diff -r 43f8a0f5edd3 perf.py
--- a/perf.py	Fri Nov 23 23:40:58 2012 +0100
+++ b/perf.py	Sun Mar 10 10:54:20 2013 -0400
@@ -2109,7 +2109,7 @@
                         "regex", "richards", "spectral_norm", "startup_nosite",
                         "telco", "threading", "unpack_sequence"],
                 # After 2to3-conversion
-                "py3k": ["2to3", "2n3", "chameleon", "mako_v2"]
+                "py3k": ["2to3", "2n3", "chameleon", "mako_v2"],
                }
 
 SLOW_BENCHMARKS = ["hexiom2"]
@@ -2138,6 +2138,7 @@
 
     Args:
         benchmarks_opt: the string passed to the -b option on the command line.
+        bench_groups: the collection of benchmark groups to pull from
 
     Returns:
         A set() of the names of the benchmarks to run.
@@ -2215,19 +2216,30 @@
     parser.values.output_style = value
 
 
-def main(argv, bench_funcs=BENCH_FUNCS, bench_groups=BENCH_GROUPS):
+def CreateBenchGroups(bench_funcs=BENCH_FUNCS, bench_groups=BENCH_GROUPS):
     bench_groups = bench_groups.copy()
     all_benchmarks = bench_funcs.keys()
     bench_groups["all"] = all_benchmarks
+    return bench_groups
+
+
+def main(argv, bench_funcs=BENCH_FUNCS, bench_groups=BENCH_GROUPS):
+    bench_groups = CreateBenchGroups(bench_funcs, bench_groups)
+
+    # Calculate the lengths of expanded benchmark names for all groups
+    bench_counts = {}
+    for name in bench_groups:
+        bench_counts[name] = sum(1 for _ in
+                                 _ExpandBenchmarkName(name, bench_groups))
 
     # Prettify the displayed benchmark list: first the benchmark groups by
     # decreasing number of benches, then individual benchmarks by
     # lexicographic order.
     pretty_benchmarks = ["%s(%d)" % (name, nbenchs)
         for nbenchs, name in sorted(
-            ((len(v), k) for (k, v) in bench_groups.items()),
+            ((v, k) for (k, v) in bench_counts.items()),
             reverse=True)]
-    pretty_benchmarks.extend(sorted(all_benchmarks))
+    pretty_benchmarks.extend(sorted(bench_groups["all"]))
 
     parser = optparse.OptionParser(
         usage="%prog [options] baseline_python changed_python",
diff -r 43f8a0f5edd3 test_perf.py
--- a/test_perf.py	Fri Nov 23 23:40:58 2012 +0100
+++ b/test_perf.py	Sun Mar 10 10:54:20 2013 -0400
@@ -112,36 +112,49 @@
 
     def testParseBenchmarksOption(self):
         # perf.py, no -b option.
-        should_run = perf.ParseBenchmarksOption("")
-        self.assertEqual(should_run, set(["2to3", "django", "slowpickle",
-                                          "slowspitfire", "slowunpickle"]))
+        bench_groups = perf.CreateBenchGroups()
+        should_run = perf.ParseBenchmarksOption("", bench_groups)
+        self.assertEqual(should_run, set(["2to3", "django", "nbody",
+                                          "slowpickle", "slowspitfire",
+                                          "slowunpickle", "spambayes"]))
 
         # perf.py -b 2to3
-        should_run = perf.ParseBenchmarksOption("2to3")
+        should_run = perf.ParseBenchmarksOption("2to3", bench_groups)
         self.assertEqual(should_run, set(["2to3"]))
 
         # perf.py -b 2to3,pybench
-        should_run = perf.ParseBenchmarksOption("2to3,pybench")
+        should_run = perf.ParseBenchmarksOption("2to3,pybench", bench_groups)
         self.assertEqual(should_run, set(["2to3", "pybench"]))
 
         # perf.py -b -2to3
-        should_run = perf.ParseBenchmarksOption("-2to3")
-        self.assertEqual(should_run, set(["django", "slowspitfire",
-                                          "slowpickle", "slowunpickle"]))
+        should_run = perf.ParseBenchmarksOption("-2to3", bench_groups)
+        self.assertEqual(should_run, set(["django", "nbody", "slowspitfire",
+                                          "slowpickle", "slowunpickle",
+                                          "spambayes"]))
 
         # perf.py -b all
-        should_run = perf.ParseBenchmarksOption("all")
+        should_run = perf.ParseBenchmarksOption("all", bench_groups)
         self.assertTrue("django" in should_run, should_run)
         self.assertTrue("pybench" in should_run, should_run)
 
         # perf.py -b -2to3,all
-        should_run = perf.ParseBenchmarksOption("-2to3,all")
+        should_run = perf.ParseBenchmarksOption("-2to3,all", bench_groups)
         self.assertTrue("django" in should_run, should_run)
         self.assertTrue("pybench" in should_run, should_run)
         self.assertFalse("2to3" in should_run, should_run)
 
         # Error conditions
-        self.assertRaises(ValueError, perf.ParseBenchmarksOption, "-all")
+        self.assertRaises(ValueError, perf.ParseBenchmarksOption, "-all",
+                          bench_groups)
+
+
+    def testBenchmarkCounts(self):
+        bench_groups = {"top": ["middle1", "middle2"],
+                        "middle1": ["bottom1", "bottom2"],
+                        "middle2": ["bottom3"]}
+        found = list(perf._ExpandBenchmarkName("top", bench_groups))
+        self.assertEqual(["bottom1", "bottom2", "bottom3"], found)
+
 
 if __name__ == "__main__":
     unittest.main()