test_power_draw: wait for power instead of fixed warmup time (#31318)

* wait for valid

* reuuse that

* assert warmup

* Wip

* fix that

* reset

* none

* timeout

* maxlen

* consecutive

* proc

* fix

* local

* same as before

* mean

* fixes

* SA

* this is the real warmup time

* this is the real warmup time

* Cleanup

* Cleanup

* monotonic + len

* reduce duplication

* fix

* fix

---------

Co-authored-by: Comma Device <device@comma.ai>
old-commit-hash: caa9dff610

system/hardware/tici/tests/test_power_draw.py

@@ -1,4 +1,6 @@
 #!/usr/bin/env python3
+from collections import defaultdict, deque
+import sys
 import pytest
 import unittest
 import time
@@ -15,7 +17,8 @@ from openpilot.system.hardware.tici.power_monitor import get_power
 from openpilot.selfdrive.manager.process_config import managed_processes
 from openpilot.selfdrive.manager.manager import manager_cleanup
 
-SAMPLE_TIME = 8   # seconds to sample power
+SAMPLE_TIME = 8       # seconds to sample power
+MAX_WARMUP_TIME = 30  # seconds to wait for SAMPLE_TIME consecutive valid samples
 
 @dataclass
 class Proc:
@@ -24,7 +27,6 @@ class Proc:
   msgs: List[str]
   rtol: float = 0.05
   atol: float = 0.12
-  warmup: float = 6.
 
 PROCS = [
   Proc('camerad', 2.1, msgs=['roadCameraState', 'wideRoadCameraState', 'driverCameraState']),
@@ -48,41 +50,82 @@ class TestPowerDraw(unittest.TestCase):
   def tearDown(self):
     manager_cleanup()
 
+  def get_expected_messages(self, proc):
+    return int(sum(SAMPLE_TIME * SERVICE_LIST[msg].frequency for msg in proc.msgs))
+
+  def valid_msg_count(self, proc, msg_counts):
+    msgs_received = sum(msg_counts[msg] for msg in proc.msgs)
+    msgs_expected = self.get_expected_messages(proc)
+    return np.core.numeric.isclose(msgs_expected, msgs_received, rtol=.02, atol=2)
+
+  def valid_power_draw(self, proc, used):
+    return np.core.numeric.isclose(used, proc.power, rtol=proc.rtol, atol=proc.atol)
+
+  def tabulate_msg_counts(self, msgs_and_power):
+    msg_counts = defaultdict(lambda: 0)
+    for _, counts in msgs_and_power:
+      for msg, count in counts.items():
+        msg_counts[msg] += count
+    return msg_counts
+
+  def get_power_with_warmup_for_target(self, proc, prev):
+    socks = {msg: messaging.sub_sock(msg) for msg in proc.msgs}
+    for sock in socks.values():
+      messaging.drain_sock_raw(sock)
+
+    msgs_and_power = deque([], maxlen=SAMPLE_TIME)
+
+    start_time = time.monotonic()
+
+    while (time.monotonic() - start_time) < MAX_WARMUP_TIME:
+      power = get_power(1)
+      iteration_msg_counts = {}
+      for msg,sock in socks.items():
+        iteration_msg_counts[msg] = len(messaging.drain_sock_raw(sock))
+      msgs_and_power.append((power, iteration_msg_counts))
+
+      if len(msgs_and_power) < SAMPLE_TIME:
+        continue
+
+      msg_counts = self.tabulate_msg_counts(msgs_and_power)
+      now = np.mean([m[0] for m in msgs_and_power])
+
+      if self.valid_msg_count(proc, msg_counts) and self.valid_power_draw(proc, now - prev):
+        break
+
+    return now, msg_counts, time.monotonic() - start_time - SAMPLE_TIME
+
   @mock_messages(['liveLocationKalman'])
   def test_camera_procs(self):
     baseline = get_power()
 
     prev = baseline
     used = {}
+    warmup_time = {}
     msg_counts = {}
-    for proc in PROCS:
-      socks = {msg: messaging.sub_sock(msg) for msg in proc.msgs}
-      managed_processes[proc.name].start()
-      time.sleep(proc.warmup)
-      for sock in socks.values():
-        messaging.drain_sock_raw(sock)
 
-      now = get_power(SAMPLE_TIME)
+    for proc in PROCS:
+      managed_processes[proc.name].start()
+      now, local_msg_counts, warmup_time[proc.name] = self.get_power_with_warmup_for_target(proc, prev)
+      msg_counts.update(local_msg_counts)
+
       used[proc.name] = now - prev
       prev = now
-      for msg,sock in socks.items():
-        msg_counts[msg] = len(messaging.drain_sock_raw(sock))
 
     manager_cleanup()
 
-    tab = [['process', 'expected (W)', 'measured (W)', '# msgs expected', '# msgs received']]
+    tab = [['process', 'expected (W)', 'measured (W)', '# msgs expected', '# msgs received', "warmup time (s)"]]
     for proc in PROCS:
       cur = used[proc.name]
       expected = proc.power
       msgs_received = sum(msg_counts[msg] for msg in proc.msgs)
-      msgs_expected = int(sum(SAMPLE_TIME * SERVICE_LIST[msg].frequency for msg in proc.msgs))
-      tab.append([proc.name, round(expected, 2), round(cur, 2), msgs_expected, msgs_received])
+      tab.append([proc.name, round(expected, 2), round(cur, 2), self.get_expected_messages(proc), msgs_received, round(warmup_time[proc.name], 2)])
       with self.subTest(proc=proc.name):
-        np.testing.assert_allclose(msgs_expected, msgs_received, rtol=.02, atol=2)
-        np.testing.assert_allclose(cur, expected, rtol=proc.rtol, atol=proc.atol)
+        self.assertTrue(self.valid_msg_count(proc, msg_counts), f"expected {self.get_expected_messages(proc)} msgs, got {msgs_received} msgs")
+        self.assertTrue(self.valid_power_draw(proc, cur), f"expected {expected:.2f}W, got {cur:.2f}W")
     print(tabulate(tab))
     print(f"Baseline {baseline:.2f}W\n")
 
 
 if __name__ == "__main__":
-  pytest.main()
+  pytest.main(sys.argv)