A test vector by Wilkinson.
x = accsum_wilkinson ( r )
a 1-by-1 matrix of doubles, a positive integer
a 1-by-r, matrix of doubles
Returns a vector x of size 2^r where r is a flint. Exercise 4.2 in SNAA
// Example with r = 3 x = accsum_wilkinson (3) // Expected result e = [1 , 1-%eps , 1-2*%eps , 1-2*%eps , 1-4*%eps , 1-4*%eps , 1-4*%eps , 1-4*%eps]' // Use with larger r x = accsum_wilkinson(10); s=sum(x) [s,e] = accsum_compsum ( x ) [s,e] = accsum_dblcompsum ( x )
"Stability and numerical accuracy of algorithms", Nicolas Higham
"Handbook of Floating Point Computations", Muller et al