Marie-Hélène Burle

training@westgrid.ca
March 04, 2020

Brief history


Started in 2009 by Jeff Bezanson, Stefan Karpinski , Viral B. Shah , and Alan Edelman

Launched in 2012 as free and open source software

Version 1.0 released in 2018

Why another language?


Computer languages mostly fall into two categories:

         Compiled languages

         Interpreted languages

Compiled languages


Interpreted languages


JIT compilation


Just-in-time compilation (JIT) based on LLVM

Source code compiled at run time

Multiple dispatch


Built-in multiple dispatch : functions apply different methods at run time based on the type of the operands


Optional type declaration

Documentation


Julia website

The official Julia manual

Online training material

The Julia YouTube channel

The Julia Wikibook

A blog aggregator for Julia

Getting help

Discourse forum

[julia] tag on Stack Overflow

Slack  

#julialang hashtag on Twitter

Subreddit  

Gitter channel

#julia IRC channel on Freenode

Nice ways to run Julia

Emacs

julia-emacs with julia-repl
ESS
EIN for Jupyter notebooks

Juno

A Julia IDE built on Atom

Jupyter

Project Jupyter has a Julia kernel

REPL


REPL keybindings

C-c		cancel
C-d		quit
C-l		clear console

C-u		kill from the start of line
C-k		kill until the end of line

C-a		go to start of line
C-e		go to end of line

C-f		move forward one character
C-b		move backward one character

M-f		move forward one word
M-b		move backward one word

C-d		delete forward one character
C-h		delete backward one character

M-d		delete forward one word
M-Backspace	delete backward one word

C-p		previous command
C-n		next command

C-r		backward search
C-s		forward search

Where to find packages?


Easy search engine for registered packages (all on GitHub)

Managing packages in Pkg mode


(env) pkg> add <package>        # install <package>
(env) pkg> rm <package>         # uninstall <package>
(env) pkg> up <package>         # upgrade <package>

(env) pkg> st                   # check which packages are installed
(env) pkg> up                   # upgrade all packages


By default, installed in ~/.julia

Loading a package


> using <package>

Data types


> typeof(2)

> typeof(2.0)

> typeof("hello")

> typeof(true)

Indexing


Indexing starts at 1 , not 0

> a = [1 2; 3 4]

> a[1, 1]

> a[1, :]

For loops


> for i in 1:10
    println(i)
end


> for i in 1:3, j = 1:2
    println(i * j)
end

Conditionals


> a = 2
> b = 2.0

> if a == b
    println("It's true")
else
    println("It's false")
end

# Terse format
> a == b ? println("It's true") : println("It's false")

Functions

> function addTwo(a)
    a + 2
end

> addTwo(3)

# Terse format
> addtwo = a -> a + 2

# With default argument
> function name(a = "unknown")
    println("My name is $a.")
end

> name("Julia")
> name()

Plotting


Fun: plots in the command line!

> using UnicodePlots

> UnicodePlots.histogram(randn(1000), nbins=40)


This can be useful in remote sessions

Plotting


Nicer looking plots

> using Plots, Distributions, StatsPlots
> gr() # Using the GR framework as backend

> x = 1:10; y = rand(10, 2);
> p1 = Plots.histogram(randn(1000), nbins=40)
> p2 = plot(Normal(0, 1))
> p3 = scatter(x, y)
> p4 = plot(x, y)
> plot(p1, p2, p3, p4)


The Plots site has demos

Parallel programming

Launching Julia on multiple threads


Set the environment variable:

$ export JULIA_NUM_THREADS=n


Or launch a julia session with:

$ JULIA_NUM_THREADS=n julia


See how many threads are used in a julia session:

> Threads.nthreads()

When is parallelism happening?


Non parallel code

> for i = 1:10
    println("Iteration $i ran on thread $(Threads.threadid())")
end


Parallel code

> Threads.@threads for i = 1:10
    println("Iteration $i ran on thread $(Threads.threadid())")
end

Effect on timing


Let’s do a simple loop with 10,000,000 iterations

Non parallel code

> @time for i = 1:10000000
    i ^ i
end


Parallel code

> @time Threads.@threads for i = 1:10000000
    i ^ i
end

Let’s move on to the cluster

Loading the Julia module


# Look for available julia modules
$ module spider julia

# See modules required to load julia 1.3
$ module spider julia/1.3.0

# Load required gcc module and julia module
$ module load gcc/7.3.0 julia/1.3.0

Job script


#!/bin/bash
#SBATCH --job-name=julialoop		# job name
#SBATCH --time=00:00:30				# max walltime 30s
#SBATCH --cpus-per-task=32   		# number of cores
#SBATCH --mem=100					# max memory (in MB)
#SBATCH --output=julialoop%j.out	# output file name
#SBATCH --error=julialoop%j.err		# errors file name

echo Running non parallel loop on $SLURM_CPUS_PER_TASK cores
JULIA_NUM_THREADS=$SLURM_CPUS_PER_TASK julia loop.jl
echo Running parallel loop on $SLURM_CPUS_PER_TASK cores
JULIA_NUM_THREADS=$SLURM_CPUS_PER_TASK julia ploop.jl

Submit job


$ sbatch job_julialoop.sh


Check its status


$ sq

PD : pending       R : running

Results


Running non parallel loop on 32 cores
0.810377 seconds
Running parallel loop on 32 cores
0.093013 seconds (31.92 k allocations: 1.785 MiB)


89% faster