New settings for governors, +30% memory bandwidth (#2590)

* new settings for governors, +30% memory bandwidth

* CPU waste is better

* warm up frame_drop

* optional CORE arg for rtshield

* back off priorities of processes not critical to driving

* if it doesn't drive, it's not realtime

* err, resource. this should have been included

* reverse order to not spike

Co-authored-by: Comma Device <device@comma.ai>

launch_chffrplus.sh

@@ -19,6 +19,19 @@ function two_init {
   # openpilot gets all the cores
   echo 0-3 > /dev/cpuset/app/cpus
 
+  # set up governors
+  # +50mW offroad, +500mW onroad for 30% more RAM bandwidth
+  echo "performance" > /sys/class/devfreq/soc:qcom,cpubw/governor
+  echo 1056000 > /sys/class/devfreq/soc:qcom,m4m/max_freq
+  echo "performance" > /sys/class/devfreq/soc:qcom,m4m/governor
+
+  # unclear if these help, but they don't seem to hurt
+  echo "performance" > /sys/class/devfreq/soc:qcom,memlat-cpu0/governor
+  echo "performance" > /sys/class/devfreq/soc:qcom,memlat-cpu2/governor
+
+  # /sys/class/devfreq/soc:qcom,mincpubw is the only one left at "powersave"
+  # it seems to gain nothing but a wasted 500mW
+
   # Collect RIL and other possibly long-running I/O interrupts onto CPU 1
   echo 1 > /proc/irq/78/smp_affinity_list # qcom,smd-modem (LTE radio)
   echo 1 > /proc/irq/33/smp_affinity_list # ufshcd (flash storage)

scripts/waste.c

@@ -36,8 +36,23 @@ void waste(int pid) {
   double sec = seconds_since_boot();
   while (1) {
     for (int i = 0; i < 8; i++) {
-      for (int j = 0; j < 0x1000000; j+=2) {
-        out = vmlaq_f32(out, tmp[j], tmp[j+1]);
+      for (int j = 0; j < 0x1000000; j+=0x20) {
+        out = vmlaq_f32(out, tmp[j+0], tmp[j+1]);
+        out = vmlaq_f32(out, tmp[j+2], tmp[j+3]);
+        out = vmlaq_f32(out, tmp[j+4], tmp[j+5]);
+        out = vmlaq_f32(out, tmp[j+6], tmp[j+7]);
+        out = vmlaq_f32(out, tmp[j+8], tmp[j+9]);
+        out = vmlaq_f32(out, tmp[j+10], tmp[j+11]);
+        out = vmlaq_f32(out, tmp[j+12], tmp[j+13]);
+        out = vmlaq_f32(out, tmp[j+14], tmp[j+15]);
+        out = vmlaq_f32(out, tmp[j+16], tmp[j+17]);
+        out = vmlaq_f32(out, tmp[j+18], tmp[j+19]);
+        out = vmlaq_f32(out, tmp[j+20], tmp[j+21]);
+        out = vmlaq_f32(out, tmp[j+22], tmp[j+23]);
+        out = vmlaq_f32(out, tmp[j+24], tmp[j+25]);
+        out = vmlaq_f32(out, tmp[j+26], tmp[j+27]);
+        out = vmlaq_f32(out, tmp[j+28], tmp[j+29]);
+        out = vmlaq_f32(out, tmp[j+30], tmp[j+31]);
       }
     }
     double nsec = seconds_since_boot();

selfdrive/loggerd/loggerd.cc

@@ -599,7 +599,7 @@ int main(int argc, char** argv) {
   int err;
 
 #ifdef QCOM
-  set_realtime_priority(50);
+  setpriority(PRIO_PROCESS, 0, -12);
 #endif
 
   if (argc > 1 && strcmp(argv[1], "--bootlog") == 0) {

selfdrive/modeld/dmonitoringmodeld.cc

@@ -3,6 +3,7 @@
 #include <unistd.h>
 #include <signal.h>
 #include <cassert>
+#include <sys/resource.h>
 
 #include "common/visionbuf.h"
 #include "common/visionipc.h"
@@ -23,7 +24,7 @@ static void set_do_exit(int sig) {
 
 int main(int argc, char **argv) {
   int err;
-  set_realtime_priority(51);
+  setpriority(PRIO_PROCESS, 0, -15);
 
 #ifdef QCOM2
   set_core_affinity(5);

selfdrive/modeld/modeld.cc

@@ -157,6 +157,7 @@ int main(int argc, char **argv) {
     uint32_t frame_id = 0, last_vipc_frame_id = 0;
     double last = 0;
     int desire = -1;
+    uint32_t run_count = 0;
     while (!do_exit) {
       VIPCBuf *buf;
       VIPCBufExtra extra;
@@ -179,6 +180,8 @@ int main(int argc, char **argv) {
 
       double mt1 = 0, mt2 = 0;
       if (run_model_this_iter) {
+        run_count++;
+
         float vec_desire[DESIRE_LEN] = {0};
         if (desire >= 0 && desire < DESIRE_LEN) {
           vec_desire[desire] = 1.0;
@@ -198,6 +201,7 @@ int main(int argc, char **argv) {
         // tracked dropped frames
         uint32_t vipc_dropped_frames = extra.frame_id - last_vipc_frame_id - 1;
         frames_dropped = (1. - frame_filter_k) * frames_dropped + frame_filter_k * (float)std::min(vipc_dropped_frames, 10U);
+        if (run_count < 10) frames_dropped = 0;    // let frame drops warm up
         float frame_drop_ratio = frames_dropped / (1 + frames_dropped);
 
         model_publish(pm, extra.frame_id, frame_id,  vipc_dropped_frames, frame_drop_ratio, model_buf, extra.timestamp_eof, model_execution_time);

selfdrive/rtshield.py

@@ -1,16 +1,20 @@
 #!/usr/bin/env python3
+import os
 import time
 from common.realtime import set_core_affinity, set_realtime_priority
 
-
 # RT shield - ensure CPU 3 always remains available for RT processes
 #   runs as SCHED_FIFO with minimum priority to ensure kthreads don't
 #   get scheduled onto CPU 3, but it's always preemptible by realtime
 #   openpilot processes
 
 def main():
-  set_core_affinity(3)
+  set_core_affinity(int(os.getenv("CORE", "3")))
   set_realtime_priority(1)
 
   while True:
     time.sleep(0.000001)
+
+if __name__ == "__main__":
+  main()
+

selfdrive/thermald/thermald.py

@@ -377,7 +377,6 @@ def thermald_thread():
       if started_ts is None:
         started_ts = sec_since_boot()
         started_seen = True
-        os.system('echo performance > /sys/class/devfreq/soc:qcom,cpubw/governor')
     else:
       if startup_conditions["ignition"] and (startup_conditions != startup_conditions_prev):
         cloudlog.event("Startup blocked", startup_conditions=startup_conditions)
@@ -387,7 +386,6 @@ def thermald_thread():
       started_ts = None
       if off_ts is None:
         off_ts = sec_since_boot()
-        os.system('echo powersave > /sys/class/devfreq/soc:qcom,cpubw/governor')
 
     # Offroad power monitoring
     pm.calculate(health)