DiffEqGPU.jl VS hn-search

Indeed, and this year we created a system for compiling ODE code not just optimized CUDA kernels but also OneAPI kernels, AMD GPU kernels, and Metal. Peer reviewed version is here (https://www.sciencedirect.com/science/article/abs/pii/S00457...), open access is here (https://arxiv.org/abs/2304.06835), and the open source code is at https://github.com/SciML/DiffEqGPU.jl. The key that the paper describes is that in this case kernel generation is about 20x-100x faster than PyTorch and Jax (see the Jax compilation in multiple ways in this notebook https://colab.research.google.com/drive/1d7G-O5JX31lHbg7jTzz..., extra overhead though from calling Julia from Python but still shows a 10x).

The point really is that while deep learning libraries are amazing, at the end of the day they are DSL and really pull towards one specific way of computing and parallelization. It turns out that way of parallelizing is good for deep learning, but not for all things you may want to accelerate. Sometimes (i.e. cases that aren't dominated by large linear algebra) building problem-specific kernels is a major win, and it's over-extrapolating to see ML frameworks do well with GPUs and think that's the only thing that's required. There are many ways to parallelize a code, ML libraries hardcode a very specific way, and it's good for what they are used for but not every problem that can arise.

Link to GitHub repo from the abstract: https://github.com/SciML/DiffEqGPU.jl

This is a very popular article that get submitted every now and then (read nearly every year).

Past:

https://hn.algolia.com/?query=Computational%20Complexity%20o...

Interestingly, [X has died] seems to be among some of the topmost upvoted posts of HN. (Based on https://hn.algolia.com/)

This HN post is seven days old but presented as if it was posted 9 hours ago. The déjà vu effect is disconcerting and an absolute mind fuck. Please stop doing this, ffs. The person who thought this would be a good idea is a madman.

https://hn.algolia.com/?query=Silicon%20Valley%20crosswalk%2...

I really like the Barcelona Superblocks model [1], but paying people to move closer to schools is also an option imho.

[1] https://hn.algolia.com/?q=barcelona+superblocks

Political interest and intellectual interest are not the same thing. HN is for the latter.

There is overlap, of course [1], but there is also a huge amount of political and social material which is not primarily about intellectual curiosity. Most of that is off topic on HN (as the site guidelines say), even though much of it is far more important—as you say—than nearly anything else on HN.

[1] https://hn.algolia.com/?dateRange=all&page=0&prefix=false&so...

SomaFM NEEDS to put up a stream of these atop ambient electronica, please!

https://hn.algolia.com/?dateRange=all&page=0&prefix=false&qu...

Any particular reason we want to revisit this (2018) now?

https://hn.algolia.com/?q=jslinux

Might be some thread merging, dang sometimes does it, or might be a newish automatic feature I haven't heard about.

If you search the story name it says it was posted 3 days ago:

https://hn.algolia.com/?dateRange=all&page=0&prefix=false&qu...

There's a large number of very well-capitalized startups in China trying to build reusable rockets, in competition to SpaceX. It's a significant national security priority for them. I'm pretty sure one of them will succeed soon. I don't understand your argument why they will all fail—I don't see where you articulated such an argument, really, beyond "many people have failed, so it must be impossible or someting".

But SpaceX itself is proof it's possible to do what SpaceX did. And: the market in China is wide-open for the taking, since, SpaceX obviously isn't selling there.

Chinese newspace seems to be at inflection point right now. I understand there's about five different private-sector teams that will attempt orbital launch and booster recovery, in either this year or 2026. That's: Deep Blue (Nebula); Landspace (Zhuque-3); Space Pioneer (Tianlong-3); iSpace (Hyperbola-3); Galactic (Pallas-1). Four of those five startups have already gone to orbit—Landscape even beat SpaceX (and Blue Origin) to the first orbital methane/oxygen rocket. (Not that you'd hear about that on HN[0]. China's space startups are a blindspot for us, and, I think we're going to end up blindsided by them, in exactly the same way we were blindsided by Huawei, by DeepSeek, by BYD. We're just ignore all the fast-moving developments there up to the point they're impossible to ignore...)

[0] https://hn.algolia.com/?q=zhuque-2

woww. wonders will never cease.

>This is not appropriate language to use on HN

if, by "not appropriate language", you meant my (mild) use of the word "fuck", first of all, who the hell are you to decide what is appropriate language? you are a nobody, just like me. so don't give yourself airs. you don't get to decide what is right around here. the mods do. and the users also do. if you don't like how things are done around here, you can vote with your feet, and buzz off. I do that myself, sometimes, on any forum where the general attitude displeases me. it's my right, and yours too.

and, maybe you grew up in a nunnery, and should head back there, pronto?

because, you know, that word "fuck" is commonly accepted as a mild expletive nowadays, and is widely and freely used by all sorts of people, in all sorts of situations, both formal and informal, including by young girls (!), from the last 20 to 50 years or more, depending on the country. so it is well established and accepted. you might be in the minority of people who think that it is inappropriate. the word "fuck" is used heavily, even on a chi-chi site like hn.

check this as proof:

https://hn.algolia.com/?dateRange=all&page=0&prefix=false&qu...

there are over a hundred pages of comments using that word, at that above link!

change the top left dropdown from Comments to Stories, and you will see more of the same.

>and I'm not your "bro".

and the same as what I said above, about the word "fuck", is the case for my use of the word "bro". so I am not going to belabor the point any further, sis.