Integers turn out to be more valuable than I had thought initially. We already have basic scaffolding in place (see DESIGN.md), such as the wide regfile being designed to hold both floats and ints, but the operators and compiler support are missing.
Once we have integers, we will need float bitlevel operators (they are very simple): ldexp/frexp
True division first converts its arguments into float and the uses fdiv.
There is a number of examples that currently substitute integers with floats; also, integer support is needed for for loop back-edge implementation (we don't want floats for iteration counting).
Integers turn out to be more valuable than I had thought initially. We already have basic scaffolding in place (see DESIGN.md), such as the wide regfile being designed to hold both floats and ints, but the operators and compiler support are missing.
Once we have integers, we will need float bitlevel operators (they are very simple): ldexp/frexp
True division first converts its arguments into float and the uses fdiv.
There is a number of examples that currently substitute integers with floats; also, integer support is needed for
forloop back-edge implementation (we don't want floats for iteration counting).