------- Comment #17 from igaztanaga at gmail dot com 2008-07-23 08:03 -------
Andrew, I think I do write some crappy code, but the slowdown is also present
in the non-optimized version.
FWIW, there are STL implementations that embed the color bit in rbtrees in the
parent pointer so the optimization is not that strange. Maybe my implementation
is not correct and that's why the size-optimized version ends in an infinite
loop.
And we have tons of two-way pointers in production code. Too much crappy code
to ignore, I'm afraid.
--
igaztanaga at gmail dot com changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |igaztanaga at gmail dot com
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36861
