Intermediate code is executable on a virtual machine. Hence, source code that is compiled on a first computing device can be distributed to a second computing device in an intermediate form as long as the second computing device has the appropriate virtual machine. In this way, the compilation can be done without regard to the underlying native processor on the second computing device. However, the intermediate code loses the original type information defined in the source code. By determining the variable types in the intermediate code, the intermediate code can be converted to a different intermediate language.