Re: [Intel-gfx] [PATCH] drm/i915/gt: Fix up clock frequency

[Mika Kuoppala <mika.kuoppala@linux.intel.com>]

Chris Wilson <chris@chris-wilson.co.uk> writes:

> Quoting Mika Kuoppala (2020-04-27 13:15:23)
>> Chris Wilson <chris@chris-wilson.co.uk> writes:
>> 
>> > The bspec lists both the clock frequency and the effective interval. The
>> > interval corresponds to observed behaviour, so adjust the frequency to
>> > match.
>> >
>> > v2: Mika rightfully asked if we could measure the clock frequency from a
>> > selftest.
>> 
>> This kind of delivery times might impact on demand :P
>> 
>> >
>> > Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>
>> > Cc: Mika Kuoppala <mika.kuoppala@linux.intel.com>
>> > ---
>> >  .../gpu/drm/i915/gt/intel_gt_clock_utils.c    |  12 +-
>> >  drivers/gpu/drm/i915/gt/selftest_gt_pm.c      |   1 +
>> >  drivers/gpu/drm/i915/gt/selftest_rps.c        | 125 ++++++++++++++++++
>> >  drivers/gpu/drm/i915/gt/selftest_rps.h        |   1 +
>> >  4 files changed, 133 insertions(+), 6 deletions(-)
>> >
>> > diff --git a/drivers/gpu/drm/i915/gt/intel_gt_clock_utils.c b/drivers/gpu/drm/i915/gt/intel_gt_clock_utils.c
>> > index 852a7d731b3b..999079686846 100644
>> > --- a/drivers/gpu/drm/i915/gt/intel_gt_clock_utils.c
>> > +++ b/drivers/gpu/drm/i915/gt/intel_gt_clock_utils.c
>> > @@ -7,9 +7,9 @@
>> >  #include "intel_gt.h"
>> >  #include "intel_gt_clock_utils.h"
>> >  
>> > +#define MHZ_12   12000000 /* 12MHz (24MHz/2), 83.333ns */
>> > +#define MHZ_12_5 12500000 /* 12.5MHz (25MHz/2), 80ns */
>> >  #define MHZ_19_2 19200000 /* 19.2MHz, 52.083ns */
>> > -#define MHZ_24 24000000 /* 24MHz, 83.333ns */
>> > -#define MHZ_25 25000000 /* 25MHz, 80ns */
>> >  
>> >  static u32 read_clock_frequency(const struct intel_gt *gt)
>> >  {
>> > @@ -21,19 +21,19 @@ static u32 read_clock_frequency(const struct intel_gt *gt)
>> >               config >>= GEN11_RPM_CONFIG0_CRYSTAL_CLOCK_FREQ_SHIFT;
>> >  
>> >               switch (config) {
>> > -             case 0: return MHZ_24;
>> > +             case 0: return MHZ_12;
>> >               case 1:
>> >               case 2: return MHZ_19_2;
>> >               default:
>> > -             case 3: return MHZ_25;
>> > +             case 3: return MHZ_12_5;
>> >               }
>> >       } else if (INTEL_GEN(gt->i915) >= 9) {
>> >               if (IS_GEN9_LP(gt->i915))
>> >                       return MHZ_19_2;
>> >               else
>> > -                     return MHZ_24;
>> > +                     return MHZ_12;
>> >       } else {
>> > -             return MHZ_25;
>> > +             return MHZ_12_5;
>> >       }
>> >  }
>> >  
>> > diff --git a/drivers/gpu/drm/i915/gt/selftest_gt_pm.c b/drivers/gpu/drm/i915/gt/selftest_gt_pm.c
>> > index e02fdec58826..242181a5214c 100644
>> > --- a/drivers/gpu/drm/i915/gt/selftest_gt_pm.c
>> > +++ b/drivers/gpu/drm/i915/gt/selftest_gt_pm.c
>> > @@ -53,6 +53,7 @@ int intel_gt_pm_live_selftests(struct drm_i915_private *i915)
>> >  {
>> >       static const struct i915_subtest tests[] = {
>> >               SUBTEST(live_rc6_manual),
>> > +             SUBTEST(live_rps_clock_interval),
>> >               SUBTEST(live_rps_control),
>> >               SUBTEST(live_rps_frequency_cs),
>> >               SUBTEST(live_rps_frequency_srm),
>> > diff --git a/drivers/gpu/drm/i915/gt/selftest_rps.c b/drivers/gpu/drm/i915/gt/selftest_rps.c
>> > index e13cbcb82825..fe92c55572c1 100644
>> > --- a/drivers/gpu/drm/i915/gt/selftest_rps.c
>> > +++ b/drivers/gpu/drm/i915/gt/selftest_rps.c
>> > @@ -208,6 +208,131 @@ static void show_pstate_limits(struct intel_rps *rps)
>> >       }
>> >  }
>> >  
>> > +int live_rps_clock_interval(void *arg)
>> > +{
>> > +     struct intel_gt *gt = arg;
