160ijk
161ijk
162ijk
163ijk
164ijk
165ijk
166ijk
167ijk
Floating products of
(Sj)
and
(Vk
elements) to
Vi
elements
Floating products of
(Vj
elements)
and
(Vk
elements)
to
Vi
elements
Half-precision
rounded
floating products of (Sj)
and
(Vk
elements) to
Vi
elements
Half-precision
rounded
floating products of
(Vj
elements)
and
(Vk
elements) to
Vi
elements
Rounded
floating products of (Sj)
and
(Vk
elements)
to
Vi
elements
Rounded
floating products of
(Vj
elements)
and
(Vk
elements) to
Vi
elements
Reciprocal
iterations;
2 - (Sj) *
(Vk
elements) to
Vi
elements
Reciprocal
iterations;
2 -
(Vj
elements) *
(Vk
elements) to
Vi
elements
~-----------------------------------------------------
----------------
These
instructions are executed in the floating point mUltiply
unit.
The
number
of operations performed
by
an
instruction
is
determined
by
the contents of the
VL
register.
All
operations
start
with element
zero of the
V
registers
and
increment the element
number
by
one
for
each
success operation.
Operands
are
assumed
to
be
in floating point format.
Even-numbered
instructions in the
group
deliver (Sj) to the functional
unit
for
each
operation
as
one
of the operands.
The
other operand
is
an
element of
Vk.
For
odd-numbered
instructions in the group, both operands are
obtained
from
V
registers.
All
results
are delivered to elements of
Vi.
If
the operands are
unnorma11zed.
there
is
no
guarantee
that
the products will
be
normalized.
Out
of range conditions are described in Section 3.
2240004
4-63
E