Bend: a high-level language that runs on GPUs (via HVM2)

I'm personally putting a LOT of effort to make our claims as accurate and truthful as possible, in every single place. Documentation, website, demos. I spent hours in meetings to make sure everything is correct. Yet, sometimes it feels that no matter how much effort I put, people will just find ways to misinterpret it.

We published the real benchmarks, checked and double checked. And then you complained some benchmarks are not so good. Which we acknowledged, and provided causes, and how we plan to address them. And then you said the benchmarks need more evaluation? How does that make sense in the context of them being underwhelming?

We're not going to compare to Mojo or other languages, specifically because it generates hate.

Our only claim is:

HVM2 is the first version of our Interaction Combinator evaluator that runs with linear speedup on GPUs. Running closures on GPUs required colossal amount of correctness work, and we're reporting this milestone. Moreover, we finally managed to compile a Python-like language to it. That is all that is being claimed, and nothing else. The codegen is still abysmal and single-core performance is bad - that's our next focus. If anything else was claimed, it wasn't us!

I thought you was talking about the DEMO example, which ran ~30% slower than expected. Instead, you were talking about the README, which was actually incorrect. I noticed the error and edited it. I explained the issue in another comment.

We still see this today in how languages go out of their way to implement higher order method libraries (map/reduce/filter) but then under the hood there is no multithreading, they just expect the developer to annotate their loops to be parallel because the languages aren't formal enough to know about side effects in the innermost logic, and don't support immutability or performant pass-by-value semantics with copy-on-write anyway. So we end up with handwavy languages like Rust that put all of that mental load onto the developer for basically no gain, they just save memory by performing computation in-place imperatively.

I also like how Bend sidesteps the nonexistence of highly scaled symmetric multiprocessing CPUs by supporting GPUs. It makes the argument moot that GPUs can't be stopped because they're too big to fail. Julia is the only other language I've seen that tries this. I wish Clojure did, although it's been a long time since I followed it so maybe it has some parallelism?

I would have dearly loved to work on something like Bend, had someone solved the funding issue. Nobody wants to pay for pure research, and nobody sees the need for languages that do what everyone else is doing except easier. We have Kickstarter for widgets and Patreon for influencers, but makers have to bootstrap themselves or learn everything about finance or live in the right city or have a large network to hopefully meet an angel investor or work in academia and lose all rights to what they invent while spending the majority of their time hustling for grants anyway. So it just never happens and we're stuck with the same old busted techniques. Like how Hollywood only has money for sequels and reboots or the recording industry only has money for canned corporate music and hits from already famous artists and yet another cover that yanks the original better song off the radio.

A quarter of a century can go by in the blink of an eye if you get suckered into building other people's dreams as a people-pleaser. Be careful what you work on.

The other link on the front page says:

"Welcome to the Parallel Future of Computation"