January 28th, 2011, 10:46 am
@CuchulainnMatrix algebra is usually memory limited. Also, make sure that C++ code has debug checks disabled. MS compilers use security checks even in release mode. The same is true for UBLAS.Straightforward ALU or memory limed code won't run slower on C# or Java. The proper way to do numeral benchmarking is by comparing complete optimized modules. Compare MKL (Fortran/C), IPP (C/C++), Atlas (Fortran/C), Eigen (C++) or FFTW (C) vs the best C# implementation (is there any?).@jikan. C can't be faster than C++, since C++ includes C as its subset. All C++ compilers can compile C code as well. Apples to apples C++ is no faster or slower than C, since they are the apples from the same tree.
Last edited by
renorm on January 27th, 2011, 11:00 pm, edited 1 time in total.