Exploiting exceptions

Most programmers use exceptions to handle exceptional events. However, exceptions can also be used to simplify the control logic of a program and enhance readability [1]. As we will demonstrate, there are situations where exceptions can be used as a standard programming pattern to make programs execute faster. These situations are opportunities to increase the speed of Java programs by changing bytecode within methods with a space cost of, at most, a few extra instructions. Such modifications apply knowledge of which run-time actions and checks are performed by a virtual machine as bytecode instructions are executed. For instance, within Sun’s Java Virtual Machine (JVM), all object dereferences are preceded with a run-time check for a null value; if null, a NullPointerException is thrown; if not null, the object dereference proceeds. As observed by Orchard [2], this exception can be exploited in loops whose control expressions involve an explicit null check since the expressions may be redundant—the exception thrown as a result of dereferencing a null pointer may be used to transfer control out of the loop (Figures 1(a), (b)). Array-bounds checks present another opportunity. Before accessing any element of an array, the runtime system first checks whether the array index is within the array’s bounds; if not, the JVM throws an ArrayIndexOutOfBounds exception; otherwise the array access proceeds. In a similar manner to the previous example, the exception may be exploited in any loop whose control expression involves