I found a bit more time in the day and worked out how to run the C# compiler from the command line. I used the following command:
csc /optimize /out:commandline.exe *.cs Then when I run commandline I get: SciGMark 1.0 - C# - specialized FFT (1024): 388.779673030229 SOR (100x100): 630.880482605193 Monte Carlo : 76.1505509057327 Sparse matmult (N=1000, nz=5000): 576.581135920473 LU (100x100): 508.721326139402 PolyMult (N=40): 327.208768923184 Composite Score: 418.053656254036 Platform Information CLR Version: 2.0.50727.1433 Working Set: 5840896 Which is a considerable improvement over before, but still 1.5 times slower than Java. However on a personal note; I don't usually bother optimising further when I get within a factor of 2, so I would say that C# is fast enough. On Apr 23, 1:07 pm, hlovatt <[EMAIL PROTECTED]> wrote: > On my Mac Book Pro running Windows XP via Parallels the figures I get > are: > > For C# > > SciGMark 1.0 - C# - specialized > > FFT (1024): 127.693325911912 > SOR (100x100): 451.688663887376 > Monte Carlo : 53.8252351904638 > Sparse matmult (N=1000, nz=5000): 287.33058461682 > LU (100x100): 281.689806463528 > PolyMult (N=40): 129.801981762775 > Composite Score: 222.004932972147 > > Platform Information > CLR Version: 2.0.50727.1433 > Working Set: 17170432 > > For java -server: > > SciGMark 1.0 - Java - specialized > FFT (1024): 296.13593880108886 > SOR (100x100): 895.0093438244637 > Monte Carlo : 237.23858121300037 > Sparse matmult (N=10, nz=50): 467.9837360457056 > LU (100x100): 1304.666388568605 > PolyMult (N=40): 563.4980943142291 > > Composite Score: 627.4220137945155 > > java.vendor: Sun Microsystems Inc. > java.version: 1.6.0_06 > os.arch: x86 > os.name: Windows XP > os.version: 5.1 > > Which makes the Java version about 3 times quicker. I used the code > given for the paper I previously referenced, since this code uses > exactly the same algorithms and avoids system calls. It is quite > possible that I did not give the C# compiler the right options; I > simply ran the code from Visual Studio Express, I am far from a C# > expert. > > PS We are probably both is breach of the .NET user agreement; see > section 8, it points considerable restrictions on running benchmarks. > > On Apr 22, 7:44 pm, Jon Harrop <[EMAIL PROTECTED]> wrote: > > > On Tuesday 22 April 2008 08:39:03 hlovatt wrote: > > > > @John, > > > > Your benchmarking does not seem consistent with this paper: > > > >http://www.orcca.on.ca/~ldragan/synasc2005/2005-synasc-scigmark-final... > > > > They show Java faster than C# on most of the benchmarks in the SciMark > > > suite. But not the Monte Carlo simulation to calculate Pi, which is > > > presumably the benchmark you are talking about (are you using just > > > this benchmark or all of the SciMark benchmarks?). > > > I quoted the combined figures for all benchmarks. The individual figures > > are: > > > Java: > > FFT 326 > > Jacobi 499 > > Monte C 71.8 > > Sparse 446 > > LU 579 > > > C# .NET: > > FFT 325 > > Jacobi 505 > > Monte C 96.5 > > Sparse 415 > > LU 629 > > > As you can see, the Monte Carlo benchmark is several times faster (was 27.0) > > without the unnecessary lock and the performance is basically identical > > between Java and C#. > > > > Note the authors of > > > this paper used an identical, non-synchronised random number generator > > > for all languages, therefore your comments about syncronization are > > > addressed by their approach. > > > They benchmarked an extremely old version of .NET that predated generics. > > > -- > > Dr Jon D Harrop, Flying Frog Consultancy > > Ltd.http://www.ffconsultancy.com/products/?e --~--~---------~--~----~------------~-------~--~----~ You received this message because you are subscribed to the Google Groups "JVM Languages" group. To post to this group, send email to [email protected] To unsubscribe from this group, send email to [EMAIL PROTECTED] For more options, visit this group at http://groups.google.com/group/jvm-languages?hl=en -~----------~----~----~----~------~----~------~--~---
