We have a hole in our busy-stat accounting if the pmu is enabled during a long running batch, the pmu will not start accumulating busy-time until the next context switch. This then fails tests that are only sampling a single batch.
Testcase: igt/perf_pmu/start-busy # To Be Written Signed-off-by: Chris Wilson <[email protected]> Cc: Tvrtko Ursulin <[email protected]> --- We should also construct a queue of batches just to complicate things. spin1 = spin_batch; for (QLEN - 1) { rvsd1 = create_context; resubmit_spin1_batch; delete_context(rsvd1); } rsvd1 = create_context; spin2 = spin_batch() delete_context(rsvd1); start_pmu cancel_spin1 usleep(batch_duration) cancel_spin2 end_pmu assert(busy == batch_duration); --- drivers/gpu/drm/i915/intel_engine_cs.c | 9 ++++++++- 1 file changed, 8 insertions(+), 1 deletion(-) diff --git a/drivers/gpu/drm/i915/intel_engine_cs.c b/drivers/gpu/drm/i915/intel_engine_cs.c index ebdcbcbacb3c..c2f01ff96ce1 100644 --- a/drivers/gpu/drm/i915/intel_engine_cs.c +++ b/drivers/gpu/drm/i915/intel_engine_cs.c @@ -1946,8 +1946,15 @@ int intel_enable_engine_stats(struct intel_engine_cs *engine) spin_lock_irqsave(&engine->stats.lock, flags); if (engine->stats.enabled == ~0) goto busy; - if (engine->stats.enabled++ == 0) + if (engine->stats.enabled++ == 0) { engine->stats.enabled_at = ktime_get(); + + /* XXX submission method oblivious */ + engine->stats.active = port_count(&engine->execlists.port[1]); + engine->stats.active += port_count(&engine->execlists.port[0]); + if (engine->stats.active) + engine->stats.start = engine->stats.enabled_at; + } spin_unlock_irqrestore(&engine->stats.lock, flags); return 0; -- 2.15.1 _______________________________________________ Intel-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/intel-gfx
