Macros¶

Detailed information on macros in julia is found here.

@def¶

Given some parameters:

using Parameters

@with_kw immutable Vpara @deftype Float64
    m  = 2.6887e+03
end
pa = Vpara(); # initialize parameter set

Instead of unpackaging the same parameters each time in a nested function like this:

function outer_f(pa::Vpara)
  num = zeros(Float64, (10,1))
  for i in 1:10
   num[i] = inner_f(pa::Vpara,i)
  end
  return num
end

with:

function inner_f(pa::Vpara,i)
  @unpack_Vpara pa
   m + i + 0.1
end

We define a macro as:

macro def(name, definition)
  return quote
    macro $name()
      esc($(Expr(:quote, definition)))
    end
  end
end

then redefine inner_f1() as:

@def inner_f2 begin
  m + i + 0.1
end

We also need to modify the outer_f() as:

function outer_f(pa::Vpara)
  @unpack_Vpara pa
  num = zeros(Float64, (10,1))
  for i in 1:10
    num[i] = @inner_f2
  end
  return num
end

The @def macro is functionally equivalent to copying and pasting the contents of inner_f() into outer_f().

rr=RemoteChannel()
# stores a value -> like a container that you can put something in and take something out

adding something to the container:

put(rr,0)

to add some processors:

addprocs()

calling a processor:

remotecall(function, proc,data)
# don't use this all of the time -> sort of low level

Example:

addprocs(1)
rc=remotecall(rand,2,100,100)
# then later...
f=fetch(rc)

An easier way to do this example is using @span:

r=@spawn rand(100,100)
# it picks a processor and sets this up for us

When you are typing you are typing to processor #1.

Number of processors:

nprocs()

Be able to run something everywhere:

@everywhere println(myid())

Also try:

fetch(@spawn myid())

Another useful macro:

@parallel
# works with ``for`` loops
#a=zeros(20) # wrong --> returns all 0's
a = SharedArray(Float64,20)
@parallel for i=1:20
  a[i]=i;
end