This are results obtained on a NEC TX7 (16 proc.s, 1.5 GHz processors) with the
NEC FTN 3.4 compiler (very similar to the Intel 7.1 compiler) for the
OpenMP EuroBen Benchmark V1.0.

Remarks:
General - The results mod[1|2]x.p01.log are 1-processor OpenMP results,
          NOT sequential single-CPU results. Generally, the OpenMP results
          are worse than the sequential ones because of OpenMP overhead.
mod2b   - Solution of a dense linear system. In addition to the provided Fortran
          version, it was also run with the linear algebra routines from NEC's
	  Math Keisan Library with large improvements in performance (see 
	  mod2b_lib[p01,...p16].log).
mod2d   - Solution of a dense eigenvalue problem. In addition to the provided
	  Fortran version, it was also run with the linear algebra routines from
	  NEC's Math Keisan Library.  There were significant improvements in
	  performance (see  mod2d_lib.log). For the largest problem size, however,
	  N = 2000, the timings were incorrect, probably because of timer
	  overflow.
mod2f   - 1-D complex-to-complex FFT did not work properly because of incorrect
          code that is generated by the FTN 3.4 compiler for the radix-8
	  transform subroutine.
	
In most cases the speedup with respect to the single-CPU obtained with OpenMP
prove to be very poor (See Log.tx7).				
