Polyhedron Fortran 2005 benchmarking (AMD64)

The following results were obtained with the Polyhedron 2005 Benchmark Suite on an AMD Athlon64 4800+ (dual core, 1MB L2 cache each, SSE, SSE2, SSE3), 4GB DDR400 CL2.5, SATA-II running openSUSE Factory in x86_64 mode.

The results have been obtained by averaging over three results (parameters: 200 0.1 3 3; since 2007-06-11: 300 0.5 5 10).

Other results

Results for AMD64 on the website of Polyhedron
Daily runs of gfortran (trunk) by openSUSE; further benchmark runs at openSUSE
CP2K GCC/gfortran benchmark and other GCC benchmarks

Compilers

GNU Compiler Collection gfortran 4.4.0 (daily trunk build)
Compiler options [64]: gfortran -march=opteron -ffast-math -funroll-loops -ftree-loop-linear -ftree-vectorize -msse3 -O3
Compiler options [LTO] gfortran -march=native -ffast-math -funroll-loops -flto -fwhole-program -O3
Compiler options [ACML] gfortran -march=opteron -ffast-math -funroll-loops -ftree-vectorize -ftree-loop-linear -msse3 -mveclibabi=acml [ACML 4.0.0]
Compiler options [Prof]: gfortran -fprofile-use -march=opteron -ffast-math -funroll-loops -ftree-loop-linear -ftree-vectorize -O3 (before: a single run with -fprofile-generate)
Compiler options [32]: gfortran -m32 -march=opteron -ffast-math -funroll-loops -ftree-loop-linear -ftree-vectorize -O3
Compiler options [32SSE]: gfortran -m32 -march=opteron -ffast-math -funroll-loops -ftree-loop-linear -ftree-vectorize -mfpmath=sse -msse3 -O3
New option since 2007-04-29: -ftree-loop-linear
GNU Compiler Collection gfortran 4.2 branch
Compiler options [64bit]: gfortran -march=opteron -ffast-math -funroll-loops -ftree-loop-linear -ftree-vectorize -msse3 -O3
GNU Compiler Collection gfortran 4.1.3 (openSUSE system compiler; 20070430)
Compiler options [64bit]: gfortran -march=opteron -ffast-math -funroll-loops -ftree-loop-linear -ftree-vectorize -msse3 -O3
Open64 4.0
Compiler options: openf95 -march=opteron -msse3 -O3
Intel Fortran Compiler ifort 9.1.40 (20061101)
Compiler options: ifort -O3 -xW -ipo -static
Intel Fortran Compiler ifort 10.0.013 Beta (20061208)
After February 20, 2007: ifort 10.0.014 Beta (20070209)
After March 19, 2007: ifort 10.0.017 Beta (20070307)
After June 6, 2007: ifort 10.0.023 (20070426)
Compiler options: ifort -O3 -xW -ipo -ip
After July 2, 2009: ifort 11.1.038 (20090511)
Compiler options: ifort -O3 -xHost -ipo -ip
Intel Fortran Compiler ifort 10.1.011 (20071116)
Compiler options: ifort -O3 -xW -ipo -ip
Intel Fortran Compiler ifort 11.0.069 (20080930)
Compiler options: ifort -O3 -xW -ipo -ip
(SUN Studio 11) Sunstudio sunf95 8.3 Build27_2 (2006/08/02)
Sunstudio sunf95 8.3 Build35_2 (2006/12/04) [Since 20 December 2006]
Compiler options: sunf95 -w4 -fast -xarch=amd64a -xipo=0
(SUN Studio 12) Sunstudio sunf95 8.3 Linux_i386 2007/05/03
NAG f95 5.1 (302)
Uses system C compiler as backend: gcc (GCC) 4.1.2 20061115 (prerelease) (SUSE Linux)
Compiler options: f95 -O4 -ieee=full -Bstatic -march=opteron -ffast-math -funroll-loops -ftree-vectorize -msse3
g95 4.0.3 (g95 0.91!) Jan 17 2007
Compiler options: g95 -march=opteron -ffast-math -funroll-loops -ftree-vectorize -msse3 -O3
PathScale EKOPath(TM) Compiler Suite Version 2.9.99 (beta)
Compiler options: pathf95 -Ofast -OPT:fast_math=on -WOPT:if_conv=0 -LNO:fu=9:full_unroll_size=7000 -static
2009:New version used: 3.2.99; same options.

As only the gfortran trunk binaries changes daily, daily variations of the other compilers should not happen!

