Improve benchmark accuracy and output.

.github/workflows/ci.yml

@@ -78,7 +78,7 @@ jobs:
 
  - name: Running benchmark
  run: |
- python benchmark/telco_fractions.py
+ python benchmark/telco_fractions.py -n 250
  python benchmark/microbench.py create pydigits
 
  Linux:
@@ -201,7 +201,7 @@ jobs:
 
  - name: Running benchmark
  run: |
- python benchmark/telco_fractions.py
+ python benchmark/telco_fractions.py -n 250
  python benchmark/microbench.py create pydigits
 
  merge-wheels:

benchmark/telco_fractions.py

@@ -104,18 +104,20 @@ def run(cls):
 def main(n, cls=Fraction):
  for _ in range(5):
  run(cls) # warmup
- times = []
- for _ in range(n):
- times.append(run(cls))
+ times = [run(cls) for _ in range(n)]
  return times
 
 
+def percentile(values, percent):
+ return values[len(values) * percent // 100]
+
+
 if __name__ == "__main__":
  import optparse
  parser = optparse.OptionParser(
  usage="%prog [options]",
  description="Test the performance of the Telco fractions benchmark")
- parser.add_option("-n", "--num_runs", action="store", type="int", default=16,
+ parser.add_option("-n", "--num_runs", action="store", type="int", default=200,
  dest="num_runs", help="Number of times to repeat the benchmark.")
  parser.add_option("--use-decimal", action="store_true", default=False,
  dest="use_decimal", help="Run benchmark with Decimal instead of Fraction.")
@@ -130,10 +132,12 @@ def main(n, cls=Fraction):
  from fractions import Fraction as num_class
 
  results = main(options.num_runs, num_class)
- for result in results:
- print(result)
+ #for result in results:
+ # print(result)
+ #print()
 
- print()
  results.sort()
+ print('%.4f (15%%)' % percentile(results, 15))
+ print('%.4f (median)' % percentile(results, 50))
+ print('%.4f (85%%)' % percentile(results, 85))
  print('%.4f (mean)' % (fsum(results[1:-1]) / (len(results) - 2)))
- print('%.4f (median)' % (results[len(results) // 2]))