Takes two matrix descriptors in X and Y, and a
real number z. Returns
.
Thus a scalar multiple of one matrix is added to
another matrix. The multiply adds are done in-
ternally in high precision and results should be
exactly rounded.
Takes a descriptor of a square matrix in X and
inverts the matrix in-situ. Doesn't alter the stack.
Takes a matrix descriptor in X , saves it in L,
and returns a value suitable for ISG or DSL loop-
ing in X. The loop processes all elements in the
matrix. The loop index is DSL if the descriptor is
negative and ISG else.
Takes a matrix descriptor in X and a column
number in Y. Returns a loop counter in X, drop-
ping the stack. The matrix descriptor is saved in
L. The loop processes all elements m
iy
only. The
loop index is DSL if the descriptor is negative
and ISG else.
Takes a matrix descriptor in X, saves it in L, and
returns a loop counter in X. The loop processes
all elements along the matrix diagonal, i.e. all
elements m
ii
. The loop index is DSL if the de-
scriptor is negative and ISG else.
Takes a matrix descriptor in X and a row num-
ber in Y. Returns a loop counter in X, dropping
the stack and setting last L like all two-argument
commands. The loop processes all elements m
yi
only. The loop index is DSL if the descriptor is
negative and ISG else.
Takes two matrix descriptors in Y and Z and the
integer part of x as the base address of the re-
sult. Returns
. The fractional part
of x is updated to match the resulting matrix no
overlap checking is performed.
All calculations are done internally in high preci-
sion, although it would still be possible to trick
the code up and produce bad results. It would be
very difficult to get the same degree of accuracy
in RPN since the best that can easily be
achieved there is a·b+c·d and a matrix multiply
adds more terms than this.
To Next Page
Loading...
Need help?
General