Table 3-9 Vu Divide Instruction Summary - Nintendo Ultra64 Programmer's Manual

Vector Unit Instructions
Type
vnop
vmov
vrcp, vrsq
vrcph,
vrsqh
76
The reciprocal (rcp) or reciprocal of the square root (rsq) of the scalar
vt
element of
of vd .
The scalar element of
contents of vs
Single (16 bit) and double (32 bit) precision source values are supported,
with double precision sources supplied in two instructions.
The destination is a 32 bit value, written to two register elements in two
instructions.
For single precision sources, vrcp/vrsq supplies the source operand in
vt element
[ ] and the low 16 bits of the result is written to
16 bits of the result is written by a subsequent vrcph/vrsqh.
For double precision sources, vrcph/vrsqh supplies the upper 16 bits of
the source (and writes the upper 16 bits of a previous vrcp/vrsq or
vrcpl/vrsql). A subsequent vrcpl/vrsql supplies the low 16 bits of the
source and writes the low 16 bits of the result.
The vmov type simply copies
reordering scalar data.
vnop ignores
The following table shows the source and destination operand bits from
vt
each the and

Table 3-9 VU Divide Instruction Summary

vt[element] vd[vs]
NA
low low
high high
is computed by table lookup and written to the scalar element
vd
is selected by the register number
, but the instruction field
vt element
[
vd
, and no register is written.
vd
elements.
no operation
write source to result
lookup source and write result
set source, write previous result
vs
(0-7). Not the
vs bits.
vd vs
[
vd vs
] to [ ], and is useful for
]. The upper