Using vectors in single NLP constraint inside proc iml with call statement

[ayaa]

I have two variables that are stored in a data set and also are stored
as matrices and vectors. I want to formulate only one constraint that
is a function of all the data set's observations (or elements of these
vectors) and a decision variable. Suppose the two stored variables
(two vectors) are a and b, and x is a decision variable. My constraint
is: sum over i for [ (ai)^2 * bi/ (ai+x)] = 1, can you help me to
include this constraint inside proc iml in order to use call nlpqn ?