> +
> + if (status & GEN8_CTX_STATUS_PREEMPTED)
> + return CSB_PREEMPT;
> +
> + if (*execlists->active)
> + return CSB_COMPLETE;
> +
> + return CSB_NOP;
> +}
> +
> static void process_csb(struct intel_engine_cs *engine)
> {
> struct intel_engine_execlists * const execlists = &engine->execlists;
> @@ -1316,8 +1340,6 @@ static void process_csb(struct intel_engine_cs *engine)
> rmb();
>
> do {
> - unsigned int status;
> -
> if (++head == num_entries)
> head = 0;
>
> @@ -1343,10 +1365,16 @@ static void process_csb(struct intel_engine_cs *engine)
> engine->name, head,
> buf[2 * head + 0], buf[2 * head + 1]);
>
> - status = buf[2 * head];
> - if (status & GEN8_CTX_STATUS_IDLE_ACTIVE) {
> + switch (csb_parse(execlists, buf + 2 * head)) {
> + case CSB_PREEMPT: /* cancel old inflight, prepare for switch */
> + trace_ports(execlists, "preempted", execlists->active);
> +
> + while (*execlists->active)
> + execlists_schedule_out(*execlists->active++);
> +
> + /* fallthrough */
> + case CSB_PROMOTE: /* switch pending to inflight */
> GEM_BUG_ON(*execlists->active);
> -promote:
> GEM_BUG_ON(!assert_pending_valid(execlists, "promote"));
> execlists->active =
> memcpy(execlists->inflight,
> @@ -1355,25 +1383,17 @@ static void process_csb(struct intel_engine_cs *engine)
> sizeof(*execlists->pending));
> execlists->pending[0] = NULL;
>
> + trace_ports(execlists, "promoted", execlists->active);
> +
> if (enable_timeslice(engine))
> mod_timer(&execlists->timer, jiffies + 1);
>
> if (!inject_preempt_hang(execlists))
> ring_set_paused(engine, 0);
> - } else if (status & GEN8_CTX_STATUS_PREEMPTED) {
> - struct i915_request * const *port = execlists->active;
> -
> - trace_ports(execlists, "preempted", execlists->active);
> -
> - while (*port)
> - execlists_schedule_out(*port++);
> -
> - goto promote;
> - } else if (*execlists->active) {
> - struct i915_request *rq = *execlists->active++;
> + break;
>
> - trace_ports(execlists, "completed",
> - execlists->active - 1);
> + case CSB_COMPLETE: /* port0 completed, advanced to port1 */
> + trace_ports(execlists, "completed", execlists->active);
>
> /*
> * We rely on the hardware being strongly
> @@ -1381,11 +1401,15 @@ static void process_csb(struct intel_engine_cs *engine)
> * coherent (visible from the CPU) before the
> * user interrupt and CSB is processed.
> */
> - GEM_BUG_ON(!i915_request_completed(rq));
> - execlists_schedule_out(rq);
> + GEM_BUG_ON(!i915_request_completed(*execlists->active));
> + execlists_schedule_out(*execlists->active++);
>
> GEM_BUG_ON(execlists->active - execlists->inflight >
> execlists_num_ports(execlists));
> + break;
> +
> + case CSB_NOP:
> + break;
> }
> } while (head != tail);
>
> --
> 2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@lists.freedesktop.org
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
next prev parent reply other threads:[~2019-07-01 11:49 UTC|newest]
Thread overview: 23+ messages / expand[flat|nested] mbox.gz Atom feed top
2019-07-01 10:04 [PATCH 01/12] drm/i915/guc: Avoid reclaim locks during reset Chris Wilson
2019-07-01 10:04 ` [PATCH 02/12] drm/i915: Markup potential lock for i915_active Chris Wilson
2019-07-01 10:04 ` [PATCH 03/12] drm/i915: Mark up vma->active as safe for use inside shrinkers Chris Wilson
2019-07-01 10:04 ` [PATCH 04/12] drm/i915/execlists: Refactor CSB state machine Chris Wilson
2019-07-01 11:49 ` Mika Kuoppala [this message]
2019-07-01 13:50 ` Chris Wilson
2019-07-02 8:36 ` Mika Kuoppala
2019-07-01 18:28 ` Daniele Ceraolo Spurio
2019-07-01 10:04 ` [PATCH 05/12] drm/i915/execlists: Hesitate before slicing Chris Wilson
2019-07-01 10:04 ` [PATCH 06/12] drm/i915/selftests: Lock the drm_mm while modifying Chris Wilson
2019-07-01 10:04 ` [PATCH 07/12] drm/i915: Teach execbuffer to take the engine wakeref not GT Chris Wilson
2019-07-01 10:04 ` [PATCH 08/12] drm/i915/gt: Track timeline activeness in enter/exit Chris Wilson
2019-07-01 10:04 ` [PATCH 09/12] drm/i915/gt: Convert timeline tracking to spinlock Chris Wilson
2019-07-01 10:05 ` [PATCH 10/12] drm/i915/gt: Guard timeline pinning with its own mutex Chris Wilson
2019-07-01 10:05 ` [PATCH 11/12] drm/i915: Protect request retirement with timeline->mutex Chris Wilson
2019-07-01 10:05 ` [PATCH 12/12] drm/i915: Replace struct_mutex for batch pool serialisation Chris Wilson
2019-07-01 11:15 ` ✗ Fi.CI.CHECKPATCH: warning for series starting with [01/12] drm/i915/guc: Avoid reclaim locks during reset Patchwork
2019-07-01 11:20 ` ✗ Fi.CI.SPARSE: " Patchwork
2019-07-01 11:44 ` ✓ Fi.CI.BAT: success " Patchwork
2019-07-01 12:36 ` [PATCH 01/12] " Michal Wajdeczko
2019-07-01 13:48 ` Chris Wilson
2019-07-01 18:12 ` Daniele Ceraolo Spurio
2019-07-02 14:06 ` ✓ Fi.CI.IGT: success for series starting with [01/12] " 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=875zom3rfn.fsf@gaia.fi.intel.com \
--to=mika.kuoppala@linux.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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox