These are the results for the EuroBen Distributed Memory Benchmark V1.0
with the following testing circumstances:
------------------------------------------------------------------------
 Computer                  SGI Origin3000-R14K 500 MHz                     
 Memory size               512 GB                                          
 Compiler version          MIPSpro Fortran version 7.3.1.2m                
 Compiler options          -O3 -64                   
 Operating System version  IRIX 6.5.10                                     
 Working precision          64-bits precision 
------------------------------------------------------------------------
Remarks:
1) Program 'mod1h', the bandwidth measurement of all kinds of 
   communication patterns, is run on 4, 6, and 8 processors, as it
   should.
2) Program 'mod1i': the results for 1--32 processors for 3 implementations
   of the distributed dot product.
3) Program 'mod1j': very precise bandwidth measurement that enables to
   calculate the latency of point-to-point communication.
4) Program 'mod2a': dense matrix-vector multiply on 1--64 processors.
5) Program 'mod2b': dense matrix-matrix multiply on 1--64 processors. 
6) Program 'mod2f': 1-D FFT on 1, 2, 4, ..., 64 processors.
7) Program 'mod2g': 2-D Haar Wavelet Transform on 1, 2, 4, ..., 64
   processors.
8) Program 'mod2i': Parallel Quicksort of Doubles and Integers on
   1--20 processors (not fit for all numbers of processors).
                    
