I was looking at the source code for the VxWorks port of dlmalloc
(http://www.xs4all.nl/~borkhuis/vxwor..._pt5.html#5.4A), and
noticed that by default it isn't task safe. Isn't the default
implementation of of malloc task safe?

If so, are most people using it with the semaphores enabled and how
does that affect performance? Doug's comments indicate that the
allocator really isn't geared for concurrent environments, so I'm
wondering if some of the major performance improvements people are
seeing isn't related to going from a task safe (default malloc) to a
non-task safe allocator (default dlmalloc port).