GSL Shell main data types are real and complex matrices and many of the GSL routines works on this kind of data. GSL Shell does not have a separate type to rapresent a vector but we use column matrix for this purpose.
With matrix you can do basic algebraic operation just by using the ‘+’ and ‘*’ operators to perform element by element operations with the matrix operands. You can argue that for multiplication the element-wise multiplication is not very interesting from the algebric point of view. No problem about that since with GSL shell you have two function to perform matrix multiplications, the mul() and the prod() fuctions. The first one take an arbitrary number of arguments and perform the matrix products of its arguments. The prod() function instead takes exactly two arguments and perform the matrix product of the transpose of the first with the second argument. So, for example, if x and y are two column matrices the expression prod(x, y) gives their scalar product.
In order to create a new matrix you should use the new() function for a real matrix and cnew() for a complex matrix. Most of GSL Shell function comes in two variants for complex and real number. Generally the complex variant has the same name of the real one but with ‘c’ prefix.
The new() function takes two mandatory arguments, the number of rows and of columns and an optional third argument. If the third argument is not given all the elements of the matrix are initialised to 0. Otherwise, if you provide a function of two variables, lets say i and j, it will be used to initialise the element of the matrix. The provided function will be called for each element of the matrix with the index of row and column as an argument. This provides a very useful way to initialise matrices. As an example, let us suppose that we want to define a matrix m of dimension whose elements are given by
.
In GSL Shell we can define the matrix with the following command:
-- we assume that n is a positive integer number
m = cnew(n, n, |j,k| exp(2i*pi*(j-1)*(k-1)/n))
where you have previously defined n to be a small integer number.
Here an examples that shows how to calculate the exponential of a 2x2 matrix by using the elementary Taylor expansion of the exponentian function:
function mexp(m, nmax)
local n = m:dims()
local r = unit(n)
local p, f = r, 1
for i=1, nmax do
p, f = mul(p, m), f/i
r = r + f * p
end
return r
end
This function get two arguments.
All the functions described in this section have an equivalent function for complex valued matrix. The functions for complex matrix are obtained by adding the ‘c’ prefix to the name. So for example the function mul() has an equivalent for complex matrix whose name is cmul().