Can someone explain how Julia MT got a super-linear improvement in performance by using 4 cores + HT vs 1 core? Fortran’s MT performance is more along the lin

Can someone explain how Julia MT got a super-linear improvement in performance by using 4 cores + HT vs 1 core? Fortran’s MT performance is more along the lines of what I would expect.