[pypy-svn] r54004 - pypy/dist/pypy/interpreter
arigo at codespeak.net
arigo at codespeak.net
Tue Apr 22 12:57:07 CEST 2008
Author: arigo
Date: Tue Apr 22 12:57:06 2008
New Revision: 54004
Modified:
pypy/dist/pypy/interpreter/pyframe.py
Log:
Reverting unless there is evidence that it's an improvement.
To me it looks like the result is worse:
* allocating an empty list is faster than pre-allocating items, so
for frames that don't contain loops we loose
* if there are loops, it's very uncommon to have more than 4 nested
levels (which is the default allocation size).
Modified: pypy/dist/pypy/interpreter/pyframe.py
==============================================================================
--- pypy/dist/pypy/interpreter/pyframe.py (original)
+++ pypy/dist/pypy/interpreter/pyframe.py Tue Apr 22 12:57:06 2008
@@ -10,7 +10,7 @@
from pypy.rlib.objectmodel import we_are_translated, instantiate
from pypy.rlib.jit import we_are_jitted, hint
from pypy.rlib import rstack # for resume points
-from pypy.rlib.objectmodel import newlist
+
# Define some opcodes used
g = globals()
@@ -54,7 +54,7 @@
eval.Frame.__init__(self, space, w_globals, code.co_nlocals)
self.valuestack_w = [None] * code.co_stacksize
self.valuestackdepth = 0
- self.blockstack = newlist(sizehint=20)
+ self.blockstack = []
if space.config.objspace.honor__builtins__:
self.builtin = space.builtin.pick_builtin(w_globals)
# regular functions always have CO_OPTIMIZED and CO_NEWLOCALS.
More information about the Pypy-commit
mailing list