The Julia Language Seminar Talk Francisco Vidal Meca

Home , FFTW, Gnuplot, IPython, Notebook interface

Languages for Scientiﬁc Computing Aachen, January 16, 2014 Why Julia?

Many languages, each one a trade-oﬀ Multipurpose language: scientiﬁc computing machine learning data mining large-scale linear algebra distributed and parallel computing

"In short, because we are greedy. (We) want to have it all." [1]

The Julia Language 3 Julia and Scientiﬁc Computing

Julia is "A fresh approach to technical computing" High level and high performance Easy syntax Parallelism & Cloud computing Graphs Free and Open source Developed in the MIT Version 1.0 released February 2012

The Julia Language 4 About Julia

Multi-paradigm Homoiconic Dynamic Easy integration with C/Fortran Multiple-dispatch Just-in-time(JIT) LLVM-based compiler Free and open source Core under MIT license Libraries under GPL, LGPL, BSD Environment under GPL Numerical accuracy

The Julia Language 6 Parallelism and Cloud computing

Key factor in Julia’s design Building blocks for Example distributed computation function TicToc(N) Multiple worker processes tic(); Local nh = @parallel (+) for i=1:N Remote int(randbool()) Diﬀerent from other end; environments such as MPI s=toc(); println("Num. Heads: $nh ") Built on remote references println("in $s seconds") and remote calls end Simpliﬁed with macros, e.g., @paralell

The Julia Language 7 Mathematical functions

Extensive Math function library

Integrates C and Fortran libraries Linear Algebra Random number generation Signal processing String processing

The Julia Language 8 Integration

Calling external functions in C/Fortran Need to be Shared Library Called directly with ccal

Example ccal((:function,"library") ,RetType,InputTypes,Inputs)

julia> t = ccall( (:clock, "libc"), Int32, ()) 2292761

julia> t 2292761

The Julia Language 9 Performance

4 10

3 10

benchmark 2 rand_mat_mul 10 rand_mat_stat pi_sum printfd 101 mandel quicksort fib parse_int 100

10-1

Julia Fortran Go JavaScript Python RMathematica Matlab Octave

Figure: Benchmark times [1] relative to C 1.

1Smaller is better, C performance = 1.0 The Julia Language 10 Plotting with Julia

inset 1 1 0.5

0.5 −0.5

−1 Not in the core Julia System 0 1 2 3 4 5 6 0

Add-on package "Winston" −0.5

−1 Julia and GNUplot

0 1 2 3 4 5 6

Figure: Error bars and plot composition.

The Julia Language 11 title!

1 title

0.5 i

Θ 0

−0.5

−1

0 2 4 6 8 2 x label Σx i

100 top 10 100 1000 a points b points 1 1 80 slope

0.5 1/2 60 right 0 0 left 40

−0.5 −1/2

−1 −1

3 4 5 6 7 8 9 0 bottom 0 20 40 60 80 100 The Julia Language 12 Developer community

External packages Active community Built in package manager Mailing list IJulia GitHub Browser-based graphical Youtube channel notebook interface “JuliaLanguage” IPython and Julia StackOverﬂow community Users meetup (today in E.g.: Prof. Edelman’s San Francisco Bay Area) notes for the Parallel ... Computing class

The Julia Language 13 Conclusion

Uses Multi-purpose language BigFloats, Combinatorics, High level and easy syntax Statistics High performance LAPACK for Linear algebra Graphs SuiteSparse for Sparse Parallelism & Cloud factorizations computing Provides BLAS functions Integration wrappers Free and Open Source Signal processing, FFT functions from FFTW