add filter-param support to analyze-archive

3 files changed, 44 insertions, 15 deletions

diff --git a/bin/analyze-archive.py b/bin/analyze-archive.py
index e04f3fe..a4af02a 100755
--- a/bin/analyze-archive.py
+++ b/bin/analyze-archive.py
@@ -61,6 +61,12 @@ Options:
     also defined for cases such as safe_inv(0) or safe_sqrt(-1). This allows
     a greater range of functions to be tried during fitting.
 
+--filter-param=<parameter name>=<parameter value>[,<parameter name>=<parameter value>...]
+    Only consider measurements where <parameter name> is <parameter value>
+    All other measurements (including those where it is None, that is, has
+    not been set yet) are discarded. Note that this may remove entire
+    function calls from the model.
+
 --hwmodel=<hwmodel.json>
     Load DFA hardware model from JSON
 
@@ -73,7 +79,7 @@ import json
 import plotter
 import re
 import sys
-from dfatool import PTAModel, RawData, pta_trace_to_aggregate
+from dfatool import PTAModel, RawData, pta_trace_to_aggregate, filter_aggregate_by_param
 from dfatool import soft_cast_int, is_numeric, gplearn_to_function
 from dfatool import CrossValidator
 
@@ -210,6 +216,7 @@ if __name__ == '__main__':
         optspec = (
             'plot-unparam= plot-param= show-models= show-quality= '
             'ignored-trace-indexes= discard-outliers= function-override= '
+            'filter-param= '
             'cross-validate= '
             'with-safe-functions hwmodel= export-energymodel='
         )
@@ -242,6 +249,11 @@ if __name__ == '__main__':
             xv_method, xv_count = opts['cross-validate'].split(':')
             xv_count = int(xv_count)
 
+        if 'filter-param' in opts:
+            opts['filter-param'] = list(map(lambda x: x.split('='), opts['filter-param'].split(',')))
+        else:
+            opts['filter-param'] = list()
+
         if 'with-safe-functions' in opts:
             safe_functions_enabled = True
 
@@ -257,6 +269,9 @@ if __name__ == '__main__':
 
     preprocessed_data = raw_data.get_preprocessed_data()
     by_name, parameters, arg_count = pta_trace_to_aggregate(preprocessed_data, ignored_trace_indexes)
+
+    filter_aggregate_by_param(by_name, parameters, opts['filter-param'])
+
     model = PTAModel(by_name, parameters, arg_count,
         traces = preprocessed_data,
         discard_outliers = discard_outliers,
diff --git a/bin/analyze-timing.py b/bin/analyze-timing.py
index 39a915f..eecb70e 100755
--- a/bin/analyze-timing.py
+++ b/bin/analyze-timing.py
@@ -77,7 +77,7 @@ import re
 import sys
 from dfatool import AnalyticModel, TimingData, pta_trace_to_aggregate
 from dfatool import soft_cast_int, is_numeric, gplearn_to_function
-from dfatool import CrossValidator
+from dfatool import CrossValidator, filter_aggregate_by_param
 import utils
 
 opts = {}
@@ -208,19 +208,7 @@ if __name__ == '__main__':
 
     utils.prune_dependent_parameters(by_name, parameters)
 
-    for param_name_and_value in opts['filter-param']:
-        param_index = parameters.index(param_name_and_value[0])
-        param_value = soft_cast_int(param_name_and_value[1])
-        names_to_remove = set()
-        for name in by_name.keys():
-            indices_to_keep = list(map(lambda x: x[param_index] == param_value, by_name[name]['param']))
-            by_name[name]['param'] = list(map(lambda iv: iv[1], filter(lambda iv: indices_to_keep[iv[0]], enumerate(by_name[name]['param']))))
-            for attribute in by_name[name]['attributes']:
-                by_name[name][attribute] = by_name[name][attribute][indices_to_keep]
-                if len(by_name[name][attribute]) == 0:
-                    names_to_remove.add(name)
-        for name in names_to_remove:
-            by_name.pop(name)
+    filter_aggregate_by_param(by_name, parameters, opts['filter-param'])
 
     model = AnalyticModel(by_name, parameters, arg_count, use_corrcoef = opts['corrcoef'], function_override = function_override)
 
diff --git a/lib/dfatool.py b/lib/dfatool.py
index e039b92..16e23ce 100755
--- a/lib/dfatool.py
+++ b/lib/dfatool.py
@@ -1520,6 +1520,32 @@ def _add_trace_data_to_aggregate(aggregate, key, element):
     for datakey, dataval in element['offline_aggregates'].items():
         aggregate[key][datakey].extend(dataval)
 
+def filter_aggregate_by_param(aggregate, parameters, parameter_filter):
+    """
+    Remove entries with certain parameter values from `aggregate`.
+
+    :param aggregate: aggregated measurement data, must be a dict conforming to
+        aggregate[state or transition name]['param'] = (first parameter value, second parameter value, ...)
+        and
+        aggregate[state or transition name]['attributes'] = [list of keys with measurement data, e.g. 'power' or 'duration']
+    :param parameters: list of parameters, used to map parameter index to parameter name. parameters=['foo', ...] means 'foo' is the first parameter
+    :param parameter_filter: [[name, value], [name, value], ...] list of parameter values to remove. Values refer to normalizad parameter data.
+    """
+    for param_name_and_value in parameter_filter:
+        param_index = parameters.index(param_name_and_value[0])
+        param_value = soft_cast_int(param_name_and_value[1])
+        names_to_remove = set()
+        for name in aggregate.keys():
+            indices_to_keep = list(map(lambda x: x[param_index] == param_value, aggregate[name]['param']))
+            aggregate[name]['param'] = list(map(lambda iv: iv[1], filter(lambda iv: indices_to_keep[iv[0]], enumerate(aggregate[name]['param']))))
+            for attribute in aggregate[name]['attributes']:
+                aggregate[name][attribute] = aggregate[name][attribute][indices_to_keep]
+                if len(aggregate[name][attribute]) == 0:
+                    names_to_remove.add(name)
+        for name in names_to_remove:
+            aggregate.pop(name)
+
+
 def pta_trace_to_aggregate(traces, ignore_trace_indexes = []):
     u"""
     Convert preprocessed DFA traces from peripherals/drivers to by_name aggregate for PTAModel.