Ttykuu

I'm just starting to learn Julia, I work primarily in physics and am used to writing most of my code in Fortran90 and occasionally Python for Tensorflow (also Mathematica but that's less relevant). Julia has been recommended to me and I started checking it out; I like it a lot in theory as a middle ground between the speed of Fortran and the syntax of Python. To test it out I wrote a simple 2D Ising model code implementing a basic single-spin-flip Metropolis Monte Carlo algorithm. However, this code runs very slowly compared to an equivalent code in Fortran. Am I doing something wrong which is significantly affecting the performance of the code? I know almost nothing beyond what I've done here. I am using the Juno IDE in Atom on Windows 10. As an aside, I would also like to know how I can make multiple plots in the Atom plot tab, but that's secondary.

using Printf
using Plots

L = 20
n_sweep = 20
n_therm = 1000
n_data = 100
temps = 4.0:-0.3:0.1
e1 = Array(1:n_data)
m1 = Array(1:n_data)
et = []
mt = []

energy = e1
magnetization = m1

s = ones(Int32,L,L)


function measure(i)
 en = 0
 m = 0
 for x = 1:L
 for y = 1:L
 u = 1+mod(y,L)
 r = 1+mod(x,L)
 en -= s[x,y]*(s[x,u]+s[r,y])
 m += s[x,y]
 end
 end
 energy[i] = en
 magnetization[i] = abs(m)
end

function sweep(n,T)
 for i = 1:n
 for x = 1:L
 for y = 1:L
 flip(x,y,T)
 end
 end
 end
end

function flip(x,y,T)
 u = 1+mod(y,L)
 d = 1+mod(y-2,L)
 r = 1+mod(x,L)
 l = 1+mod(x-2,L)
 de = 2*s[x,y]*(s[x,u]+s[x,d]+s[l,y]+s[r,y])
 if (de < 0)
 s[x,y] = -s[x,y]
 else
 p = rand()
 if (p < exp(-de/T))
 s[x,y] = -s[x,y]
 end
 end
end

for T in temps
 sweep(n_therm, T)
 energy = e1
 magnetization = m1
 for i = 1:n_data
 sweep(n_sweep, T)
 measure(i)
 end
 en1 = sum(energy)/n_data
 ma1 = sum(magnetization)/n_data
 push!(et,en1/(L*L))
 push!(mt,ma1/(L*L))
 @printf("%8.3f %8.3f n", en1/(L*L), ma1/(L*L))
end

plot(temps,mt)