Run time

As of Sun, 10 Jan 2010 15:15:34 +0100

Color code: Dark green: Within 4% of the fastest; light green: within 7% of the fastest; light red slower by more than 20% than the fastest; dark red: by more than 50%. (In order not to cheat, gfortran[Prof] results are not for comparison of the fastest.)

GCC bug reports: PR31897. The crash (internal compiler error) of g95 and of gfortran 4.1 is a backend issue triggered with -ftree-vectorize.

Note: Using -mrecip gas_dyn is sped up by 30%, but aermod does a direct comparison of two floating point numbers which fails due to slightly differened results.

	gfortran64	gfortranLTO	gfortranProf	gfortran32	gfortran32sse	gfortran4.2	gfortran4.1	Open64	Intel	ifort10	ifort10.1	ifort11	sunf95 (SS11)	sunf95 (SS12)	Pathsc3	NAG	g95
ac	13.30 [100]	13.04 [ 98]	14.24 [107]	14.67 [110]	18.29 [137]	22.15 [166]	22.98 [172]	19.15 [143]	15.09 [113]	15.23 [114]	15.35 [115]	15.08 [113]	20.25 [152]	19.94 [149]	10.15 [ 76]	24.38 [183]	20.13 [151]
aermod	35.34 [100]	31.66 [ 89]	32.53 [ 92]	44.15 [124]	38.39 [108]	34.37 [ 97]	50.60 [143]	46.81 [132]	30.32 [ 85]	29.75 [ 84]	29.14 [ 82]	24.55 [ 69]	31.13 [ 88]	30.83 [ 87]	37.41 [105]	43.97 [124]	45.85 [129]
air	15.99 [100]	15.92 [ 99]	16.66 [104]	17.38 [108]	17.69 [110]	17.30 [108]	-1.00 [ -6]	13.60 [ 85]	13.68 [ 85]	13.73 [ 85]	13.77 [ 86]	14.06 [ 87]	12.81 [ 80]	13.41 [ 83]	15.55 [ 97]	14.37 [ 89]	20.21 [126]
capacita	83.71 [100]	82.34 [ 98]	88.91 [106]	89.72 [107]	90.51 [108]	83.79 [100]	84.33 [100]	63.07 [ 75]	87.90 [105]	84.73 [101]	84.64 [101]	88.00 [105]	85.96 [102]	85.72 [102]	66.13 [ 78]	85.90 [102]	97.26 [116]
channel	16.29 [100]	15.35 [ 94]	17.99 [110]	17.16 [105]	16.77 [102]	12.92 [ 79]	12.67 [ 77]	17.20 [105]	12.56 [ 77]	14.84 [ 91]	14.83 [ 91]	15.47 [ 94]	11.97 [ 73]	12.24 [ 75]	20.56 [126]	15.91 [ 97]	25.19 [154]
doduc	43.43 [100]	41.99 [ 96]	40.93 [ 94]	43.10 [ 99]	43.46 [100]	40.94 [ 94]	41.51 [ 95]	53.88 [124]	40.51 [ 93]	41.23 [ 94]	40.91 [ 94]	39.13 [ 90]	44.64 [102]	45.27 [104]	38.34 [ 88]	49.79 [114]	50.19 [115]
fatigue	9.43 [100]	9.71 [102]	9.07 [ 96]	13.10 [138]	12.51 [132]	21.43 [227]	27.42 [290]	17.86 [189]	12.01 [127]	10.64 [112]	10.67 [113]	9.72 [103]	8.34 [ 88]	8.30 [ 88]	6.21 [ 65]	19.58 [207]	46.55 [493]
gas_dyn	11.79 [100]	11.19 [ 94]	12.13 [102]	11.79 [100]	12.45 [105]	15.25 [129]	16.74 [141]	11.23 [ 95]	8.58 [ 72]	7.58 [ 64]	7.40 [ 62]	6.78 [ 57]	12.06 [102]	13.34 [113]	9.24 [ 78]	14.29 [121]	-1.00 [ -8]
induct	36.72 [100]	35.09 [ 95]	37.95 [103]	34.40 [ 93]	34.13 [ 92]	52.31 [142]	46.64 [127]	38.83 [105]	43.33 [118]	42.94 [116]	43.96 [119]	43.44 [118]	42.53 [115]	43.35 [118]	30.18 [ 82]	38.86 [105]	42.65 [116]
linpk	32.30 [100]	30.31 [ 93]	31.66 [ 98]	33.40 [103]	32.90 [101]	26.34 [ 81]	25.93 [ 80]	26.09 [ 80]	25.98 [ 80]	25.80 [ 79]	25.90 [ 80]	25.88 [ 80]	24.68 [ 76]	26.99 [ 83]	30.64 [ 94]	25.87 [ 80]	27.50 [ 85]
mdbx	20.25 [100]	19.98 [ 98]	19.17 [ 94]	20.89 [103]	19.86 [ 98]	24.64 [121]	23.97 [118]	19.43 [ 95]	21.19 [104]	20.88 [103]	20.97 [103]	20.75 [102]	18.55 [ 91]	18.42 [ 90]	19.85 [ 98]	21.41 [105]	26.75 [132]
nf	31.97 [100]	31.38 [ 98]	32.05 [100]	36.94 [115]	35.10 [109]	33.47 [104]	38.19 [119]	24.48 [ 76]	26.27 [ 82]	25.73 [ 80]	25.83 [ 80]	25.63 [ 80]	28.13 [ 87]	31.03 [ 97]	28.83 [ 90]	28.04 [ 87]	48.72 [152]
protein	64.34 [100]	64.18 [ 99]	56.86 [ 88]	79.38 [123]	73.01 [113]	64.84 [100]	62.92 [ 97]	62.87 [ 97]	59.62 [ 92]	59.06 [ 91]	60.26 [ 93]	60.33 [ 93]	63.65 [ 98]	66.78 [103]	61.16 [ 95]	66.25 [102]	81.16 [126]
rnflow	35.83 [100]	36.04 [100]	35.85 [100]	51.12 [142]	37.81 [105]	44.64 [124]	45.35 [126]	39.33 [109]	43.93 [122]	38.95 [108]	39.15 [109]	39.07 [109]	37.93 [105]	38.02 [106]	31.18 [ 87]	41.94 [117]	52.93 [147]
test_fpu	21.27 [100]	22.49 [105]	21.11 [ 99]	22.68 [106]	22.80 [107]	19.64 [ 92]	19.53 [ 91]	18.47 [ 86]	17.96 [ 84]	18.05 [ 84]	18.08 [ 85]	17.83 [ 83]	16.15 [ 75]	16.11 [ 75]	19.19 [ 90]	19.70 [ 92]	32.08 [150]
tfft	8.50 [100]	8.20 [ 96]	8.42 [ 99]	8.62 [101]	8.83 [103]	7.74 [ 91]	7.62 [ 89]	7.68 [ 90]	7.69 [ 90]	7.50 [ 88]	7.64 [ 89]	7.77 [ 91]	7.63 [ 89]	7.45 [ 87]	8.12 [ 95]	7.87 [ 92]	-1.00 [-11]
Geometric Mean	24.40 [100]	23.82 [ 97]	24.31 [ 99]	27.03 [110]	26.39 [108]	27.30 [111]	36.76 [150]	25.06 [102]	23.09 [ 94]	22.66 [ 92]	22.72 [ 93]	22.24 [ 91]	22.99 [ 94]	22.99 [ 94]	21.89 [ 89]	26.92 [110]	59.16 [242]

