mppf.bsky.social

mppf.bsky.social

@mppf.bsky.social

My goal: make parallel programming easier for everybody. Compiler developer for the Chapel programming language for productive parallel programming ( https://chapel-lang.org/ ). Also https://github.com/femto-dev/femto I work at HPE. He/him. OSS.

230 followers72 following78 posts

Pinned

If you haven’t see the 7 Questions series interviewing users of #ChapelLang, now is a great time to check it out! The latest post describes how Chapel enabled the rapid development of an interactive data analysis tool that runs on #HPC hardware. Link to series: chapel-lang.org/blog/series/...

chapel-lang.org7 Questions for Chapel UsersThis series interviews users of Chapel about their experiences with the language.
Chapel Programming Language@chapellanguage.bsky.social

Learn about Chapel’s use in cybersecurity by enabling interactive data analysis at HPC scales in our latest “7 Questions for Chapel Users” interview with Bill Reus, one of the original co-developers of Arkouda. chapel-lang.org/blog/posts/7...

I made the new scalable sort routine for the Chapel standard library! Now it's possible to easily sort a distributed array just by calling the standard library 'sort'. Outperforms the best distributed Chapel sort to date, and it performs competitively with the KaDiS AMS sort written in MPI. #HPC

Chapel Programming Language@chapellanguage.bsky.social

Chapel 2.5 is being released today! Highlights include a new scalable sort routine, an editions feature for experimental breaking changes, a new aliasing reshape() for arrays, initial support for VSCode debugging, dynamically loaded libraries, and more! chapel-lang.org/blog/posts/a...

I made a #blog post comparing memory safety across programming languages, with examples! It covers C, C++, Rust, Python, and Chapel. And it discusses memory safety in the context of distributed-memory programming with MPI, Chapel, or OpenSHMEM for #HPC. chapel-lang.org/blog/posts/m...

chapel-lang.orgMemory Safety in ChapelA description of how Chapel’s features for memory safety strike a balance between productivity and performance, with comparisons to other languages

I’m going to be co-hosting this birds of a feather session at JuliaCon!

Chapel Programming Language@chapellanguage.bsky.social

Attending JuliaCon this year? Join us at the “Chapel ❤️ Julia” session! Chapel and Julia have similar goals despite big differences in approach. Join this BoF for a brief overview of Chapel and a discussion where the two communities can learn from each other. juliacon.org/2025/

mppf.bsky.social reposted

How does Chapel stack up against MPI and OpenSHMEM for distributed memory parallel programming? Check out this repo to see implementations and performance measurements of a Least-Significant Digit First Radix Sort in these frameworks. github.com/mppf/distrib... #ChapelLang #MPI #OpenSHMEM

github.comGitHub - mppf/distributed-lsb: Comparing Distributed Programming Frameworks with LSD Radix SortComparing Distributed Programming Frameworks with LSD Radix Sort - mppf/distributed-lsb

mppf.bsky.social reposted

I’m very excited about the new Chapel Programming Language website. It’s the little things, right? Have a look if you have not already & I hope you like it.

Chapel Programming Language@chapellanguage.bsky.social

We've made some big changes to our website! It shows off the power of Chapel with a modern, smartphone-friendly appearance. Check out all the awesome projects people are using Chapel for today: chapel-lang.org #ChapelLang #OpenSource #HPC

I’m looking forward to this talk today at NASA discussing Chapel and CFD for aerodynamic simulations. #HPC

Chapel Programming Language@chapellanguage.bsky.social

For the first time, the Chapel team at HPE and the CHAMPS team at Polytechnique Montreal will give a joint talk open to the public. Join us on February 20th at 9AM PST in the AMS Seminar Series hosted by NASA Ames Research Center! Please message @chapellanguage.bsky.social for the meeting link!

View full thread

CHAMPS’ core is aimed at aerospace application, solving the URANS equations (time and spectral) with several turbulence models with transition and roughness for aero-icing and aeroelastic models, all available within CHAMPS.

I’m looking forward to this talk today at NASA discussing Chapel and CFD for aerodynamic simulations. #HPC

Chapel Programming Language@chapellanguage.bsky.social

For the first time, the Chapel team at HPE and the CHAMPS team at Polytechnique Montreal will give a joint talk open to the public. Join us on February 20th at 9AM PST in the AMS Seminar Series hosted by NASA Ames Research Center! Please message @chapellanguage.bsky.social for the meeting link!

