I recently perused the AVM2 bytecode overview that Adobe provides and found something strange: they have special instructions for the first few local variables. Getting the first local variables is done by getlocal0, getlocal1, getlocal2, getlocal3, but then a generalized getlocal is used for local variables thereafter. This brought me to think about the performance implications of these specialized instructions. After all, why have them when you already have a generalized one? Read on to see the performance differences and if you can get any speed boost from rearranging your local variables.

Read the rest of this entry »

Tags: bytecode, ordering, performance, registers, variable