rragno (9/23/2012)
Unlike the C# code, this always takes 0 ms to execute. It appears to simply analyze the code and conclude that only the last iteration of the loop will matter. If you change the Mult function to accumulate instead, it no longer can do that. Still, it then takes only 6 ms to execute, which is much faster than the C# version.)
So the Java optimiser is somewhat better than the C# optimiser.
Of course I would have expected pre-compiled code in either of these languages to run in 0 ms for the version with the accumulation as well, so these numbers suggest that optimisation techniques (for procedural languages like these) have deteriorated badly in the last four decades, or everything is doing recompilation all the time, or people have decide that some of the possible optimisations are rather pointless.
And it seems pretty clear that the C# optimisation is really bad.
Tom