A program value has both an abstract meaning and a concrete representation. The representation determines how much memory the value occupies and what information must travel with it at run time.

Fixed-width values

Machine-oriented integer and floating-point types commonly have a fixed width such as 8, 32 or 64 bits. Their size is known independently of the particular value stored.

Larger representations can encode a larger range or more precision, but they consume more memory and move more data through caches and memory buses.

Representation is not the type name

The same mathematical kind of value can be represented differently by different languages or runtimes. A value may be stored directly in a fixed-size slot, boxed behind a reference, or accompanied by run-time metadata such as a type tag.

Dynamic typing therefore does not make values physically "sizeless"; it changes where and when representation information is carried.

Collections amplify representation cost

A few extra bytes may be irrelevant for one value but significant across millions of elements. Representation choices therefore connect language semantics with memory footprint and cache behaviour.