Performance analysis revealed that the standard RTEMS objects are a
major bottleneck. The object get mechanism and attribute checks at
runtime have a significant overhead. Use a custom implementation for
synchronization primitives. This drops also the size of the
synchronization primitives considerably.
