Gaussian elimination for binary matrices ( all elements in GF(2) ) implemented in numba python and numpy for efficiency.

gf2elim.py

import numpy as np
import numba

@numba.jit(nopython=True, parallel=True) #parallel speeds up computation only over very large matrices
# M is a mxn matrix binary matrix 
# all elements in M should be uint8 
def gf2elim(M):

    m,n = M.shape

    i=0
    j=0

    while i < m and j < n:
        # find value and index of largest element in remainder of column j
        k = np.argmax(M[i:, j]) +i

        # swap rows
        #M[[k, i]] = M[[i, k]] this doesn't work with numba
        temp = np.copy(M[k])
        M[k] = M[i]
        M[i] = temp


        aijn = M[i, j:]

        col = np.copy(M[:, j]) #make a copy otherwise M will be directly affected

        col[i] = 0 #avoid xoring pivot row with itself

        flip = np.outer(col, aijn)

        M[:, j:] = M[:, j:] ^ flip

        i += 1
        j +=1

    return M