[tools] Add capability of generating log2-based histograms to eval_gc_nvp.py

tools/eval_gc_nvp.py

Index: tools/eval_gc_nvp.py
diff --git a/tools/eval_gc_nvp.py b/tools/eval_gc_nvp.py
index 047a6dfae751108c874b2eab1410c3fcdb9d6da0..8a9b8e70728cc58c53970cc4f8184d18a486a543 100755
--- a/tools/eval_gc_nvp.py
+++ b/tools/eval_gc_nvp.py
@@ -10,17 +10,47 @@
 from argparse import ArgumentParser
 from copy import deepcopy
 from gc_nvp_common import split_nvp
+from math import log
 from sys import stdin
 
 
-class Histogram:
-  def __init__(self, granularity, fill_empty):
+class LinearBucket:
+  def __init__(self, granularity):
     self.granularity = granularity
+
+  def value_to_bucket(self, value):
+    return int(value / self.granularity)
+
+  def bucket_to_range(self, bucket):
+    return (bucket * self.granularity, (bucket + 1) * self.granularity)
+
+
+class Log2Bucket:
+  def __init__(self, start):
+    self.start = int(log(start, 2)) - 1
+
+  def value_to_bucket(self, value):
+    index = int(log(value, 2))
+    index -= self.start
+    if index < 0:
+      index = 0
+    return index
+
+  def bucket_to_range(self, bucket):
+    if bucket == 0:
+      return (0, 2 ** (self.start + 1))
+    bucket += self.start
+    return (2 ** bucket, 2 ** (bucket + 1))
+
+
+class Histogram:
+  def __init__(self, bucket_trait, fill_empty):
     self.histogram = {}
     self.fill_empty = fill_empty
+    self.bucket_trait = bucket_trait
 
   def add(self, key):
-    index = int(key / self.granularity)
+    index = self.bucket_trait.value_to_bucket(key)
     if index not in self.histogram:
       self.histogram[index] = 0
     self.histogram[index] += 1
@@ -29,20 +59,17 @@ class Histogram:
     ret = []
     keys = self.histogram.keys()
     keys.sort()
-    last = -self.granularity
-    for key in keys:
-      min_value = key * self.granularity
-      max_value = min_value + self.granularity
-
-      if self.fill_empty:
-        while (last + self.granularity) != min_value:
-          last += self.granularity
+    last = keys[len(keys) - 1]
+    for i in range(0, last + 1):
+      (min_value, max_value) = self.bucket_trait.bucket_to_range(i)
+      if i == keys[0]:
+        keys.pop(0)
+        ret.append("  [{0},{1}[: {2}".format(
+          str(min_value), str(max_value), self.histogram[i]))
+      else:
+        if self.fill_empty:
           ret.append("  [{0},{1}[: {2}".format(
-            str(last), str(last + self.granularity), 0))
-
-      ret.append("  [{0},{1}[: {2}".format(
-        str(min_value), str(max_value), self.histogram[key]))
-      last = min_value
+            str(min_value), str(max_value), 0))
     return "\n".join(ret)
 
 
@@ -73,27 +100,37 @@ class Category:
 def main():
   parser = ArgumentParser(description="Process GCTracer's NVP output")
   parser.add_argument('keys', metavar='KEY', type=str, nargs='+',
-                      help='the keys (names) to process')
-  parser.add_argument('--histogram-granularity', metavar='GRANULARITY',
-                      type=int, nargs='?', default=5,
-                      help='histogram granularity (default: 5)')
-  parser.add_argument('--no-histogram-print-empty', dest='histogram_print_empty',
-                      action='store_false',
-                      help='print empty histogram buckets')
-  feature_parser = parser.add_mutually_exclusive_group(required=False)
-  feature_parser.add_argument('--histogram', dest='histogram',
-                              action='store_true',
-                              help='print histogram')
-  feature_parser.add_argument('--no-histogram', dest='histogram',
-                              action='store_false',
-                              help='do not print histogram')
+                      help='the keys of NVPs to process')
+  parser.add_argument('--histogram-type', metavar='<linear|log2>',
+                      type=str, nargs='?', default="linear",
+                      help='histogram type to use (default: linear)')
+  linear_group = parser.add_argument_group('linear histogram specific')
+  linear_group.add_argument('--linear-histogram-granularity',
+                            metavar='GRANULARITY', type=int, nargs='?',
+                            default=5,
+                            help='histogram granularity (default: 5)')
+  log2_group = parser.add_argument_group('log2 histogram specific')
+  log2_group.add_argument('--log2-histogram-init-bucket', metavar='START',
+                          type=int, nargs='?', default=64,
+                          help='initial buck size (default: 64)')
+  parser.add_argument('--histogram-omit-empty-buckets',
+                      dest='histogram_omit_empty',
+                      action='store_true',
+                      help='omit empty histogram buckets')
+  parser.add_argument('--no-histogram', dest='histogram',
+                      action='store_false', help='do not print histogram')
   parser.set_defaults(histogram=True)
-  parser.set_defaults(histogram_print_empty=True)
+  parser.set_defaults(histogram_omit_empty=False)
   args = parser.parse_args()
 
   histogram = None
   if args.histogram:
-    histogram = Histogram(args.histogram_granularity, args.histogram_print_empty)
+    bucket_trait = None
+    if args.histogram_type == "log2":
+      bucket_trait = Log2Bucket(args.log2_histogram_init_bucket)
+    else:
+      bucket_trait = LinearBucket(args.linear_histogram_granularity)
+    histogram = Histogram(bucket_trait, not args.histogram_omit_empty)
 
   categories = [ Category(key, deepcopy(histogram))
                  for key in args.keys ]