aggregated performance updated push

bin/pbt_to_ctf.py

@@ -28,7 +28,7 @@ def read_pbt(pbt_files_list):
 
     print('There are ' + str(len(trace.events)) + ' events in this trace', end=' ')
     for e in range(len(trace.events)):
-        print('id===', trace.events.id[e], ' node_id=', trace.events.node_id[e],' stream_id=',trace.events.stream_id[e], 'key=' ,trace.events.key[e],' type=',trace.events.type[e],' b=',trace.events.begin[e],' e=',trace.events.end[e])
+        print('id===', trace.events.id[e], ' node_id=', trace.events.node_id[e], ' stream_id=', trace.events.stream_id[e], 'key=' ,trace.events.key[e], ' type=', trace.events.type[e], ' b=', trace.events.begin[e], ' e=', trace.events.end[e])
 
 import json
 import re
@@ -41,18 +41,18 @@ def bool(str):
 def pbt_to_ctf(pbt_files_list, ctf_filename, skip_parsec_events, skip_mpi_events):
     ctf_data = {"traceEvents": []}
     # Dictionary to store aggregated durations
-    aggregated_durations = {}    
+    aggregated_durations = {}
 
     ptt_filename = pbt2ptt.convert(pbt_files_list, multiprocess=False)
     trace = ptt.from_hdf(ptt_filename)
 
     for e in range(len(trace.events)):
-        # print('id=',trace.events.id[e],' node_id=',trace.events.node_id[e],' stream_id=',trace.events.stream_id[e],'key=',trace.events.key[e],' type=',trace.events.type[e],' b=',trace.events.begin[e],' e=',trace.events.end[e])
+        # print('id=', trace.events.id[e], ' node_id=', trace.events.node_id[e], ' stream_id=', trace.events.stream_id[e], 'key=', trace.events.key[e], ' type=', trace.events.type[e], ' b=', trace.events.begin[e], ' e=', trace.events.end[e])
         # print('\n')
 
-        if(skip_parsec_events == True and trace.event_names[trace.events.type[e]].startswith("PARSEC")):
+        if skip_parsec_events == True and trace.event_names[trace.events.type[e]].startswith("PARSEC"):
             continue
-        if(skip_mpi_events == True and trace.event_names[trace.events.type[e]].startswith("MPI")):
+        if skip_mpi_events == True and trace.event_names[trace.events.type[e]].startswith("MPI"):
             continue
 
         ctf_event = {}
@@ -61,9 +61,9 @@ def pbt_to_ctf(pbt_files_list, ctf_filename, skip_parsec_events, skip_mpi_events
         ctf_event["dur"] = 0.001 * (trace.events.end[e] - trace.events.begin[e]) # when we started, in ms
         ctf_event["name"] = trace.event_names[trace.events.type[e]]
 
-        if trace.events.key[e] != None:
+        if trace.events.key[e] is not None:
             ctf_event["args"] = trace.events.key[e].decode('utf-8').rstrip('\x00')
-            ctf_event["name"] = trace.event_names[trace.events.type[e]]+"<"+ctf_event["args"]+">"
+            ctf_event["name"] = trace.event_names[trace.events.type[e]] + "<" + ctf_event["args"] + ">"
 
         ctf_event["pid"] = trace.events.node_id[e]
         tid = trace.streams.th_id[trace.events.stream_id[e]]
@@ -78,16 +78,17 @@ def pbt_to_ctf(pbt_files_list, ctf_filename, skip_parsec_events, skip_mpi_events
         if index_of_open_bracket != -1:
             name = event_trace["name"][:index_of_open_bracket]
             duration = event_trace["dur"]
-            seconds = duration / 1000
-
+
             if name in aggregated_durations:
-                aggregated_durations[name] += duration
+                aggregated_durations[name]["duration"] += duration
+                aggregated_durations[name]["count"] += 1
             else:
-                aggregated_durations[name] = duration
+                # If name doesn't exist, create a new entry
+                aggregated_durations[name] = {"duration": duration, "count": 1}
 
-    # Print aggregated durations
-    for name, duration in aggregated_durations.items():
-        print(f"{name}: {duration} ms, equivalent to: {seconds} seconds")
+    # Calculate the averages for each name
+    averages = {name: data["duration"] / data["count"] for name, data in aggregated_durations.items()}
+    print(averages)
 
     with open(ctf_filename, "w") as chrome_trace:
         json.dump(ctf_data, chrome_trace)
@@ -105,10 +106,10 @@ def pbt_to_ctf(pbt_files_list, ctf_filename, skip_parsec_events, skip_mpi_events
         skip_mpi_events = bool(sys.argv[4])
 
     # iterate over all files within the directory that start with sys.argv[1]
-    pbt_files_list=[]
+    pbt_files_list = []
     dirname = os.path.dirname(pbt_file_prefix)
     for file in os.listdir(dirname):
-        file_fullname = os.path.join(dirname,file)
+        file_fullname = os.path.join(dirname, file)
         if file_fullname.startswith(pbt_file_prefix) and ".prof" in file_fullname and file_fullname != ctf_file_name:
             print("found file ", file_fullname)
             pbt_files_list.append(file_fullname)