--- Begin Message ---
Hi all,
François-Xavier Coudert wrote:
> I wanted to report some results for the Polyhedron benchmark** on
> Opteron [...]
>
> gfortran -march=k8 -ffast-math -funroll-loops -static -O3
> ifort -O3 -xW -ipo -static -V
A comparison gives the following
(When another compiler according to
www (dot) polyhedron (dot) com/pb05/linux/f90bench_AMD.html is much faster
then the Intel, I also marked it; polyhedron used:
gfortran -march=opteron -ffast-math -funroll-loops -O3
ifort -O3 -xW -ipo -no-prec-div -static -V
[I used (dot) to trick the SPAM filter.])
Benchmark gfortran ifort slow/fast-1
--------- -------- ----- -----------
ac 18.10 20.06 11% FASTER (Absoft: 30% faster than Intel;
10.44 vs. 13.60)
aermod 58.61 40.62 44% slower
air 19.30 14.89 30% slower
capacita 108.15 108.88 1% FASTER (Pathscale: 46% faster than
Intel; 71.02 vs. 104.03)
channel 12.76 13.18 3% FASTER
doduc 51.20 54.28 6% FASTER
fatigue 18.89 15.44 22% slower (Pathscale: 67% faster than
Intel; 6.38 vs. 10.64)
gas_dyn 22.17 10.48 212% slower (Absoft: 23% faster than Intel;
6.35 vs. 7.84)
induct 53.02 58.06 10% FASTER (Pathscale: 45% faster than
Intel; 30.56 vs. 44.24)
linpack 25.97 26.10 1% FASTER (Pathscale: 11% faster than
Intel; 25.76 vs. 28.59)
mdbx 31.31 27.97 20% slower
nf 37.39 28.93 29% slower (Pathscale: 11% fasther than
Intel; 25.76 vs. 28.59)
protein 82.50 75.46 9% slower
rnflow 58.41 58.10 1% slower (Pathscale: 49% faster than
Intel; 29.80 vs. 44.40)
test_fpu 24.72 22.35 11% slower
tfft 9.24 9.25 0% FASTER
Tobias
--- End Message ---