Skip to content

Latest commit

 

History

History
70 lines (49 loc) · 3.53 KB

index.md

File metadata and controls

70 lines (49 loc) · 3.53 KB
Raw

@def title = "Dagger.jl Home" @def tags = ["parallel", "hpc", "julia"]

<img alt="Dagger.jl Logo" src="assets/logo-final-medium.jpg" style="width:100%;padding:0%" />

Unleashing Parallelism with Dagger.jl

Making code run fast used to be difficult - no longer! Dagger.jl supercharges your code, letting it run in parallel on any and all hardware you have at your disposal.

Dagger doesn't expect you to learn about multithreading, distributed communication, GPU programming, atomics and memory models, etc. It could take years just to get up to speed on accelerating your code by hand! Instead, Dagger does that hard work for you - all you have to do is write your code!

Does this sound interesting? If so, keep reading to find out how to make your code go fast with Dagger.jl!

Installing Dagger.jl

Once you've installed Julia, run the following to start using Dagger:

(MyJuliaProject) pkg> add Dagger
julia> using Dagger
julia> mycode!() = println("I'm using Dagger!")
julia> fetch(Dagger.@spawn mycode!())

Using Dagger.jl

See Dagger.jl's documentation on GitHub to get going, or check out any of our other excellent recommended resources for a different learning experience!

Why use Dagger.jl instead of ?

Other libraries and frameworks make you work hard to get good performance:

  • dask makes you choose between multithreaded and distributed acceleration
  • MPI forces you to write your code in an SPMD style
  • OpenMP expects your code to be statically compiled C, C++, or Fortran
  • Spark tries to get you to pay for extra libraries to be productive
  • Hadoop is slow and is painful to configure

Why use Dagger.jl instead of accelerating by hand?

It's never a bad thing to learn how to write fast, parallel code, but it's also time- and experience-intensive. Sometimes you just can't afford to do all the learning and still be productive!

Don't get me wrong, Julia has great facilities for writing fast code:

  • A state-of-the-art compiler with support for many CPU architectures
  • Built-in multitasking and multithreading (Threads.@spawn)
  • Easily accessible distributed computing with Distributed.jl
  • Vendor-agnostic GPU computing with KernelAbstractions.jl
  • Supercomputer-scale communication with MPI.jl

Each of these tools is amazing in isolation, and in many cases, can be mixed-and-matched to gain the benefits of each tool. But as your code grows in features, so too it grows in complexity - and that complexity is amplified by the assumptions and quirks of each tool that your code relies on.

Dagger is no stranger to these tools - it uses them under the hood to make your code fast! But Dagger frees you - the busy developer, scientist, engineer - from having to learn how to best utilize these tools and maintain support for them. Dagger gives you a small (but powerful!) set of tools that cooperate to make your code fast, and keep you productive!

Contributing to Dagger.jl

Dagger.jl is a Free and Open Source, MIT-licensed project, written entirely in the Julia programming language, and hosted on GitHub. We welcome users to file issues or suggest improvements, or if you have the time and comfort with Julia programming, we'd love to review your contributions!