Java Performance Tuning

Switches

The Java bytecode specification allows a switch statement to be compiled into one of two different bytecodes. One compiled switch type works as follows:

Given a particular value passed to the switch block to be compared, the passed value is successively compared against the value associated with each case statement in order. If, after testing all cases, no statements match, then the default label is matched. When a case statement that matches is found, the body of that statement and all subsequent case bodies are executed (until one body exits the switch statement, or the last one is reached).

The operation of this switch statement is equivalent to holding an ordered collection of values that are compared to the passed value, one after the other in order, until a match is determined. This means that the time taken for the switch to find the case that matches depends on how many case statements there are and where in the list the matched case is. If no cases match, and the default must be used, that always takes the longest matching time.

The other switch bytecode works for switch statements where the case values all lie in a particular range (or can be made to lie in a particular range). It works as follows:

Given a particular value passed to the switch block to be compared, the passed value is tested to see if it lies in the range. If it does not, the default label is matched; otherwise, the offset of the case is calculated and the corresponding case is matched directly. ...