Polyhedron Benchmarks (AMD+Linux)

Linux/64 on AMD Processor

Absoft
15.0.2
Absoft(AP)
15.0.2
gfortran
5.2.0
Intel
16.0
Intel(AP)
16.0
NAG
6.0
Oracle
12.4
PGI
15.7
open64
4.5.2.1
AC 7.50 7.46 8.54 6.86 6.87 20.86 19.68 8.72 7.32
AERMOD 16.65 17.10 23.35 16.71 17.23 22.89 15.40 17.25 24.19
AIR 4.54 3.12 5.24 4.29 2.17 6.22 4.69 5.68 4.83
CAPACITA 32.93 33.41 38.46 34.05 34.23 40.38 39.89 30.30 31.64
CHANNEL2 137.72 129.71 136.89 141.50 114.96 212.97 142.31 139.51 188.36
DODUC 26.38 26.33 25.64 21.62 21.64 28.08 23.24 25.36 25.01
FATIGUE2 80.51 84.41 83.77 76.57 73.42 120.49 128.18 114.84 100.77
GAS_DYN2 113.02 137.03 132.56 97.83 104.09 205.98 118.38 149.18 104.03
INDUCT2 96.77 55.30 102.67 109.76 45.01 193.00 188.17 197.23 166.37
LINPK 9.04 9.09 8.22 9.45 9.46 9.91 9.88 9.94 10.55
MDBX 12.98 12.52 12.72 10.52 10.44 12.74 12.06 12.98 12.32
MP_PROP_DESIGN 134.25 70.83 263.03 110.53 34.15 360.59 207.02 125.39 140.57
NF 12.04 12.02 10.87 11.25 11.31 13.96 12.08 12.13 11.73
PROTEIN 30.35 30.85 31.73 30.30 30.35 34.60 31.23 36.02 30.46
RNFLOW 17.30 19.86 20.68 18.07 18.27 22.16 26.59 22.61 25.44
TEST_FPU2 98.76 88.34 86.13 83.44 77.35 116.77 102.55 94.03 90.66
TFFT2 157.74 155.27 112.31 157.57 160.14 159.98 156.42 161.81 159.53
Geometric Mean 33.69 31.09 36.13 31.94 26.86 47.44 40.50 38.10 37.34

 

Compiler Switches
Absoft af90 -m64 -O4 -speed_math=10 -march=barcelona -xINTEGER
Absoft(AP) af90 -m64 -O5 -speed_math=10 -march=barcelona -xINTEGER
(These settings enable auto-parallelization, see below)
gfortran gfortran -ffast-math -funroll-loops  –param max-unroll-times=2 -Ofast -march=native
Intel ifort -fast -ipo
Intel(AP) ifort -fast -parallel -ipo (These settings enable auto-parallelization, see below)
NAG nagfor -O4 -s -v -V
PGI pgf90 -V -fast -Mipa=fast,inline -Msmartalloc -Mfprelaxed -Mstack_arrays
Oracle sunf95 -fast -xtarget=native -m64
Open64 openf95 -Ofast

 

Notes
All figures are execution times in seconds – measured on a machine with an AMD Phenom II X4 955 processor (3.2 GHz), with 8GBytes memory, running Linux Mint 17.1. Each figure is the average over at least 10 runs (many more for some). Measurement error is typically <1%. Green cells highlight figures within 10% of the fastest. Red cells indicate figures which are more than 150% of the fastest.  So far as possible, we have used the compiler switches which give the best overall results. We have not attempted to tune individual benchmarks, and, in particular cases, different switch settings may give better results. The settings used for the Intel and Absoft compilers enable autoparallelization. Autoparallelization settings are not used on any other compilers because we found that they produced no significant performance benefits on this benchmark set.  Thanks are due to Jos Bergervoet for permission to use his CAPACITA benchmark, to Quetzal Associates for permission to use their CHANNEL, FATIGUE, GAS_DYN, INDUCT, PROTEIN and RNFLOW benchmarks, to David Frank for his TEST_FPU benchmark, to Anthony Falzone for the use of MP_PROP_DESIGN, and to Ted Addison of McVehil-Monnett Associates for permission to use AERMOD, an air quality model used by the US Environmental Protection Agency.  All the benchmarks have been modified slightly to fit into our benchmarking harness.