Troels Henriksen [email protected]

Do you also not understand parallel functional programming or language-based security? Actually nobody does, which is why researchers try to figure it out! Come and help as a PhD student at #DIKU/#UCPH: jobportal.ku.dk/phd/?show=1609…

@adamnemecek1 The package structure is not important. I think you need to write a main function that invokes the tests in Math.Test.TestAll (if it does not already exist), and then reference the file with that function in the .cabal file, like here: #L494" target="_blank" rel="nofollow noopener">github.com/diku-dk/futhar…

@sigkill_dk I didn’t mean testing in particular, for example I’m trying to fix the package structure of some old package and it’s really painful. What is the proper structure for this package so that I can run cabal test? github.com/Xodarap/Haskel…

Same with Haskell.

a completely unhinged response to a person raising money for children

A nice blog post by @sigkill_dk about `smlfut` which allows SML (especially MLton and MPL) to easily call out to Futhark. I've used it and it works really well! Lots of interesting language interop details. Check it out: futhark-lang.org/blog/2023-10-1…

@manchaary12 @dima_savin @code_report All Remora is missing is combinators, and that's not too hard to add. Although I would argue that no statically typed language is Iversonian.

@dima_savin @code_report my guess is that static types are too cumbersome, and compilation step doesn't bring too much benefits for array languages (my wild guess), because "working on data directly all the time" - you're not really writing functions, nothing to compile away. Also, REPL

I added Uiua to my Array Language Comparison repo. It is officially my second favorite programming language after BQN! Maybe soon to be #1. This language is seriously awesome! github.com/codereport/arr…

@yminsky @shwestrick It is viable for non-hobby uses. The language and the compiler is stable. But it is still a tiny niche language with all that this implies; e.g. very austere tooling.

@sigkill_dk @shwestrick Would you recommend it for non-hobby uses? Do you think it's too unstable to use for real applications?

So, other than Rust, are there practical languages that have data race freedom guarantees? Haskell? Scala? Is Haskell's memory model documented somewhere?

@yminsky @shwestrick I don't know of any use in industrial production applications, though some users have mentioned that from time to time. All non-hobby uses I see are prototypes or research, where future maintenance (or existing code bases) is not a concern.

@shwestrick Do you know where futhark is being used for real production applications? I'd love to know more...

@rtfeldman @sw_unscripted Sure, sounds fun!

@sigkill_dk Hi! I don't know if you have any interest in being on podcasts, but I host @sw_unscripted and we talk about a lot of different programming languages. We haven't done an episode on Futhark yet, but I'd love to chat with you about it sometime if you're ever interested!

Join us at the Functional HP and Numerical Computing Workshop icfp23.sigplan.org/home/FHPNC-2023 at @icfp_conference. It's just a month away, with @satnam6502 and @sperbsen as invited speakers — early registration closes tomorrow!

PyTorch and Dex with Adam Paszke. A view from the inside. Can the Array languages learn anything from the ascendency of PyTorch and Dex? Host: Conor Hoekstra Guest: Adam Paszke Panel: Marshall Lochbaum, Stephen Taylor and Bob Therriault. arraycast.com/episodes/episo…

Most popular array programming languages 🥳 plrank.com

@blelbach @tristanbrindle @cpponsea Generalisation: any subsequence of statements in your program is an ad-hoc unspecified named subroutine.

Tristan's Law: Every for loop in your program is an ad-hoc unspecified named algorithm. @tristanbrindle @cpponsea

@blelbach Spicy holes.

This is a great example of an interface footgun.

@shwestrick Also, I remember Guy built a whole language that could do irregular parallelism on computers even more limited than GPUs!

@shwestrick We should write algorithms that are agnostic to that! Or better, have a cost model that allows reasoning about the cost of irregularity, even on hardware where it is well supported.

Guy giving his award keynote at SPAA 2023. Packed room!

@hannelita @PLDI You should put your affiliation as "the PL community" or something.

@hannelita While I know you best from being a highly helpful volunteer at many PL conferences over the years, I also remember you being fun at taking PL stuff at PLTea.

@shwestrick Why only multicore?

Are parallel algorithms ready for prime time? Guy: yes, for multicore! So let’s get out and teach multicore parallelism. CMU starts early (~3rd semester undergrad), and it works.

@haskellhutt @RanjitJhala @nilehmann @atom_geller @nikivazou @PLDI One shows that it ought to work, the other that it seems to work.

@RanjitJhala @nilehmann @atom_geller @nikivazou @PLDI I scrolled through looking for the performance table that apparently has to be in every PLDI paper, just as every POPL paper has to have typing rules. And, yup, there it is :-) Nice work!

I've enjoyed working with @nilehmann @atom_geller and @nikivazou on this new paper on #Flux to appear @PLDI next week; maybe you'll enjoy reading it too! ranjitjhala.github.io/static/flux-pl…