Gawelek said:
Month ago I carried out some testes, and MC++ was not far behind C++.
And C# was 4x slower then C++, but faster then Java.
My tests were not professional, I am sure, but they show something.
It will be entirely dependent on what you're doing - and how you
measure it. (For instance, in Java you need to let the JIT "warm up"
and find hot spots; in C# you need to at least run the code once so it
will have been JITted; for both you should only measure the time taken
to execute the code rather than the total time for the program, etc.)
You should also be aware that some benchmarks will be optimised away
entirely by the compiler - there was a recent benchmark where someone
was saying how much faster C++ was than .NET (or Java, I can't remember
which) for trigonometric functions: it turned out that because he
wasn't using the results of the trig functions, the C++ compiler wasn't
even calculating them. Once the benchmark was corrected, the speeds
were fairly comparable.
Ultimately, it's only the performance of the real-world app you're
interested in which is relevant. If my app runs find under .NET, that
doesn't necessarily mean that yours will - or vice versa.