| Message ID | 20210108105608.18424-1-chris@chris-wilson.co.uk (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | drm/i915/selftests: Rearrange ktime_get to reduce latency against CS | expand |
Chris Wilson <chris@chris-wilson.co.uk> writes: > In our tests where we measure the elapsed time on both the CPU and CS > using a udelay, our CS results match the udelay much more accurately > than the ktime (even when using ktime_get_fast_ns). With preemption > disabled, we can go one step lower than ktime and use local_clock. > > Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/2919 > Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> > --- > drivers/gpu/drm/i915/gt/selftest_engine_pm.c | 4 ++-- > 1 file changed, 2 insertions(+), 2 deletions(-) > > diff --git a/drivers/gpu/drm/i915/gt/selftest_engine_pm.c b/drivers/gpu/drm/i915/gt/selftest_engine_pm.c > index ca080445695e..c3d965279fc3 100644 > --- a/drivers/gpu/drm/i915/gt/selftest_engine_pm.c > +++ b/drivers/gpu/drm/i915/gt/selftest_engine_pm.c > @@ -112,11 +112,11 @@ static int __measure_timestamps(struct intel_context *ce, > > /* Run the request for a 100us, sampling timestamps before/after */ > preempt_disable(); Do you need to promote this to local_irq_disable() ? -Mika > - *dt = ktime_get_raw_fast_ns(); > + *dt = local_clock(); > write_semaphore(&sema[2], 0); > udelay(100); > + *dt = local_clock() - *dt; > write_semaphore(&sema[2], 1); > - *dt = ktime_get_raw_fast_ns() - *dt; > preempt_enable(); > > if (i915_request_wait(rq, 0, HZ / 2) < 0) { > -- > 2.20.1 > > _______________________________________________ > Intel-gfx mailing list > Intel-gfx@lists.freedesktop.org > https://lists.freedesktop.org/mailman/listinfo/intel-gfx
Quoting Mika Kuoppala (2021-01-12 19:19:34) > Chris Wilson <chris@chris-wilson.co.uk> writes: > > > In our tests where we measure the elapsed time on both the CPU and CS > > using a udelay, our CS results match the udelay much more accurately > > than the ktime (even when using ktime_get_fast_ns). With preemption > > disabled, we can go one step lower than ktime and use local_clock. > > > > Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/2919 > > Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> > > --- > > drivers/gpu/drm/i915/gt/selftest_engine_pm.c | 4 ++-- > > 1 file changed, 2 insertions(+), 2 deletions(-) > > > > diff --git a/drivers/gpu/drm/i915/gt/selftest_engine_pm.c b/drivers/gpu/drm/i915/gt/selftest_engine_pm.c > > index ca080445695e..c3d965279fc3 100644 > > --- a/drivers/gpu/drm/i915/gt/selftest_engine_pm.c > > +++ b/drivers/gpu/drm/i915/gt/selftest_engine_pm.c > > @@ -112,11 +112,11 @@ static int __measure_timestamps(struct intel_context *ce, > > > > /* Run the request for a 100us, sampling timestamps before/after */ > > preempt_disable(); > > Do you need to promote this to local_irq_disable() ? Good suggestion. Will try to remember if we still see discrepancies... Interrupt handlers are meant to <5us, right??? -Chris
Chris Wilson <chris@chris-wilson.co.uk> writes: > Quoting Mika Kuoppala (2021-01-12 19:19:34) >> Chris Wilson <chris@chris-wilson.co.uk> writes: >> >> > In our tests where we measure the elapsed time on both the CPU and CS >> > using a udelay, our CS results match the udelay much more accurately >> > than the ktime (even when using ktime_get_fast_ns). With preemption >> > disabled, we can go one step lower than ktime and use local_clock. >> > >> > Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/2919 >> > Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> >> > --- >> > drivers/gpu/drm/i915/gt/selftest_engine_pm.c | 4 ++-- >> > 1 file changed, 2 insertions(+), 2 deletions(-) >> > >> > diff --git a/drivers/gpu/drm/i915/gt/selftest_engine_pm.c b/drivers/gpu/drm/i915/gt/selftest_engine_pm.c >> > index ca080445695e..c3d965279fc3 100644 >> > --- a/drivers/gpu/drm/i915/gt/selftest_engine_pm.c >> > +++ b/drivers/gpu/drm/i915/gt/selftest_engine_pm.c >> > @@ -112,11 +112,11 @@ static int __measure_timestamps(struct intel_context *ce, >> > >> > /* Run the request for a 100us, sampling timestamps before/after */ >> > preempt_disable(); >> >> Do you need to promote this to local_irq_disable() ? > > Good suggestion. Will try to remember if we still see discrepancies... > > Interrupt handlers are meant to <5us, right??? With both test types, we might sometimes find out what they are :) Reviewed-by: Mika Kuoppala <mika.kuoppala@linux.intel.com> > -Chris
diff --git a/drivers/gpu/drm/i915/gt/selftest_engine_pm.c b/drivers/gpu/drm/i915/gt/selftest_engine_pm.c index ca080445695e..c3d965279fc3 100644 --- a/drivers/gpu/drm/i915/gt/selftest_engine_pm.c +++ b/drivers/gpu/drm/i915/gt/selftest_engine_pm.c @@ -112,11 +112,11 @@ static int __measure_timestamps(struct intel_context *ce, /* Run the request for a 100us, sampling timestamps before/after */ preempt_disable(); - *dt = ktime_get_raw_fast_ns(); + *dt = local_clock(); write_semaphore(&sema[2], 0); udelay(100); + *dt = local_clock() - *dt; write_semaphore(&sema[2], 1); - *dt = ktime_get_raw_fast_ns() - *dt; preempt_enable(); if (i915_request_wait(rq, 0, HZ / 2) < 0) {
In our tests where we measure the elapsed time on both the CPU and CS using a udelay, our CS results match the udelay much more accurately than the ktime (even when using ktime_get_fast_ns). With preemption disabled, we can go one step lower than ktime and use local_clock. Closes: https://gitlab.freedesktop.org/drm/intel/-/issues/2919 Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk> --- drivers/gpu/drm/i915/gt/selftest_engine_pm.c | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-)