On 16.09.2024 11:03, Janusz Krzysztofik wrote: > When we expect an engine to be busy, we check if its reported busy time > falls within a +/-5% tolerance range of measurement time period. > However, when we expect the engine to be idle, we compare its reported > busy time against zero, still with a +/-5% tolerance range, but now > calculated against the zero value, then no tolerance at all. Obviously, > such check fails when the reported busy time is not exactly zero. > > Compare engine idle time against measurement time period instead of > comparing its busy time against zero when we expect the busy time to be > close to zero. As a debugging aid, display messages with the compared > values when requested via --debug option or when a failure occurs. > Reviewed-by: Peter Senna Tschudin <peter.senna@xxxxxxxxxxxxxxx> > Closes: https://gitlab.freedesktop.org/drm/i915/kernel/-/issues/7742 > Suggested-by: Chris Wilson <chris.p.wilson@xxxxxxxxxxxxxxx> > Signed-off-by: Janusz Krzysztofik <janusz.krzysztofik@xxxxxxxxxxxxxxx> > --- > tests/intel/drm_fdinfo.c | 51 ++++++++++++++++++++-------------------- > 1 file changed, 25 insertions(+), 26 deletions(-) > > diff --git a/tests/intel/drm_fdinfo.c b/tests/intel/drm_fdinfo.c > index 43216a64e..45d17aaaa 100644 > --- a/tests/intel/drm_fdinfo.c > +++ b/tests/intel/drm_fdinfo.c > @@ -107,12 +107,18 @@ static const char *engine_map[] = { > }; > > #define __assert_within_epsilon(x, ref, tol_up, tol_down) \ > - igt_assert_f((double)(x) <= (1.0 + (tol_up)) * (double)(ref) && \ > - (double)(x) >= (1.0 - (tol_down)) * (double)(ref), \ > - "'%s' != '%s' (%f not within +%.1f%%/-%.1f%% tolerance of %f)\n",\ > - #x, #ref, (double)(x), \ > - (tol_up) * 100.0, (tol_down) * 100.0, \ > - (double)(ref)) > + do { \ > + igt_assert_f((double)(x) <= (1.0 + (tol_up)) * (double)(ref) && \ > + (double)(x) >= (1.0 - (tol_down)) * (double)(ref), \ > + "'%s' != '%s' (%f not within +%.1f%%/-%.1f%% tolerance of %f)\n",\ > + #x, #ref, (double)(x), \ > + (tol_up) * 100.0, (tol_down) * 100.0, \ > + (double)(ref)); \ > + igt_debug("%f within +%.1f%%/-%.1f%% tolerance of %f\n",\ > + (double)(x), \ > + (tol_up) * 100.0, (tol_down) * 100.0, \ > + (double)(ref)); \ > + } while (0) > > #define assert_within_epsilon(x, ref, tolerance) \ > __assert_within_epsilon(x, ref, tolerance, tolerance) > @@ -241,10 +247,8 @@ single(int gem_fd, const intel_ctx_t *ctx, > else > end_spin(spin_fd, spin, FLAG_SYNC); > > - assert_within_epsilon(val, > - (flags & TEST_BUSY) && !(flags & TEST_ISOLATION) ? > - slept : 0.0f, > - tolerance); > + assert_within_epsilon((flags & TEST_BUSY) && !(flags & TEST_ISOLATION) ? val : slept - val, > + slept, tolerance); > > /* Check for idle after hang. */ > if (flags & FLAG_HANG) { > @@ -255,7 +259,7 @@ single(int gem_fd, const intel_ctx_t *ctx, > slept = measured_usleep(batch_duration_ns / 1000); > val = read_engine_time(gem_fd, e->class) - val; > > - assert_within_epsilon(val, 0, tolerance); > + assert_within_epsilon(slept - val, slept, tolerance); > } > > igt_spin_free(spin_fd, spin); > @@ -328,11 +332,8 @@ busy_check_all(int gem_fd, const intel_ctx_t *ctx, > > log_busy(num_classes, val); > > - for (i = 0; i < num_classes; i++) { > - double target = i == e->class ? slept : 0.0f; > - > - assert_within_epsilon(val[i], target, tolerance); > - } > + for (i = 0; i < num_classes; i++) > + assert_within_epsilon(i == e->class ? val[i] : slept - val[i], slept, tolerance); > > gem_quiescent_gpu(gem_fd); > } > @@ -405,9 +406,9 @@ most_busy_check_all(int gem_fd, const intel_ctx_t *ctx, > log_busy(num_classes, val); > > for (i = 0; i < num_classes; i++) { > - double target = slept * busy_class[i]; > + double target = slept * busy_class[i] ?: slept; > > - assert_within_epsilon(val[i], target, tolerance); > + assert_within_epsilon(busy_class[i] ? val[i] : slept - val[i], target, tolerance); > } > gem_quiescent_gpu(gem_fd); > } > @@ -460,9 +461,9 @@ all_busy_check_all(int gem_fd, const intel_ctx_t *ctx, > log_busy(num_classes, val); > > for (i = 0; i < num_classes; i++) { > - double target = slept * busy_class[i]; > + double target = slept * busy_class[i] ?: slept; > > - assert_within_epsilon(val[i], target, tolerance); > + assert_within_epsilon(busy_class[i] ? val[i] : slept - val[i], target, tolerance); > } > gem_quiescent_gpu(gem_fd); > } > @@ -601,10 +602,8 @@ virtual(int i915, const intel_ctx_cfg_t *base_cfg, unsigned int flags) > else > end_spin(i915, spin, FLAG_SYNC); > > - assert_within_epsilon(val, > - flags & TEST_BUSY ? > - slept : 0.0f, > - tolerance); > + assert_within_epsilon(flags & TEST_BUSY ? val : slept - val, > + slept, tolerance); > > /* Check for idle after hang. */ > if (flags & FLAG_HANG) { > @@ -616,7 +615,7 @@ virtual(int i915, const intel_ctx_cfg_t *base_cfg, unsigned int flags) > 1000); > val = read_engine_time(i915, class) - val; > > - assert_within_epsilon(val, 0, tolerance); > + assert_within_epsilon(slept - val, slept, tolerance); > } > > igt_spin_free(i915, spin); > @@ -724,7 +723,7 @@ virtual_all(int i915, const intel_ctx_cfg_t *base_cfg, unsigned int flags) > 1000); > val = read_engine_time(i915, class) - val; > > - assert_within_epsilon(val, 0, tolerance); > + assert_within_epsilon(slept - val, slept, tolerance); > } > > igt_spin_free(i915, spin);
