Show newer

@js @objfw

Something like that, but I don't get why you need the first shufps,

@js @objfw

And doing it via memory (well, l1 cache really) would be even slower?

@js @objfw

Is vectorizing point-wise multiplications, and doing horizontal additions in a nonvectorized fashion (still using sse1) still obviously worse than nonvectorized everything (or impossible, because you can't pull out single elements of vectors in the way you'd need to)?

@js @objfw

What hardware are you comparing on? I would expect the advantage on hardware that supported only sse1 to be greater than on anything more modern.

@lauren As a good rule of thumb, a random idea that wants to make social/people-affecting things more uniform that I can come up with is often one that has been tried either by the First French Republic or the Soviet Union.

@sophieschmieg @saraislet

Ah, so this is why the US has so many drivers -- after all not everyone wants to text /s

@grrrr_shark
a "fun" thing is that, in face of participants who have blocked you/your instance, this is not even in principle possible (because inReplyTo is a singular field and there's no equivalent of email References header).

@8petros

I assume you meant to say that those who want it more will pay more for it (otherwise, well, yes? people who don't want a good will not pay for it by definition, ones who do want it will pay something, so they will pay more).

What I find hard to reason about in this area is comparing how much _different people_ want something. I can tell how (in a world of perfectly honest and omniscient people who will answer all questions) to tell whether one person wants X more than Y ("what would you do if you could pick one of them only?"). Do you have a candidate definition (or something that points in the direction of one) for comparisons of wants between different people?

@sophieschmieg @filippo @neilmadden @dangoodin @ryanc

For some stupid reason I thought that you can do a few "steps" of Grover in parallel (as in starting from not copies, but entangled copies of the previous step). This is not trivially possible though, and the paper @filippo pointed out claims to prove that it's impossible (as in, that oracle search with a quantum oracle has to be at least sqrt(search space size) oracle invocations deep).

@filippo @sophieschmieg @neilmadden @dangoodin @ryanc

Huh, I don't know how I got the notion that you can trivially parallelize it if you pass qubits between the computers; this paper seems to prove _that_ impossible. Thanks.

@filippo @sophieschmieg @neilmadden @dangoodin @ryanc

Do you have a pointer to the proof (or a way of searching for it: Grover does trivially parallelize if you can transfer qubits between the various parallel computers, and I don't know how the setup of multiple quantum computers with only classical links between them is called)?

@sophieschmieg @neilmadden @dangoodin @ryanc @filippo

It shows that quantum computers can't magically find a value accepted by an _oracle_, but doesn't say anything about finding a value accepted by a program, and so also about BQP vs. BPP, NP or coNP though (even if many (most?) people think it's very likely that BQP is not larger than NP).

@smitten

Yeah, I wanted to suggest that if we consider water to be arbitrary, then we should consider standard pressure even more so.

I was surprised that hydrogen's triple point is at less than 0.1atm (engineeringtoolbox.com/docs/do)

@_dm

I meant legal paranoia as in thinking that the scope of searches that can be legally executed in various cases would be limited by not having the "more important" phone on you.

Why do people require signed fetch, but allow the same posts to be viewed via a web browser?

@_dm

Isn't the two phones thing mainly because of legal paranoia (I have no opinion on whether it makes any sense in that context)?

I don't think it's bad if people have wrong opinions, as long as they know what would it take to change their mind and are open about that (that's slightly stronger than "can explain why they think so"): often the choice is between having some opinions that might be wrong, or doing the default thing which might also be wrong. If one adheres to this rule, they are very likely to be convinced out of the wrong opinions.

(If someone publicizes their opinions using any sort of authority, then I agree this is weird, because it's actively harming society's knowledge.)

@adfichter

Handelt es sich um Fragen über etwas, dass Öffentlichkeitsprinzip unterliegen soll, oder etwas anderes?

Show older
Qoto Mastodon

QOTO: Question Others to Teach Ourselves
An inclusive, Academic Freedom, instance
All cultures welcome.
Hate speech and harassment strictly forbidden.