View full thread

Though Chapel code tends to be concise and readable, the resulting performance can compete with or beat standard high-performance computing (HPC) approaches, such as Fortran/C/C++, OpenMP, MPI, and/or CUDA.

Apps written in Chapel span various domains, including data analytics, biodiversity detection, hydrology, cosmology. One example, the largest application developed using Chapel, is the CFD solver CHAMPS (Chapel Multi-Physics Software) in the laboratory of Prof. Laurendeau at Polytechnique Montreal.

I’m looking forward to this talk today at NASA discussing Chapel and CFD for aerodynamic simulations. #HPC

Chapel Programming Language@chapellanguage.bsky.social

For the first time, the Chapel team at HPE and the CHAMPS team at Polytechnique Montreal will give a joint talk open to the public. Join us on February 20th at 9AM PST in the AMS Seminar Series hosted by NASA Ames Research Center! Please message @chapellanguage.bsky.social for the meeting link!

View full thread

Chapel promotes scalable parallelism, permitting code to be developed using a laptop’s multi-core processor, and then recompiled to run across the nodes and cores of a commodity cluster, cloud instance, or massive-scale supercomputer.

I’m looking forward to this talk today at NASA discussing Chapel and CFD for aerodynamic simulations. #HPC

Chapel Programming Language@chapellanguage.bsky.social

For the first time, the Chapel team at HPE and the CHAMPS team at Polytechnique Montreal will give a joint talk open to the public. Join us on February 20th at 9AM PST in the AMS Seminar Series hosted by NASA Ames Research Center! Please message @chapellanguage.bsky.social for the meeting link!

mppf.bsky.social reposted

Learn about Chapel’s use in cybersecurity by enabling interactive data analysis at HPC scales in our latest “7 Questions for Chapel Users” interview with Bill Reus, one of the original co-developers of Arkouda. chapel-lang.org/blog/posts/7...

chapel-lang.org7 Questions for Bill Reus: Interactive Supercomputing with Chapel for CybersecurityAn interview with Bill Reus about the creation of Arkouda, a Python library supporting interactive data analysis on HPC systems

How far does the desire for absurd performance go? I would say, fully absurd performance is only achievable if the program is parallel. Parallel code certainly has additional concerns. Do those represent too much of a cost? 🧵 (1/4)

Sam Rose@samwho.dev

In fact, I think we do have glimpses of this world in tools like uv, bun, Zed. At least for me, the difference is so noticeable that I’m happy to accept a worse experience in most other areas provided the perf is sufficiently absurd.

View full thread

Recently, I rewrote a suffix array construction code I had written in C++ and OpenMP years ago. The OpenMP version only parallelized the slowest parts. Using Chapel, I made every part parallel, because it was easy. The result was 8x faster in about 1/4 the code. (3/4)

It certainly helped that Chapel’s standard library sort is really fast. I wrote a blog post comparing standard library sorts in chapel-lang.org/blog/posts/s... . A parallel standard library helps make performance easy to achieve, but for many languages it isn’t possible or isn’t a goal. (4/4)

chapel-lang.orgComparing Standard Library Sorts: The Impact of ParallelismThis blog post compares the performance of standard library sort routines in several programming languages.

How far does the desire for absurd performance go? I would say, fully absurd performance is only achievable if the program is parallel. Parallel code certainly has additional concerns. Do those represent too much of a cost? 🧵 (1/4)

Sam Rose@samwho.dev

In fact, I think we do have glimpses of this world in tools like uv, bun, Zed. At least for me, the difference is so noticeable that I’m happy to accept a worse experience in most other areas provided the perf is sufficiently absurd.

I didn’t know this was possible but the other day I broke a cast iron griddle. Normally it covers two burners on our induction cooktop. I was was cooking something with just one burner on and I turned it up to 11. Turned around to do something, and then there was a loud bang. A big crack appeared!

mppf.bsky.social reposted

OK, let's try this to show the HPC community already has momentum and can thrive here on Bluesky ... If #HPC or #supercomputing is of interest to you, please interact with this post in some way - either repost, quote, like, reply with insights or just your fav gif, tag friends, etc.

I gave a talk at the APL colloquium, now available online! It’s an introduction to the Chapel Programming language aimed at people with some Python programming experience. #hpc

This post is unavailable.