[pypy-svn] r64653 - pypy/branch/pyjitpl5/pypy/jit/metainterp

fijal at codespeak.net fijal at codespeak.net
Fri Apr 24 20:46:52 CEST 2009


Author: fijal
Date: Fri Apr 24 20:46:52 2009
New Revision: 64653

Modified:
   pypy/branch/pyjitpl5/pypy/jit/metainterp/compile.py
Log:
don't store pseudo-loops in possible compiled merge points. I must admit I have
a bit no clue what I'm doing, except I'm pretty sure optimize_bridge should
not randomly return pseudo loop. Also don't know how to write a test.


Modified: pypy/branch/pyjitpl5/pypy/jit/metainterp/compile.py
==============================================================================
--- pypy/branch/pyjitpl5/pypy/jit/metainterp/compile.py	(original)
+++ pypy/branch/pyjitpl5/pypy/jit/metainterp/compile.py	Fri Apr 24 20:46:52 2009
@@ -244,7 +244,8 @@
         # XXX it's probably useless to do so when optimizing
         glob = metainterp_sd.globaldata
         old_loops = glob.compiled_merge_points.setdefault(greenkey, [])
-        old_loops.append(new_loop)
+        if new_loop not in map_loop2descr:
+            old_loops.append(new_loop)
 
 
 def compile_fresh_bridge(metainterp, old_loops, resumekey):



More information about the Pypy-commit mailing list