CUDA implementation of NPB-MZ 3.3 (NPB3.3-MZ-CUDA)
--------------------------------------------------

NPB3.3-MZ-CUDA is a sample CUDA implementation based on the 
sequential implementation of the NAS Parallel Benchmarks
Multi-Zone version 3.3.1. This implementation (NPB3.3-MZ-CUDA) 
contains three benchmarks: LU-MZ, SP-MZ, and BT-MZ.
Additionally, the standard NPB (non multi-zone) benchmarks
are supported.

1. Compilation and Execution
  
   Prerequisites: C++ compiler and a working CUDA installation
   are required.

   To compile the benchmarks, change into the appropriate
   subdirectory, adjust the settings in the Makefile to your
   system and type make. This produces a single executable
   that includes all benchmark classes. The benchmark class
   is specified by a command line parameter (see below for
   an example). If no parameter is provided, the default
   class 'S' is used.

   Example for LU-MZ benchmark:
   cd LU-MZ
   make
   ./lu-mz A    # execute LU-MZ benchmark class A


