Abstract

We describe an implementation of floating point numbers in Reversible Forth using immutable references, and outline the advantages and disadvantages of this approach for the user and system. Via a probabilistic algorithm example which requires a large number of random trials to create a sufficient sample size, we demonstrate the disadvantage of a naive program with regard to garbage creation, and its semi-automatic resolution using inbuilt features of the reversible virtual machine. This allows otherwise prohibitively memory-intensive operations to be split into manageable pieces, interim results being saved while garbage is collected after each stage.

Citations