Allow individual energy/power traces to be exported to the file system

2 files changed, 57 insertions, 7 deletions

diff --git a/bin/analyze-archive.py b/bin/analyze-archive.py
index 770e5d5..f7847c5 100755
--- a/bin/analyze-archive.py
+++ b/bin/analyze-archive.py
@@ -21,6 +21,22 @@ Options:
     parameters. Also plots the corresponding measurements.
     If gplearn function is set, it is plotted using dashed lines.
 
+--export-traces=<directory>
+    Export power traces of all states and transitions to <directory>.
+    Creates a JSON file for each state and transition. Each JSON file
+    lists all occurences of the corresponding state/transition in the
+    benchmark's PTA trace. Each occurence contains the corresponding PTA
+    parameters (if any) in 'parameter' and measurement results in 'offline'.
+    As measurements are typically run repeatedly, 'offline' is in turn a list
+    of measurements: offline[0]['uW'] is the power trace of the first
+    measurement of this state/transition, offline[1]['uW'] corresponds t the
+    second measurement, etc. Values are provided in microwatts.
+    For example, TX.json[0].offline[0].uW corresponds to the first measurement
+    of the first TX state in the benchmark, and TX.json[5].offline[2].uW
+    corresponds to the third measurement of the sixth TX state in the benchmark.
+    WARNING: Several GB of RAM and disk space are required for complex measurements.
+             (JSON files may grow very large -- we trade efficiency for easy handling)
+
 --param-info
     Show parameter names and values
 
@@ -220,6 +236,7 @@ if __name__ == '__main__':
     show_quality = []
     pta = None
     energymodel_export_file = None
+    trace_export_dir = None
     xv_method = None
     xv_count = 10
 
@@ -227,6 +244,7 @@ if __name__ == '__main__':
         optspec = (
             'plot-unparam= plot-param= param-info show-models= show-quality= '
             'ignored-trace-indexes= discard-outliers= function-override= '
+            'export-traces= '
             'filter-param= '
             'cross-validate= '
             'with-safe-functions hwmodel= export-energymodel='
@@ -275,10 +293,26 @@ if __name__ == '__main__':
         print(err)
         sys.exit(2)
 
-    raw_data = RawData(args)
+    raw_data = RawData(args, with_traces=('export-traces' in opts))
 
     preprocessed_data = raw_data.get_preprocessed_data()
 
+    if 'export-traces' in opts:
+        uw_per_sot = dict()
+        for trace in preprocessed_data:
+            for state_or_transition in trace['trace']:
+                name = state_or_transition['name']
+                if name not in uw_per_sot:
+                    uw_per_sot[name] = list()
+                for elem in state_or_transition['offline']:
+                    elem['uW'] = list(elem['uW'])
+                uw_per_sot[name].append(state_or_transition)
+        for name, data in uw_per_sot.items():
+            target = f"{opts['export-traces']}/{name}.json"
+            print(f'exporting {target} ...')
+            with open(target, 'w') as f:
+                json.dump(data, f)
+
     if raw_data.preprocessing_stats['num_valid'] == 0:
         print('No valid data available. Abort.')
         sys.exit(2)
diff --git a/lib/dfatool.py b/lib/dfatool.py
index 177bd7b..8fb41a5 100644
--- a/lib/dfatool.py
+++ b/lib/dfatool.py
@@ -332,7 +332,7 @@ class CrossValidator:
 
 def _preprocess_mimosa(measurement):
     setup = measurement['setup']
-    mim = MIMOSA(float(setup['mimosa_voltage']), int(setup['mimosa_shunt']))
+    mim = MIMOSA(float(setup['mimosa_voltage']), int(setup['mimosa_shunt']), with_traces=measurement['with_traces'])
     try:
         charges, triggers = mim.load_data(measurement['content'])
         trigidx = mim.trigger_edges(triggers)
@@ -489,14 +489,16 @@ class RawData:
 
     Expects a specific trace format and UART log output (as produced by the
     dfatool benchmark generator). Loads data, prunes bogus measurements, and
-    provides preprocessed data suitable for PTAModel.
+    provides preprocessed data suitable for PTAModel. Results are cached on the
+    file system, making subsequent loads near-instant.
     """
 
-    def __init__(self, filenames):
+    def __init__(self, filenames, with_traces=False):
         """
         Create a new RawData object.
 
-        Each filename element corresponds to a measurement run. It must be a tar archive with the following contents:
+        Each filename element corresponds to a measurement run.
+        It must be a tar archive with the following contents:
 
         Version 0:
 
@@ -540,8 +542,12 @@ class RawData:
           `.opt.configs`: ....
         * EnergyTrace log files (`*.etlog`) as specified in `.opt.files`
 
+        If a cached result for a file is available, it is loaded and the file
+        is not preprocessed, unless `with_traces` is set.
+
         tbd
         """
+        self.with_traces = with_traces
         self.filenames = filenames.copy()
         self.traces_by_fileno = []
         self.setup_by_fileno = []
@@ -562,7 +568,8 @@ class RawData:
                     break
 
         self.set_cache_file()
-        self.load_cache()
+        if not with_traces:
+            self.load_cache()
 
     def set_cache_file(self):
         cache_key = hashlib.sha256('!'.join(self.filenames).encode()).hexdigest()
@@ -580,6 +587,8 @@ class RawData:
                 self.preprocessed = True
 
     def save_cache(self):
+        if self.with_traces:
+            return
         try:
             os.mkdir(self.cache_dir)
         except FileExistsError:
@@ -920,6 +929,7 @@ class RawData:
     def get_preprocessed_data(self, verbose=True):
         """
         Return a list of DFA traces annotated with energy, timing, and parameter data.
+        The list is cached on disk, unless the constructor was called with `with_traces` set.
 
         Each DFA trace contains the following elements:
          * `id`: Numeric ID, starting with 1
@@ -1000,6 +1010,7 @@ class RawData:
                                 'fileno': i,
                                 'info': member,
                                 'setup': self.setup_by_fileno[i],
+                                'with_traces': self.with_traces,
                             })
 
             elif version == 1:
@@ -1049,6 +1060,7 @@ class RawData:
                                 'setup': self.setup_by_fileno[j],
                                 'repeat_id': repeat_id,
                                 'expected_trace': ptalog['traces'][j],
+                                'with_traces': self.with_traces,
                             })
                 self.filenames = new_filenames
 
@@ -2507,7 +2519,7 @@ class MIMOSA:
     Resulting data is a list of state/transition/state/transition/... measurements.
     """
 
-    def __init__(self, voltage: float, shunt: int, verbose=True):
+    def __init__(self, voltage: float, shunt: int, verbose=True, with_traces=False):
         """
         Initialize MIMOSA loader for a specific voltage and shunt setting.
 
@@ -2518,6 +2530,7 @@ class MIMOSA:
         self.voltage = voltage
         self.shunt = shunt
         self.verbose = verbose
+        self.with_traces = with_traces
         self.r1 = 984  # "1k"
         self.r2 = 99013  # "100k"
         self.errors = list()
@@ -2855,6 +2868,9 @@ class MIMOSA:
                 'us': (idx - previdx) * 10,
             }
 
+            if self.with_traces:
+                data['uW'] = range_ua * self.voltage
+
             if 'states' in substates:
                 data['substates'] = substates
                 ssum = np.sum(list(map(lambda x: x['duration'], substates['states'])))