Introduce epoch, collect statistics based on epoch, restyle plots (#46)

* Plot the start of each epoch as a vertical line. The color of the line (green/red) shows if there is regression for the epoch
* Only plot min/max data point for the latest epoch.
* For all the data points, render it as red/green/black to show if it has regresssion/improvement from the best point in the same epoch.
* No longer show hover info for moving average and standard deviation
* Always consider standard deviation when checking regression.
* Allow notes/epochs to include a start time, in addition to a start date
* Shrink the size of each plot to roughly 500px * 1200px

Author: qinsoon

configs/jikesrvm-plot.yml

@@ -8,12 +8,20 @@ plans:
     baseline: []
 notes:
   - date: "20230116"
+    time: "0000"
     note: "Move to Ubuntu 22.04"
-  - date: "20231015"
+  - date: "20231017"
+    time: "0000"
     note: "Speculative RAS Overflow mitigation on Zen1/Zen2"
-  - date: "20231102"
+  - date: "20231103"
+    time: "0800"
     note: "Move to running-ng"
   - date: "20240403"
+    time: "0000"
     note: "Move to Rust 1.77.0"
+  - date: "20240802"
+    time: "0000"
+    note: "Microcode update for Zenbleed"
   - date: "20240903"
+    time: "0000"
     note: "Move to Linux Kernel 6.8.0"

configs/openjdk-plot.yml

@@ -16,12 +16,20 @@ plans:
     baseline: ['jdk-g1', 'jdk-zgc']
 notes:
   - date: "20230116"
+    time: "0000"
     note: "Move to Ubuntu 22.04"
-  - date: "20231015"
+  - date: "20231017"
+    time: "0000"
     note: "Speculative RAS Overflow mitigation on Zen1/Zen2"
-  - date: "20231102"
+  - date: "20231103"
+    time: "0800"
     note: "Move to running-ng. Use -Xcomp. Use image build."
   - date: "20240403"
+    time: "0000"
     note: "Move to Rust 1.77.0"
+  - date: "20240802"
+    time: "0000"
+    note: "Microcode update for Zenbleed"
   - date: "20240903"
+    time: "0000"
     note: "Move to Linux Kernel 6.8.0"

scripts/history_report.py

@@ -73,7 +73,7 @@
     print("Plan: %s" % plan)
     print("Last run: %s" % last_run)
     print("Benchmarks: %s" % benchmarks)
-    print(logs)
+    # print(logs)
 
     # figure out the baseline and get the result for the baseline
     plan_config = parse.get_config_for_plan(config, plan)
@@ -83,8 +83,10 @@
 
     baseline = plot.calculate_baseline(baseline_results, baseline_builds, "execution_times")
     pp.pprint(baseline)
-    
+
+    build_info = prefix
+
     # plot
-    fig = plot.plot_history(runs, plan, benchmarks, from_date, to_date, "execution_times", baseline, config['notes'].copy())
+    fig = plot.plot_history(build_info, runs, plan, benchmarks, from_date, to_date, "execution_times", baseline, config['notes'].copy())
     path = os.path.join(output_dir, "%s_%s_history.html" % (prefix, plan))
     fig.write_html(path)

scripts/parse.py

@@ -94,11 +94,14 @@ def parse_run_date(run_id):
         return datetime(int(matcher['year']), int(matcher['month']), int(matcher['day']), int(matcher['hour']), int(matcher['minute']), int(matcher['second']))
 
 # Given a note date, return the date object
-def parse_note_date(note_date):
+def parse_note_date(note_date, note_time = None):
     from datetime import datetime
-    matcher = re.match("(?P<year>\d{4})(?P<month>\d{2})(?P<day>\d{2})", note_date)
-    if matcher:
-        return datetime(int(matcher['year']), int(matcher['month']), int(matcher['day']))
+    date_matcher = re.match(r"(?P<year>\d{4})(?P<month>\d{2})(?P<day>\d{2})", note_date)
+    if note_time is None:
+        note_time = "0000"
+    time_matcher = re.match(r"(?P<hour>\d{2})(?P<minute>\d{2})", note_time)
+    if date_matcher and time_matcher:
+        return datetime(int(date_matcher['year']), int(date_matcher['month']), int(date_matcher['day']), int(time_matcher['hour']), int(time_matcher['minute']))
 
 # Given a yaml file path, return the file
 def parse_yaml(path):