The perspex machine arose from the unification of projective geometry with the Turing machine. It uses a total arithmetic, called transreal arithmetic, that contains real arithmetic and allows division by zero. Transreal arithmetic is redefined
here. The new arithmetic has both a positive and a negative infinity which lie at the extremes of the number line, and a
number nullity that lies off the number line. We prove that nullity, 0/0, is a number. Hence a number may have one of four
signs: negative, zero, positive, or nullity. It is, therefore, impossible to encode the sign of a number in one bit, as floating-point arithmetic attempts to do, resulting in the difficulty of having both positive and negative zeros and NaNs.
Transrational arithmetic is consistent with Cantor arithmetic. In an extension to real arithmetic, the product of zero, an
infinity, or nullity with its reciprocal is nullity, not unity. This avoids the usual contradictions that follow from allowing
division by zero. Transreal arithmetic has a fixed algebraic structure and does not admit options as IEEE, floating-point
arithmetic does. Most significantly, nullity has a simple semantics that is related to zero. Zero means "no value" and
nullity means "no information." We argue that nullity is as useful to a manufactured computer as zero is to a human computer.
The perspex machine is intended to offer one solution to the mind-body problem by showing how the computable aspects
of mind and, perhaps, the whole of mind relates to the geometrical aspects of body and, perhaps, the whole of body. We
review some of Turing's writings and show that he held the view that his machine has spatial properties. In particular, that
it has the property of being a 7D lattice of compact spaces. Thus, we read Turing as believing that his machine relates
computation to geometrical bodies.
We simplify the perspex machine by substituting an augmented Euclidean geometry for projective geometry. This leads to
a general-linear perspex-machine which is very much easier to program than the original perspex-machine. We then show
how to map the whole of perspex space into a unit cube. This allows us to construct a fractal of perspex machines with the
cardinality of a real-numbered line or space. This fractal is the universal perspex machine. It can solve, in unit time, the
halting problem for itself and for all perspex machines instantiated in real-numbered space, including all Turing machines.
We cite an experiment that has been proposed to test the physical reality of the perspex machine's model of time, but we
make no claim that the physical universe works this way or that it has the cardinality of the perspex machine. We leave it
that the perspex machine provides an upper bound on the computational properties of physical things, including
manufactured computers and biological organisms, that have a cardinality no greater than the real-number line.