[Mono-bugs] [Bug 487846] PPC code gen is inefficient in several areas

[bugzilla_noreply at novell.com]

http://bugzilla.novell.com/show_bug.cgi?id=487846

User mprobst at novell.com added comment
http://bugzilla.novell.com/show_bug.cgi?id=487846#c2





--- Comment #2 from Mark Probst <mprobst at novell.com>  2009-04-06 08:52:01 MDT ---
I think this is a great idea, but I don't see how it would solve the
thread-safety problem.  You still might have to patch the long sequence in the
stub.

Never mind about the patch infos - they're destroyed when the method has
finished compiling.  For patching you also wouldn't need any additional data,
because the code would provide all the information.  In the method you'd have

   ...
   bl  stub
 ret_addr:
   ...

and in the stub you'd have

   addis ...
   ori ...
   sldi ...
   oris ...
   ori ...
   mtctr ...
   bctr

When you're patching the branch, LR points to ret_addr, i.e. LR-4 is the
address of the bl.  The bl points you to the stub, so you can patch either one,
depending on whether the bl has enough reach.

-- 
Configure bugmail: http://bugzilla.novell.com/userprefs.cgi?tab=email
------- You are receiving this mail because: -------
You are the QA contact for the bug.