pps: make pps-producer RT only on freedreno

Calling sched_setscheduler twice every sample period has high CPU
overhead.  For intel and panfrost, their dump_perfcnt is preemptible and
they don't need the scheduler change.

For freedreno, simply makes the main thread RT at all time.  This solves
most of the cpu overhead issue.

v2: removed pthread_t param and just change the scheduler for the
    calling thread

Acked-by: Rob Clark <robdclark@chromium.org> (v1)
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/19668>

src/freedreno/ds/fd_pps_driver.cc

@@ -37,6 +37,12 @@ percent(uint64_t a, uint64_t b)
    return 100.f * (a / static_cast<double>(b));
 }
 
+bool
+FreedrenoDriver::is_dump_perfcnt_preemptible() const
+{
+   return false;
+}
+
 uint64_t
 FreedrenoDriver::get_min_sampling_period_ns()
 {

src/freedreno/ds/fd_pps_driver.h

@@ -20,6 +20,7 @@ namespace pps
 class FreedrenoDriver : public Driver
 {
 public:
+   bool is_dump_perfcnt_preemptible() const override;
    uint64_t get_min_sampling_period_ns() override;
    bool init_perfcnt() override;
    void enable_counter(uint32_t counter_id) override;

src/tool/pps/pps.cc

@@ -10,6 +10,8 @@
 #include <cerrno>
 #include <cstring>
 
+#include <sched.h>
+
 namespace pps
 {
 bool check(int res, const char *msg)
@@ -23,4 +25,14 @@ bool check(int res, const char *msg)
    return true;
 }
 
+void make_thread_rt()
+{
+   // Use FIFO policy to avoid preemption while collecting counters
+   int sched_policy = SCHED_FIFO;
+   // Do not use max priority to avoid starving migration and watchdog threads
+   int priority_value = sched_get_priority_max(sched_policy) - 1;
+   sched_param priority_param { priority_value };
+   sched_setscheduler(0, sched_policy, &priority_param);
+}
+
 } // namespace pps

src/tool/pps/pps.h

@@ -27,6 +27,8 @@ enum class State {
 /// @return True if ok, false otherwise
 bool check(int res, const char *msg);
 
+void make_thread_rt();
+
 /// @param num Numerator
 /// @param den Denominator
 /// @return A ratio between two floating point numbers, or 0 if the denominator is 0

src/tool/pps/pps_datasource.cc

@@ -306,18 +306,6 @@ void GpuDataSource::trace(TraceContext &ctx)
       state->was_cleared = false;
    }
 
-   // Save current scheduler for restoring later
-   int prev_sched_policy = sched_getscheduler(0);
-   sched_param prev_priority_param;
-   sched_getparam(0, &prev_priority_param);
-
-   // Use FIFO policy to avoid preemption while collecting counters
-   int sched_policy = SCHED_FIFO;
-   // Do not use max priority to avoid starving migration and watchdog threads
-   int priority_value = sched_get_priority_max(sched_policy) - 1;
-   sched_param priority_param { priority_value };
-   sched_setscheduler(0, sched_policy, &priority_param);
-
    if (driver->dump_perfcnt()) {
       while (auto gpu_timestamp = driver->next()) {
          if (gpu_timestamp <= descriptor_gpu_timestamp) {
@@ -346,9 +334,6 @@ void GpuDataSource::trace(TraceContext &ctx)
       add_timestamp(event, driver);
       last_correlation_timestamp = cpu_ts;
    }
-
-   // Reset normal scheduler
-   sched_setscheduler(0, prev_sched_policy, &prev_priority_param);
 }
 
 void GpuDataSource::trace_callback(TraceContext ctx)

src/tool/pps/pps_driver.h

@@ -45,6 +45,9 @@ class Driver
    Driver(const Driver &) = delete;
    Driver &operator=(const Driver &) = delete;
 
+   /// @return Whether dump_perfcnt is preemptible
+   virtual bool is_dump_perfcnt_preemptible() const { return true; }
+
    /// @return The minimum sampling period for the current device
    virtual uint64_t get_min_sampling_period_ns() = 0;
 

src/tool/pps/pps_producer.cc

@@ -24,6 +24,10 @@ int main(int argc, const char **argv)
       (argc > 1) ? Driver::find_driver_name(argv[1]) : Driver::default_driver_name();
    GpuDataSource::register_data_source(driver_name);
 
+   const auto &driver = Driver::get_supported_drivers().at(driver_name);
+   if (!driver->is_dump_perfcnt_preemptible())
+      make_thread_rt();
+
    while (true) {
       GpuDataSource::wait_started();
       GpuDataSource::Trace(GpuDataSource::trace_callback);