(and ideally, at a point where there is cmake support for the new name so it “just works”)
For context, I work for Arm and Linaro but not directly on flang. I have watched from the sidelines as my colleagues worked on, and continue to work on, flang.
So I would be happy to either write such a post myself and get review from the community, or to have someone closer to the project volunteer to do so instead. I just don’t want this (in my opinion) significant moment to go to waste.
I think this is a significant event because:
Flang is an llvm as a compiler framework success story.
It’s an out-of-tree llvm project success story.
It’s a project that successfully migrated into the monorepo, and has gained a lot of interest doing so.
It’s an MLIR success story.
It has many stakeholders, especially since the move into the monorepo.
It’s a great opportunity to celebrate all your hard work, promote the LLVM ecosystem and gain new users all at the same time.
A rough outline I was considering:
flang-new is now flang! go get it , or try it on Compiler Explorer (for the readers with little time or coming in from google later)
what fortran is used for (very briefly, for the “isn’t fortran dead” crowd )
where flang started (it was downstream for a while? I would need to research this)
going open source out of tree
the adoption/use of mlir in flang (briefly, maybe just show the “stack” of IRs)
the move to in tree
where flang is now that it’s “ready”, some significant milestone codes, any features flang gets “for free” for being llvm based e.g. sanitizers?
some quotes from contributors about their favorite work / what they hope to work on next
how to try flang in your existing projects, how to report bugs, contribute, etc.
Any reactions to that idea?
I will also be joining the community call this week if anyone would like to discuss the idea there.
Again, I’m not a “core contributor” by any means but I have written well received articles in the past (Tools for Learning LLVM TableGen - The LLVM Project Blog). So I am happy to draft the article and have you all review it, or step back and let someone else run with the idea.
[ongoing] we’re planning to use flang 19 as the default Fortran compiler on windows in our cross-platform distribution (I can report back on that in 1-2 months)
[ongoing] a second one that is not yet ready for sharing, but hopefully soon as well.
I joined the call today and at least from those in attendance, there were no objections to me starting to draft this post. Of course nothing is set in stone so I’m happy to hear from anyone who’s also interested in working on it.
There were concerns about the increase in, and type of, feedback that this publicity might produce from users. This overlaps with the renaming discussion, and I will leave that for those more expert than me to debate.
In the meantime I’ll start drafting something and doing the research on the project’s history. I may contact a few of you to fact check that. Whatever I come up with will sit in draft form as long as it needs to, I’m in no hurry.
Definitely. I do remember reading the SciPy post.
@omjavaid Do you have any highlights from Linaro’s work so far?
Thank you for offering to write this - I think that this is a great idea! I would, however, make sure to co-ordinate with [PROPOSAL] Rename `flang-new` to `flang`. That proposal seems to be gaining traction, so things are indeed very promising!
I am happy to either review or even contribute myself. From my perspective, as a “Flang driver” code-owner, I’d like to call out successful adoption of clangDriver (Clang’s compiler driver lib) to implement another upstream compiler driver
I have some material on the application successes of Flang as well as its ability to detect long-standing non-conformance issues that other compilers ignored. Please write me an email (it is easy to find) if you want details.
I saw that you asked to keep it short, which I’ll probably fail to achieve to your satisfaction (in the first iteration at least; we can keep trimming!).
By early next year we should also have moved conda-forge to use flang 19 as our default Fortran compiler on windows, though I’ve left that part out for now for brefity (and because it’s not done yet).
PS. At the time that I wrote that blog, I would have liked to have been aware of Resurrecting Fortran – I think it’s good meta-context for the overall resurgence of Fortran, which is a big part of the backdrop behind flang’s rise IMO.
PPS. Support for LLVM-flang has now been released in meson 1.6.0, so perhaps we might even be able to add building SciPy as a test in the flang CI until the release (an idea floated by @kiranchandramohan a while back).
I’d put 4 in “a few” still so this will likely be fine.
I will wait until I have some reasonable amount of quotes before I add any to the article. At that point I’ll decide if and how to edit (with the author’s permission) or categorise them.
Also as the author of the rename proposal (at least, the final version of it) would be good to get a quote from you (@everythingfunctional) about that.
I’m happy to schedule a call to chat about it, or if you’ve got a draft you’d like to send my way I’d be happy to collaborate on the article. Just let me know.
I told some of you over email that it was 3 weeks from now, that was the branching date not the 1.0 date, sorry for that!
I would like to get the article stable before the release date, so now is the time to review if you intend to, and think about whether you/your employer/your community would like to contribute a quote to the article. I will be adding quotes in the coming days, so you will have some examples for inspiration
@h-vetinari I have added this in the Reflections on Flang section of the draft. I did some edits for brevity, let me know if they are ok and if you’re ok with how I have credited you.
My plan is to get this ready to publish by the 11th of March, to go up soon after the llvm 20 releases are available. So you have about 2 weeks left.
Quotes I can easily add. Review comments suggesting significant changes will be harder to address as the deadline approaches. So please do that sooner than later if you had plans to.
)
