[Issue 2900] Array appending slowed drastically since integration of druntime

[d-bugmail at puremagic.com]

http://d.puremagic.com/issues/show_bug.cgi?id=2900


sean at invisibleduck.org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |ASSIGNED




------- Comment #3 from sean at invisibleduck.org  2009-04-29 11:42 -------
Thanks for doing this research.  The reason gc_query() is currently used is
because it provides a means of determining explicitly whether the array is an
interior slice.  However, I've begun thinking that it would make more sense
just to require that gc_sizeOf() always return zero for interior pointers (I'd
left the door open for it to return a positive value).  That would bring
performance in line with Phobos pre-Druntime.  Either way though, the caching
of BlkInfo is a valuable change.  I'll see about at least getting these patches
applied before the next release.


--