Revision 13187 switched us to using nanoTime instead of a cycle counter as the basis for all VM internal timing.
The current implementation of sysNanoTime in sys.C is the naive one that simply uses getttimeofday. On modern Linux versions, we can do better by using clock_gettime. There are also better mechanisms available on AIX. We should re-implement sysNanoTime to exploit the best available monotonic clock provided by the host OS.