TornadoVM: A practical and efficient heterogeneous programming framework | Hacker News Reader

TornadoVM: A practical and efficient heterogeneous programming framework

github.com

67 pointscapableweb11 comments

Best description of what it is and why it's interesting:

> What can TornadoVM do?

> TornadoVM accelerates parts of your Java applications on heterogeneous hardware devices such as multicore CPUs, GPUs, and FPGAs.

> TornadoVM primarily supports Java. However, with the integration with GraalVM you can call your TornadoVM-compatible Java code through other programming languages supported by GraalVM's polyglot runtime (e.g., Python, R, Ruby, Javascript, Node.js, etc).

> Automatic code generation of OpenCL and Nvidia PTX, achieving portability across multi-core CPUs, GPUs and FPGAs.

We are currently implementing Tornado on a Clojure + Graal install. Even with our partial, non-optimised version, we are seeing a speed-up greater than an order of magnitude. I presently estimate an infrastructure savings of nearly 70% versus our current Azure spend. I will post a postmortem once completed if it would be of interest.

What type of hardware are you using? As the README advertises possibilities for multicore CPUs, GPUs, and FPGAs.

That would be really great if you could write more about that! Sounds incredible.

Will do. Any recommendations on where I should post our findings? I don't have any personal blogs.

Some people write GitHub gists as that's the simplest possible way to do it, probably (or the equivalent on BitBucket/GitLab etc) and it's pretty decent with good support for code samples.

Otherwise, if you want to start a real blog, I recommend a static site generator[1] and one of the free services like Netlify, GitHub Pages or Cloudflare (which has free, cool domains like your-name.pages.dev).

[1] https://jamstack.org/generators/

Medium is dead simple. Create an account and click the icon to write an article. Publish and share.

Sure it will be a postmortem and not a postpartum?

Any benchmarks?

You can find benchmarks here:

- https://jjfumero.github.io/files/VEE2019_Fumero_Preprint.pdf

- https://www.tornadovm.org/use-cases

- https://www.research.manchester.ac.uk/portal/files/190177400...

- https://www.research.manchester.ac.uk/portal/files/190176058...

- https://arxiv.org/pdf/2010.16304v1.pdf

Besides, TornadoVM integrates a benchmark script for users to run applications with Java JMH and OpenCL/PTX/Level-Zero profilers:

- https://github.com/beehive-lab/TornadoVM/blob/master/assembl...