ChapelCon '25 Retrospective #30
Conversation
Signed-off-by: Brandon Neth <[email protected]>
bradcray
left a comment
bradcray
left a comment
There was a problem hiding this comment.
This generally looks nice, and most of my embedded comments are small/minor/local. The "biggest" is probably that a number of the YouTube links are pointing to the original versions of the talks that had that frozen frame issue at the start and should be updated.
Outside of reading what's here, the main thing that strikes me is that it's a very non-visual article, making me wonder whether we could include a compelling slide/screenshot from some or all of the talks you call out. Could alternatively use the title slide and have it hyperlink to the slides or talk. This is obviously completely optional, but I think it would make the article a bit more enticing to look at and read through.
One other thing to consider is whether there are main points you'd like to highlight as pull quotes, which is another way to punch up the article for short attention spans / non-readers / their AIs. (not exclusive with the previous idea, you could do both).
| authors: ["Brandon Neth"] | ||
| --- | ||
|
|
||
| ChapelCon '25 is behind us, and another year of productive HPC programming lies ahead! |
There was a problem hiding this comment.
Something about "is behind us" sounds a bit like "good riddance" to me. Maybe "has wrapped up"
| --- | ||
|
|
||
| ChapelCon '25 is behind us, and another year of productive HPC programming lies ahead! | ||
| Don't worry if you missed a talk, you can find the recordings and slides on the ChapelCon '25 [webpage](https://chapel-lang.org/chapelcon25/). |
There was a problem hiding this comment.
Could use the https://chapel-lang.org/chapelcon25/#program anchor to focus more directly on where the recording/slide links appear.
| Before reflecting on the conference, I'd like to thank everyone who made ChapelCon '25 the exciting week it was. | ||
| Thank you to contributors and participants for the exciting dialogue during our demo sessions and conference days. | ||
| Thank you to the program committee for your support in reviewing a record-breaking number of submissions. | ||
| And, of course, huge thank yous to Luca Ferranti for his work as Program Committee Chair, Daniel Fedorin for his as Tutorial Day Chair, and Jade Abraham for theirs as Office Hours Chair. |
There was a problem hiding this comment.
Should that be "Tutorial Days Chair"? (or if that sounds awkward could just say "Tutorials Chair")
|
|
||
| ### Second Conference, First Experiments | ||
|
|
||
| As the second instance of ChapelCon, this year's conference was our first opportunity to experiment with some changes to the conference format. |
There was a problem hiding this comment.
Could have "second instance of" link either to last year's description or the article that announced the launch of ChapelCon.
| ### Second Conference, First Experiments | ||
|
|
||
| As the second instance of ChapelCon, this year's conference was our first opportunity to experiment with some changes to the conference format. | ||
| We tried out four changes to the conference: a season change from early summer to early autumn, an expansion from three days to four, a new format for tutorial days, and a new approach to the submission and review process. |
There was a problem hiding this comment.
I'd suggest turning these four things into bullets, just to add a little visual variance to this article and break up the text. It'd also make them a little easier to visually parse and iterate over.
| The second highlight came from Daniel Fedorin and demonstrates the expansive power of Chapel's type system. | ||
| Using Chapel's compile-time `param` values and a bit of clever thinking, his approach makes it possible to encode complex data structures and specialized functions, all at compile time. | ||
| This has wide-reaching implications, including eliminating runtime overhead and expanding compile-time error checking for functions like `printf`. | ||
| Check out the recording [here](https://youtu.be/-x1zYQSuaT4). |
| This has wide-reaching implications, including eliminating runtime overhead and expanding compile-time error checking for functions like `printf`. | ||
| Check out the recording [here](https://youtu.be/-x1zYQSuaT4). | ||
|
|
||
| Third, from our very own PC chair Luca Ferranti, was a talk on [automatic differentiation in Chapel](https://youtu.be/ioqxdmSprBM). |
There was a problem hiding this comment.
I might drop "very own" as it made me think "from HPE" whereas one of the things I really appreciate about Luca and his work is that he's not.
| The `ForwardModeAD` library uses Chapel's operator overloading to support derivatives, gradients, jacobians, and more, all in a composable way. | ||
| The cherry on top? A Chapel intergration for Enzyme, a library for automatic differentation at the LLVM-level! | ||
|
|
||
| Finally, closing out the conference, was [Iain Moncrief's talk on his machine learning library ChAI](https://youtu.be/bHCrfHNkW6E). |
There was a problem hiding this comment.
Another YouTube link needing updating.
|
|
||
| Finally, closing out the conference, was [Iain Moncrief's talk on his machine learning library ChAI](https://youtu.be/bHCrfHNkW6E). | ||
| ChAI offers a set of tools to support high- or low-level ML programming, defining/loading existing models, and distributed inference. | ||
| Integrated with PyTorch, ChAI supports developers at all levels, from those looking to load existing models as black boxes, build their own models entirely from scratch, or some combination of the two. |
There was a problem hiding this comment.
This "from" feels dangling in that there's not a "to" following it. Adding "to" is complicated by the fact that thelist includes three things where the third is "both the first", so maybe "whether for users looking to.."? Or there are other ways to fix…
| Second, continuing a theme present in ChapelCon '25, the community is interested in seeing more work on language interoperability. | ||
| Will 2026 be the year where we see Chapel interoperability with Rust, or even C++? | ||
| We'll have to find out! | ||
| Finally, participants were interested in the possibility of using Chapel along quantum computing systems, combining traditional HPC programming models with emerging approaches for programming these boundary-expanding systems. |
There was a problem hiding this comment.
"along" doesn't seem like the right word here?
2 participants
This PR includes a new blog post for the retrospective on ChapelCon '25.