Why the bounds check doesn't get eliminated?

Question

I wrote a simple benchmark in order to find out if bounds check can be eliminated when the array gets computed via bitwise and. This is basically what nearly all hash tables

Answer 1

In order to safely eliminate that bounds check, it is necessary to prove that

h & (table.length - 1)

is guaranteed to produce a valid index into table. It won't if table.length is zero (as you'll end up with & -1, an effective-noop). It also won't usefully do it if table.length is not a power of 2 (you'll lose information; consider the case where table.length is 17).

How can the HotSpot compiler know that these bad conditions are not true? It has to be more conservative than a programmer can be, as the programmer can know more about the high-level constraints on the system (e.g., that the array is never empty and always as a number of elements that is a power-of-two).