This are results obtained on a Bull NovaScale (16 proc.s, 1.3 GHz processors)
with the Intel ifort 8.0 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.	 
mod1d   - Program to assess (cache) memory bank conflicts. The ifort 8.0
          compiler is not able to generate a correct executable (we
	  have kept mod1d.p02.log to prove it). With the efc 7.1 compiler
	  the are no problems regarding this program.
mod2b   - The solution of a dense linear system. For p > 1, the correctness
          check	indicates errors for all problem sizes. This occurred with
	  the ifort 8.0 compiler. With the efc 7.1 compiler the execution
	  was correct (see mod2b-7.1.p[01,...,16].log).
mod2d   - The solution of a dense eigenvalue problem. For p > 1 a segmentation
          fault	occurs after a problem size of N = 700 and worse, from a problem
	  size N > 20 the results are incorrect. This occurred with
	  the ifort 8.0 compiler. With the efc 7.1 compiler the execution
	  was correct (see mod2d-7.1.p[01,...,16].log).
mod2i   - Quicksort program. For p = 1,...,16 a segmentation fault occurred for
          N > 500,000 with the ifort 8.0 compiler. The problems do not happen
	  with the efc 7.1 compiler (see mod2i-7.1.p[01,...,16].log).
	  	     
Generally speaking, the version of the ifort 8 compiler used in this test does
not seem to be very stable when using OpenMP (status as of half March 2004).
