[pypy-svn] r77827 - pypy/trunk/pypy/jit/metainterp

[cfbolz at codespeak.net]

Author: cfbolz
Date: Tue Oct 12 14:15:37 2010
New Revision: 77827

Modified:
   pypy/trunk/pypy/jit/metainterp/optimize_nopspec.py
Log:
I think this is a reason why perfect specialization needs to still be run. bit
annoying


Modified: pypy/trunk/pypy/jit/metainterp/optimize_nopspec.py
==============================================================================
--- pypy/trunk/pypy/jit/metainterp/optimize_nopspec.py	(original)
+++ pypy/trunk/pypy/jit/metainterp/optimize_nopspec.py	Tue Oct 12 14:15:37 2010
@@ -14,6 +14,9 @@
 def _optimize_loop(metainterp_sd, old_loop_tokens, loop):
     cpu = metainterp_sd.cpu
     metainterp_sd.logger_noopt.log_loop(loop.inputargs, loop.operations)
+    # XXX the following lines are probably still needed, to discard invalid
+    # loops. bit silly to run a full perfect specialization and throw the
+    # result away.
     finder = PerfectSpecializationFinder(cpu)
     finder.find_nodes_loop(loop, False)
     if old_loop_tokens:
@@ -31,6 +34,7 @@
 def _optimize_bridge(metainterp_sd, old_loop_tokens, bridge):
     cpu = metainterp_sd.cpu    
     metainterp_sd.logger_noopt.log_loop(bridge.inputargs, bridge.operations)
+    # XXX same comment as above applies
     finder = BridgeSpecializationFinder(cpu)
     finder.find_nodes_bridge(bridge)
     if old_loop_tokens: