Author: David Schneider <david.schnei...@picle.org>
Branch: extradoc
Changeset: r4475:c89413a62517
Date: 2012-08-08 17:38 +0200
http://bitbucket.org/pypy/extradoc/changeset/c89413a62517/
Log: (cfbolz, bivab) tool for guard failure calculations
diff --git a/talk/vmil2012/tool/guard_count.py
b/talk/vmil2012/tool/guard_count.py
new file mode 100644
--- /dev/null
+++ b/talk/vmil2012/tool/guard_count.py
@@ -0,0 +1,44 @@
+import json
+import csv
+MISSING=1e-30
+with file("logs/guard_summary.json") as f:
+ data = json.load(f)
+with file("logs/resume_summary.csv") as f:
+ reader = csv.DictReader(f, delimiter=',')
+ csv_data = dict([(l['bench'], l) for l in reader])
+
+
+rows = []
+max_guardcount = 0
+for bench, d in data.iteritems():
+ failures = sorted(d['results'].values())
+ total_failures = float(sum(failures))
+ normed_failures = [f/total_failures for f in failures]
+ guardcount = int(csv_data[bench]['number of guards'])
+ normed_failures.reverse()
+ normed_failures += [MISSING] * (guardcount - len(failures))
+ # marker to see where it ends
+ normed_failures += [1]
+ rows.append(normed_failures)
+ max_guardcount = max(guardcount, max_guardcount)
+
+nbenchs = len(rows)
+
+
+
+with file("logs/guard_failures.csv", "w") as out:
+ writer = csv.writer(out)
+ res = []
+ for k in data.keys():
+ res += [k,k]
+ writer.writerow(res)
+
+ for i in range(max_guardcount):
+ res = []
+ for row in rows:
+ if i < len(row):
+ res.append(i/float(len(row)))
+ res.append(row[i])
+ else:
+ row += [MISSING,MISSING]
+ writer.writerow(res)
_______________________________________________
pypy-commit mailing list
pypy-commit@python.org
http://mail.python.org/mailman/listinfo/pypy-commit