Failures per compiler

gfortran64 (2010-01-10)

gfortranVecCost (2009-04-15)

gfortranProf (2010-01-10)

gfortran32 (2010-01-10)

gfortran32sse (2010-01-10)

gfortran42 (2007-05-29)

> Value= 25.114499300     Target= 23.100000000     Tolerance= 2.0000000000    
FAIL <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
--
 
linpk FAILED    1 fails and    4 passes

gfortran41 (2007-09-06)

Warning: Errors, checking aborted

Open64 (2007-06-26)

Intel (2007-03-08)

> Value= 2173.5026900     Target= 2191.1145000     Tolerance=0.10000000000E-02
FAIL <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
--
> Value= 7894.2202100     Target= 7924.8842800     Tolerance=0.30000000000E-01
FAIL <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
--
 
aermod FAILED    2 fails and    4 passes

ifort10 (2007-06-06)

ifort10.1 (2007-12-10)

ifort11.0 (2008-11-20)

sun95 (SS11)) (2007-03-08)

> Value=0.86723900000E-06 Target=0.86774900000E-06 Tolerance=0.10000000000E-10
FAIL <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
 
gas_dyn FAILED    1 fails and    0 passes
--
> Value= 25.187506000     Target= 23.100000000     Tolerance= 2.0000000000    
FAIL <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
--
 
linpk FAILED    1 fails and    4 passes

sun95 (SS12) (2007-06-04)

> Value= 25.187506000     Target= 23.100000000     Tolerance= 2.0000000000    
FAIL <<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<<
--
 
linpk FAILED    1 fails and    4 passes