gmat-js

The implementation use TypedArray and ArrayBuffer, and matrices are passed by reference to functions, e.g al.normalize_r(o) normalize the ith row of matrix passed as o.x (run next example in RunKit )

const {al} = require("gmat-js");
const p = 10;
const i = 2;
const x = new Float64Array( new ArrayBuffer(p*p*8)); //matrix p x p in array form
for (let i=0; i < p*p; i++){
        x[i] = Math.random();
    }

al.normalize_r({
  x: x, //pass the array as reference
  p: p, //pass the dimension
  i: i  //pass the row index
});

//compute the squared norm of the ith row
console.log(
  al.scalar_r({
    x: x,
    p: p,
    i: i,
    j: i
  })
  );

Command line

If you install this module with the option -g you can then use it in the command line as,

gmat-sample 100 0.02 > matrix_100_0.02.txt 

The the matrix can be loaded in R, with the following code,

d <- read.table('matrix_1000_0.02.txt', numerals = 'warn.loss')
p <- sqrt(dim(d)[1])
K <- matrix(d[,1], nrow=p)
kappa(K)
eigK <- eigen(K)$values
hist(eigK, breaks = 100)

## spaces distribution
spaces <- p*sapply(p:2, function(i) return( eigK[i- 1] - eigK[i]))/(2 * pi) 
hist(spaces, breaks = "FD")