I finally found the time to test the free C ix86 INTEL compiler, icc, and my results corroborate those of Simon with the FORTRAN version. On data processing code, I get with icc an average speedup of 15% in execution speed compared to gcc compiled executables, with all optimization features on. Thanks to MMX and/or SSE vectorization, on Pentium3/4 and AthlonXP the speedup can reach up to 30% in some applications like SWarp.