sai-cli diagnose: plot CPU usage if available (#75)

* sai-cli diagnose: plot CPU usage if available

* Clean up

* Fix typo

Author: kaatrasa

python/cli/diagnose/diagnose.py

@@ -43,7 +43,7 @@ def generateReport(args):
         'magnetometer': {"v": [], "t": [], "td": []},
         'barometer': {"v": [], "t": [], "td": []},
         'gnss': {"v": [], "t": [], "td": []},
-        'cpu': {"v": [], "t": []},
+        'cpu': {"v": [], "t": [], "td": [], "processes": {}},
         'cameras': {}
     }
 
@@ -120,8 +120,19 @@ def addMeasurement(type, t, v):
                     if "features" in f: cameras[ind]["features"].append(len(f["features"]))
                     cameras[ind]["t"].append(t)
             elif metrics is not None and 'cpu' in metrics:
-                data["cpu"]["t"].append(t)
-                data["cpu"]["v"].append(metrics['cpu'].get('systemTotalUsagePercent', 0))
+                addMeasurement("cpu", t, metrics['cpu'].get('systemTotalUsagePercent', 0))
+                usedProcessNames = {}  # Track duplicate process names
+                for process in metrics['cpu'].get('processes', []):
+                    name = process.get('name')
+                    if not name: continue
+
+                    count = usedProcessNames.get(name, 0)
+                    usedProcessNames[name] = count + 1
+                    uniqueName = f"{name} {count + 1}" if count else name
+
+                    processData = data['cpu']["processes"].setdefault(uniqueName, {"v": [], "t": []})
+                    processData['v'].append(process['usagePercent'])
+                    processData['t'].append(t)
 
         if nSkipped > 0: print(f'Skipped {nSkipped} lines')
 
@@ -131,7 +142,7 @@ def addMeasurement(type, t, v):
     diagnoseMagnetometer(data, output)
     diagnoseBarometer(data, output)
     diagnoseGNSS(data, output)
-    diagnoseCpu(data, output)
+    diagnoseCPU(data, output)
 
     if os.path.dirname(args.output_html):
         os.makedirs(os.path.dirname(args.output_html), exist_ok=True)

python/cli/diagnose/html.py

@@ -166,7 +166,7 @@ def generateHtml(output, outputHtml):
                     camera["frequency"],
                     camera["count"])))
 
-    SENSOR_NAMES = ["accelerometer", "gyroscope", "magnetometer", "barometer", "GNSS"]
+    SENSOR_NAMES = ["accelerometer", "gyroscope", "magnetometer", "barometer", "GNSS", "CPU"]
     for sensor in SENSOR_NAMES:
         if sensor not in output: continue
         kvPairs.append((

python/cli/diagnose/sensors.py

@@ -294,7 +294,7 @@ def highpass(signal, fs, cutoff, order=3):
         plt.ylabel(measurementUnit)
 
         fig.suptitle(f"Preview of {sensorName} signal noise (mean={noiseScale:.1f}{measurementUnit}, threshold={noiseThreshold:.1f}{measurementUnit})")
-        fig.tight_layout(rect=[0, 0.03, 1, 0.95])
+        fig.tight_layout()
         self.images.append(base64(fig))
 
         if noiseScale > noiseThreshold:
@@ -365,6 +365,50 @@ def analyzeAccelerometerSignalHasGravity(self, signal):
                 "This suggests the signal may be missing gravitational acceleration."
             )
 
+    def analyzeCpuUsage(self, signal, timestamps, processes):
+        CPU_USAGE_THRESHOLD = 90.0
+
+        mean = np.mean(signal)
+        p95 = np.percentile(signal, 95)
+        p99 = np.percentile(signal, 99)
+
+        if mean > CPU_USAGE_THRESHOLD:
+            self.__addIssue(DiagnosisLevel.WARNING,
+                f"Average CPU usage {mean:.1f}% is above the threshold ({CPU_USAGE_THRESHOLD:.1f}%)."
+            )
+        elif p95 > CPU_USAGE_THRESHOLD:
+            self.__addIssue(DiagnosisLevel.WARNING,
+                f"95th percentile CPU usage {p95:.1f}% is above the threshold ({CPU_USAGE_THRESHOLD:.1f}%)."
+            )
+        elif p99 > CPU_USAGE_THRESHOLD:
+            self.__addIssue(DiagnosisLevel.WARNING,
+                f"99th percentile CPU usage {p99:.1f}% is above the threshold ({CPU_USAGE_THRESHOLD:.1f}%)."
+            )
+
+        import matplotlib.pyplot as plt
+        fig, ax = plt.subplots(figsize=(8, 6))
+
+        ax.plot(timestamps, signal, label="System total", linestyle='-')
+        ax.set_title("CPU usage")
+        ax.set_ylabel("CPU usage (%)")
+        ax.set_xlabel("Time (s)")
+
+        legend = ['System total']
+        ylim = 100
+        for name, data in processes.items():
+            if len(data['v']) == 0: continue
+            ax.plot(data['t'], data['v'], label=name, linestyle='--')
+            legend.append(name)
+            ylim = max(ylim, np.max(data['v']) * 1.1)
+
+        ax.set_ylim(0, ylim)
+
+        leg = ax.legend(legend, fontsize='large', markerscale=10)
+        for line in leg.get_lines(): line.set_linewidth(2)
+
+        fig.tight_layout()
+        self.images.append(base64(fig))
+
     def serializeIssues(self):
         self.issues = sorted(self.issues, key=lambda x: x[0], reverse=True)
         return [{
@@ -705,13 +749,22 @@ def diagnoseGNSS(data, output):
     if status.diagnosis == DiagnosisLevel.ERROR:
         output["passed"] = False
 
-def diagnoseCpu(data, output):
-    data = data["cpu"]
-    timestamps = np.array(data["t"])
-    values = data["v"]
+def diagnoseCPU(data, output):
+    sensor = data["cpu"]
+    timestamps = np.array(sensor["t"])
+    deltaTimes = np.array(sensor["td"])
+    signal = np.array(sensor['v'])
+    processes = sensor["processes"]
 
     if len(timestamps) == 0: return
 
-    output["cpu"] = {
-        "image": plotFrame(timestamps, values, "CPU system load (%)", ymin=0, ymax=100)
+    status = Status()
+    status.analyzeCpuUsage(signal, timestamps, processes)
+
+    output["CPU"] = {
+        "diagnosis": status.diagnosis.toString(),
+        "issues": status.serializeIssues(),
+        "frequency": computeSamplingRate(deltaTimes),
+        "count": len(timestamps),
+        "images": status.images
     }