Why is 10^5 more expensive than 10e5?

As Jan mentioned in this post, what's the reason behind it?

Answer 1

[dpb]

10^N is a computational expression whereas 10EN is a literal constant.  So, the former requires computation (and the exponential is moderately compute-intensive) whereas the latter simply is a conversion from character to numeric representation.  Takes (comparatively) little processing internally to do so.

[Mahdi]

Thank you :)

[dpb]

One additional comment and a difference between compiled languages and interpreted ones is that 10^5 is, while a numeric expression, ultimately a constant. Compiled languages will evaluate this at compile-, not run- time and store the result in the object file so from a computational standpoint in that case they're the same.

The difference is that in an interpreted language, the expression is not evaluated until run-time(+) and if it were in a loop it would be evaluated every time as well.

(+) This is w/o the JIT compiler, of course, which may (or again may not, TMW doesn't and won't say :) ) be able to recognize the constant and do it only once, too. But, in general, the admonition to avoid computing constants in Matlab if possible is still a good suggestion.