Nintendo Ultra64, Programmer'S Manual

Page: 35 / 331

Revision 1.0 Vector Unit Registers
35
Instructions can operate on pairs of elements, adding two vectors (8 pairs of
numbers), for example.
VU registers can also be addressed as scalars, allowing you to add 1 number
(the same number) to a vector (8 numbers), for example.
Further, registers can be broken into scalar halves and scalar quarters,
allowing you to treat pieces of VU as subsets, performing the same
operations on consecutive ranges of elements. This is best understood with
an illustrated example, see Figure 3-8, “Scalar Half and Scalar Quarter
Vector Register Elements,” on page 59.
RSP assembly language syntax for vector registers is explained in the section
“Vector Register Element Syntax” in Chapter 5.
Loads, Stores, and Moves
VU loads, stores, and moves always reference data within VU registers by
their bytes. So if you want to load a short (2 bytes) into element 3 of a VU
register, you must do this:
lsv $v1[6], 0($1)
Element 3 corresponds to byte 6, of the VU registers.
Caution:
A very common programming error is to confuse the “byte
index” of a VU load/store with the “element index” of a computational
instruction.