I’ve heard many times that jagged arrays tend to have better access speeds and also that multidimensional arrays usually use less memory and can be allocated more quickly but exactly how big are these differences (in terms of actual numerical figures)?
I’m sure it depends on the language, runtime, platform, etc. In something like Mono/C# the differences are probably minuscule to non-existent. If need hard data, you should write a test yourself in your target environment, as other people’s tests are not guaranteed to be identical. Ultimately, the choice in representation should be made based on which is a more natural expression of the functional logic. Don’t worry about minuscule performance differences like this until they show up in a profiler (which they probably will not).