>> > +     struct intel_rps *rps = &gt->rps;
>> > +     void (*saved_work)(struct work_struct *wrk);
>> > +     struct intel_engine_cs *engine;
>> > +     enum intel_engine_id id;
>> > +     struct igt_spinner spin;
>> > +     int err = 0;
>> > +
>> > +     if (!rps->enabled)
>> > +             return 0;
>> > +
>> > +     intel_gt_check_clock_frequency(gt);
>> > +
>> > +     if (igt_spinner_init(&spin, gt))
>> > +             return -ENOMEM;
>> > +
>> > +     intel_gt_pm_wait_for_idle(gt);
>> > +     saved_work = rps->work.func;
>> > +     rps->work.func = dummy_rps_work;
>> > +
>> > +     intel_gt_pm_get(gt);
>> > +     intel_rps_disable(&gt->rps);
>> > +
>> > +     for_each_engine(engine, gt, id) {
>> > +             unsigned long saved_heartbeat;
>> > +             struct i915_request *rq;
>> > +             u32 cycles, expected;
>> > +             ktime_t dt;
>> > +             u64 time;
>> > +
>> > +             if (!intel_engine_can_store_dword(engine))
>> > +                     continue;
>> > +
>> > +             saved_heartbeat = engine_heartbeat_disable(engine);
>> > +
>> > +             rq = igt_spinner_create_request(&spin,
>> > +                                             engine->kernel_context,
>> > +                                             MI_NOOP);
>> > +             if (IS_ERR(rq)) {
>> > +                     err = PTR_ERR(rq);
>> > +                     break;
>> > +             }
>> > +
>> > +             i915_request_add(rq);
>> > +
>> > +             if (!igt_wait_for_spinner(&spin, rq)) {
>> > +                     pr_err("%s: RPS spinner did not start\n",
>> > +                            engine->name);
>> > +                     igt_spinner_end(&spin);
>> > +                     engine_heartbeat_enable(engine, saved_heartbeat);
>> > +                     intel_gt_set_wedged(engine->gt);
>> > +                     err = -EIO;
>> > +                     break;
>> > +             }
>> > +
>> > +             intel_uncore_write(gt->uncore, GEN6_RP_CUR_UP_EI, 0);
>> > +
>> > +             /* Set the evaluation interval to infinity! */
>> > +             intel_uncore_write(gt->uncore,
>> > +                                GEN6_RP_UP_EI, 0xffffffff);
>> > +             intel_uncore_write(gt->uncore,
>> > +                                GEN6_RP_UP_THRESHOLD, 0xffffffff);
>> 
>> To keep it constant during the calculation?
>
> Yeah, it shouldn't conceptually make a difference. I'm just worried in
> case the interrupt and the EI stopped when it crosses threshold, as
> opposed to evaluating the counters strictly at the end of the EI.
>
>> > +
>> > +             intel_uncore_write(gt->uncore, GEN6_RP_CONTROL,
>> > +                                GEN6_RP_ENABLE | GEN6_RP_UP_BUSY_AVG);
>> > +
>> > +             if (wait_for(intel_uncore_read(gt->uncore, GEN6_RP_CUR_UP_EI),
>> > +                          10)) {
>> > +                     pr_err("%s: rps evalution interval not ticking\n",
>> > +                            engine->name);
>> > +                     igt_spinner_end(&spin);
>> > +                     engine_heartbeat_enable(engine, saved_heartbeat);
>> > +                     err = -ENODEV;
>> > +                     break;
>> > +             }
>> > +
>> > +             dt = ktime_get();
>> > +             cycles = -intel_uncore_read(gt->uncore, GEN6_RP_CUR_UP_EI);
>> 
>> Hmm you have not zeroed the cycles on the first run, but does it matter
>> is then the question. 
>
> 	cycles = -X, 
> not
> 	cycles -= X

Ah indeed. A bit of trickery but tricks are useful.
For the bspec, I didn't find a note about the actual intervals.

But evidence dictates.

Acked-by: Mika Kuoppala <mika.kuoppala@linux.intel.com>

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