spectral-norm benchmark N=5,500

Each chart bar shows how many times slower, one ↓ spectral-norm program was, compared to the fastest program.

These are not the only programs that could be written. These are not the only compilers and interpreters. These are not the only programming languages.

Column × shows how many times more each program used compared to the benchmark program that used least.

     sortsortsort
  ×   Program Source Code CPU secs Elapsed secs Memory KB Code B ≈ CPU Load
1.0Fortran Intel #2 7.857.86780513  1% 1% 2% 100%
1.0C gcc #5 7.901.99644569  100% 100% 100% 100%
1.0Fortran Intel #3 7.912.001,296638  99% 99% 99% 99%
1.0Fortran Intel 7.932.001,316568  99% 99% 100% 99%
1.0C++ g++ #5 7.972.009241044  100% 100% 100% 100%
1.0C++ g++ #6 7.972.019241044  100% 100% 100% 100%
1.2Ada 2005 GNAT #4 9.402.401,8802762  99% 98% 97% 98%
1.4C gcc #4 10.672.683801139  100% 100% 100% 100%
1.7C++ g++ #8 13.533.406441278  100% 100% 100% 100%
2.0Go #3 15.703.951,292536  99% 99% 99% 99%
2.0Haskell GHC #4 15.704.041,420984  97% 97% 97% 98%
2.0Ada 2005 GNAT #3 15.723.981,6161702  99% 99% 99% 99%
2.0C gcc #3 15.743.95644463  100% 100% 100% 100%
2.0Pascal Free Pascal 15.7615.778423  1% 1% 0% 100%
2.0Go #2 15.764.111,296668  96% 96% 96% 96%
2.0Lisp SBCL #2 15.974.1919,208906  96% 95% 95% 97%
2.0Scala 16.0415.9718,536404  99% 1% 0% 2%
2.1Java  16.5016.4416,968514  0% 1% 100% 0%
2.1Java  #2 16.564.2916,460950  96% 96% 99% 96%
2.1Scala #2 16.614.3519,928720  94% 95% 99% 94%
2.3OCaml #3 18.245.273,232938  87% 97% 95% 85%
2.4OCaml #2 18.8518.861,660377  1% 0% 0% 100%
2.6C gcc 20.7320.74364383  0% 100% 1% 1%
2.7Go 20.8120.811,032411  0% 1% 98% 3%
2.8Ada 2005 GNAT 21.7621.761,212710  1% 0% 1% 100%
2.8Dart 22.0421.9415,244457  0% 1% 100% 1%
2.9Racket #3 22.946.0019,540627  95% 95% 98% 95%
3.0Lisp SBCL #3 23.456.0920,336883  97% 96% 97% 97%
3.0Lisp SBCL 23.7823.798,824625  1% 1% 0% 100%
3.6Racket #2 28.1728.1717,904532  0% 0% 9% 92%
3.6C++ g++ 28.4628.48872452  0% 100% 1% 1%
3.8C# Mono 29.4729.4819,476459  1% 1% 0% 100%
4.0Racket 31.1431.1316,892446  0% 1% 100% 0%
4.0F# Mono #2 31.489.6821,972852  78% 82% 83% 83%
4.0C# Mono #2 31.608.0919,0361063  98% 97% 98% 98%
4.2Rust #2 32.8932.904,676525  0% 0% 100% 1%
4.2Rust 32.8932.904,808492  0% 1% 100% 0%
5.4Clojure #8 42.1811.2448,724918  93% 93% 91% 99%
5.4Clojure #6 42.3111.4645,320808  97% 90% 91% 91%
5.4Clojure #7 42.7611.5646,568762  91% 95% 94% 90%
5.7Erlang HiPE #2 45.0111.5114,436747  99% 98% 98% 98%
7.2Erlang HiPE 56.6856.6914,912507  100% 0% 0% 0%
14Haskell GHC #2 106.4836.423,784403  63% 65% 65% 100%
27Perl #4 212.5754.356,780551  97% 98% 98% 99%
29Perl 229.81229.893,860333  96% 0% 0% 4%
38Ruby JRuby #4 297.72292.36605,212326  18% 38% 33% 15%
50Ruby JRuby 6 min6 min626,968292  23% 22% 23% 35%
52PHP #2 6 min6 min4,688397  0% 0% 1% 100%
55PHP #3 7 min108.8614,3721193  99% 99% 99% 99%
82Ruby #3 10 min166.2926,908828  97% 97% 97% 97%
83Ruby 10 min10 min7,920292  0% 0% 0% 100%
93Ruby #4 12 min12 min5,436326  99% 0% 0% 1%
98Ruby JRuby #2 12 min197.76611,560776  98% 98% 98% 98%
119Python 3 #6 15 min15 min5,012328  0% 9% 91% 1%
131Python 3 #5 17 min258.4230,260437  99% 99% 99% 99%
133Python 3 #8 17 min17 min4,336449  0% 0% 1% 100%
136Perl #2 17 min17 min3,076343  99% 0% 0% 1%
136Perl #3 17 min269.367,976846  100% 99% 99% 100%
Scala #3 Failed982
Scala #4 Failed1006
"wrong" (different) algorithm / less comparable programs
0.3C gcc #2 2.282.28236,808669
1.0C++ g++ #7 8.072.039241283
1.5Python 3 #2 12.1512.16479,764233
2.1C++ g++ #2 16.644.196401330
13Python 3 #3 101.30101.81957,348379

 spectral-norm benchmark : Eigenvalue using the power method

You can write your own program for this task and contribute to the benchmarks game by following these general instructions.

More specifically:

diff program output N = 100 with this output file to check your program is correct before contributing.

We are trying to show the performance of various programming language implementations - so we ask that contributed programs not only give the correct result, but also use the same algorithm to calculate that result.

Each program should calculate the spectral norm of an infinite matrix A, with entries a11=1, a12=1/2, a21=1/3, a13=1/4, a22=1/5, a31=1/6, etc

Each program must implement 4 separate functions / procedures / methods like the C# program.

For more information see challenge #3 in Eric W. Weisstein, "Hundred-Dollar, Hundred-Digit Challenge Problems" and "Spectral Norm".

From MathWorld--A Wolfram Web Resource.
http://mathworld.wolfram.com/Hundred-DollarHundred-DigitChallengeProblems.html
http://mathworld.wolfram.com/SpectralNorm.html

Thanks to Sebastien Loisel for this benchmark.

Revised BSD license

  Home   Conclusions   License   Play