This is a UNIX port of the latest version of the benchmark described
by Larson & Krishnan in "Memory Allocation for Long-Running Server
Applications", ISMM 1998.

To see how it scales, try the following parameters, where P = 1 and
then the number of processors on your system, for driver and
driver-hoard:

Multi-threaded test driver 
C++ version (new and delete)
runtime (sec): 30
chunk size (min,max): 10 500
threads (min, max):   P P
chunks/thread:  10000
no of rounds:   10
random seed:    1
