[julia-users] Tail call optimization

[Sisyphuss]

I wrote 3 versions of factorial: by recursion, by iteration, by for-loop

rec_factorial(n) = n == one(n) ? one(n) : n * rec_factorial(n-1)

function iter_factorial(n)
  fact_iter(one(n),one(n),n)
end

function fact_iter(product, counter, max_count)
  if counter > max_count
    product
  else
    fact_iter(product * counter, counter + one(counter), max_count)
  end
end

function myfactorial(n)
  product = one(n)
  for i = one(n):n
    product *= i
  end
  product
end

Then I timed it:
@timeit rec_factorial(10000)       # 28.84 µs
@timeit iter_factorial(10000)      # 24.15 µs
@timeit myfactorial(10000)         #  8.06 µs
This result is robust.

Now, the question is how to explain tis result:
1) why version 2 is slightly faster than version 1?
2) why version 3 is mush faster than the other 2? Has the tail call 
optimization really worked here?