This computesthe same thing as the original post about 10x faster on my machine, without “cheating” using threads:
function test(s,a)
g = Vector{Float64}(s)
ap1= a+1
ap1inv = inv(ap1)
@inbounds for i =1:s
k = i+1.0
@fastmath g[i] = (k^ap1-(k-1)^ap1)*ap1inv
end
end