[pypy-commit] pypy all_ordered_dicts: Trying to overallocate dicts a little bit more eagerly

[arigo]

Author: Armin Rigo <arigo at tunes.org>
Branch: all_ordered_dicts
Changeset: r75413:4a6947ecc8fe
Date: 2015-01-17 18:03 +0100
http://bitbucket.org/pypy/pypy/changeset/4a6947ecc8fe/

Log:	Trying to overallocate dicts a little bit more eagerly

diff --git a/rpython/rtyper/lltypesystem/rordereddict.py b/rpython/rtyper/lltypesystem/rordereddict.py
--- a/rpython/rtyper/lltypesystem/rordereddict.py
+++ b/rpython/rtyper/lltypesystem/rordereddict.py
@@ -617,18 +617,13 @@
 
 def _overallocate_entries_len(baselen):
     # This over-allocates proportional to the list size, making room
-    # for additional growth.  The over-allocation is mild, but is
-    # enough to give linear-time amortized behavior over a long
-    # sequence of appends() in the presence of a poorly-performing
-    # system malloc().
-    # The growth pattern is:  0, 4, 8, 16, 25, 35, 46, 58, 72, 88, ...
-    newsize = baselen + 1
-    if newsize < 9:
-        some = 3
-    else:
-        some = 6
-    some += newsize >> 3
-    return newsize + some
+    # for additional growth.  This over-allocates slightly more eagerly
+    # than with regular lists.  The idea is that there are many more
+    # lists than dicts around in PyPy, and dicts of 5 to 8 items are
+    # not that rare (so a single jump from 0 to 8 is a good idea).
+    # The growth pattern is:  0, 8, 17, 27, 38, 50, 64, 80, 98, ...
+    newsize = baselen + (baselen >> 3)
+    return newsize + 8
 
 @jit.look_inside_iff(lambda d: jit.isvirtual(d))
 def ll_dict_grow(d):
@@ -947,7 +942,9 @@
 #
 #  Irregular operations.
 
-DICT_INITSIZE = 8
+# Start the hashtable size at 16 rather than 8, as with rdict.py, because
+# it is only an array of bytes
+DICT_INITSIZE = 16
 
 
 @specialize.memo()
diff --git a/rpython/rtyper/test/test_rordereddict.py b/rpython/rtyper/test/test_rordereddict.py
--- a/rpython/rtyper/test/test_rordereddict.py
+++ b/rpython/rtyper/test/test_rordereddict.py
@@ -115,11 +115,18 @@
         rordereddict.ll_dict_setitem(ll_d, llstr("b"), 2)
         rordereddict.ll_dict_setitem(ll_d, llstr("c"), 3)
         rordereddict.ll_dict_setitem(ll_d, llstr("d"), 4)
-        assert len(get_indexes(ll_d)) == 8
         rordereddict.ll_dict_setitem(ll_d, llstr("e"), 5)
         rordereddict.ll_dict_setitem(ll_d, llstr("f"), 6)
-        assert len(get_indexes(ll_d)) == 32
-        for item in ['a', 'b', 'c', 'd', 'e', 'f']:
+        rordereddict.ll_dict_setitem(ll_d, llstr("g"), 7)
+        rordereddict.ll_dict_setitem(ll_d, llstr("h"), 8)
+        rordereddict.ll_dict_setitem(ll_d, llstr("i"), 9)
+        rordereddict.ll_dict_setitem(ll_d, llstr("j"), 10)
+        assert len(get_indexes(ll_d)) == 16
+        rordereddict.ll_dict_setitem(ll_d, llstr("k"), 11)
+        rordereddict.ll_dict_setitem(ll_d, llstr("l"), 12)
+        rordereddict.ll_dict_setitem(ll_d, llstr("m"), 13)
+        assert len(get_indexes(ll_d)) == 64
+        for item in 'abcdefghijklm':
             assert rordereddict.ll_dict_getitem(ll_d, llstr(item)) == ord(item) - ord('a') + 1
 
     def test_dict_grow_cleanup(self):