Add geomean summary row to output from utils/compare.py.

This change adds a summary row with the geometric mean of the selected
metric to the output of utils/compare.py. The 'rhs' values are
normalized to the 'lhs' values and the geometric mean of the results
is computed.

This should be similar to the geomean row in the LNT HTML UI and allow
to compare 2 sets of runs.

Reviewers: anemet, MatzeB, cmatthews, serge-sans-paille

Reviewed By: anemet

Differential Revision: https://reviews.llvm.org/D57828

llvm-svn: 356545

utils/compare.py

@@ -6,6 +6,7 @@
 from __future__ import print_function
 
 import pandas as pd
+from scipy import stats
 import sys
 import os.path
 import re
@@ -108,28 +109,40 @@ def readmulti(filenames):
     d = pd.concat(datasets, axis=0, names=['run'], keys=datasetnames)
     return d
 
-def add_diff_column(d, absolute_diff=False):
-    values = d.unstack(level=0)
-
-    has_two_runs = d.index.get_level_values(0).nunique() == 2
+def get_values(values):
+    # Create data view without diff column.
+    if 'diff' in values.columns:
+        values = values[[c for c in values.columns if c != 'diff']]
+    has_two_runs = len(values.columns) == 2
     if has_two_runs:
-        values0 = values.iloc[:,0]
-        values1 = values.iloc[:,1]
+        return (values.iloc[:,0], values.iloc[:,1])
     else:
-        values0 = values.min(axis=1)
-        values1 = values.max(axis=1)
+        return (values.min(axis=1), values.max(axis=1))
 
+def add_diff_column(values, absolute_diff=False):
+    values0, values1 = get_values(values)
     # Quotient or absolute difference?
     if absolute_diff:
         values['diff'] = values1 - values0
     else:
         values['diff'] = values1 / values0
         values['diff'] -= 1.0
-    # unstack() gave us a complicated multiindex for the columns, simplify
-    # things by renaming to a simple index.
-    values.columns = [(c[1] if c[1] else c[0]) for c in values.columns.values]
     return values
 
+def add_geomean_row(data, dataout):
+    """
+    Normalize values1 over values0, compute geomean difference and add a
+    summary row to dataout.
+    """
+    values0, values1 = get_values(data)
+    relative = values1 / values0
+    gm_diff = stats.gmean(relative) - 1.0
+
+    gm_row = {c: '' for c in dataout.columns}
+    gm_row['diff'] = gm_diff
+    gm_row['Program'] = 'Geomean difference'
+    return dataout.append(gm_row, ignore_index=True)
+
 def filter_failed(data, key='Exec'):
     return data.loc[data[key] == "pass"]
 
@@ -209,6 +222,9 @@ def print_result(d, limit_output=True, shorten_names=True,
     # Turn index into a column so we can format it...
     dataout.insert(0, 'Program', dataout.index)
 
+    if show_diff_column:
+        dataout = add_geomean_row(d, dataout)
+
     formatters = dict()
     formatters['diff'] = format_diff
     if shorten_names:
@@ -220,7 +236,11 @@ def format_name(name, common_prefix, common_suffix):
             return "%-45s" % truncate(name, 10, 30)
 
         formatters['Program'] = lambda name: format_name(name, drop_prefix, drop_suffix)
-    float_format = lambda x: "%6.2f" % (x,)
+    def float_format(x):
+        if x == '':
+            return ''
+        return "%6.2f" % (x,)
+
     pd.set_option("display.max_colwidth", 0)
     out = dataout.to_string(index=False, justify='left',
                             float_format=float_format, formatters=formatters)
@@ -334,6 +354,12 @@ def format_name(name, common_prefix, common_suffix):
     print("Metric: %s" % (",".join(metrics),))
     if len(metrics) > 0:
         data = data[metrics]
+
+    data = data.unstack(level=0)
+    # unstack() gave us a complicated multiindex for the columns, simplify
+    # things by renaming to a simple index.
+    data.columns = [(c[1] if c[1] else c[0]) for c in data.columns.values]
+
     data = add_diff_column(data)
 
     sortkey = 'diff'