The .BASE directive is used to inform the assembler that a
    specified base register contains a specified base address. Later
    in your program, the assembler allows you to implicitly reference
    the specified base register. When the assembler knows which
    base addresses are stored in one or more base registers, it can
    convert an expression to an offset from one of the base registers
    previously specified in a .BASE directive. .BASE provides a
    convenient and more readable shorthand for accessing memory and
    constant values using base registers. .BASE also makes it easier
    for you to change your register assignments if you later modify
    your code.

    The base expression is optional. If the base expression is
    specified, this base address value is assumed by the assembler
    to be in the specified register, Rn. If the base expression is
    omitted, the contents of the specified base register, Rn, is
    considered undefined until a new base expression is associated
    with the base register.

    R31 is defined to always contain 0, according to the architecture
    definition. Therefore, R31 is known to be a predefined base
    register containing 0. For every assembly, the assembler assumes
    the following statement:

        .BASE R31, 0

    Because the contents of R31 cannot change, you cannot specify a
    base address for R31.

    You can use the .BASE directive to implicitly reference base
    registers. You can also automatically compute offsets from a base
    address known to be in a register to a base address you use in an
    instruction argument.

    Most of the memory format Alpha instructions are defined
    such that one of their arguments must have a base register and
    an offset. If the assembler encounters only an expression with
    no base register, the assembler attempts to find a base register
    that contains a base address or constant within a 16-bit signed
    offset of the value of the expression. If it finds such a base
    register, the assembler computes an offset that, when added to
    the value of the base register, results in a value equal to the
    expression specified in the instruction argument.