I don't know if you really need the full ROCm stack just to throw matrices at a card. Inference requires less "scaffolding", but maybe Fujitsu just want to outsource the whole stack and forget about it.
My only worry would be that the software model of compute that ROCm has is going to drastically different than the hardware model of compute the Fujitsu chip has. This would be like using one of those CUDA compilers that target AMD hardware. Sure you can produce something that will run, but depending on what you are compiling, it will run dog slow.
I would reject the comparison as a false dichotomy. The world's political systems can't just be bimodal distribution of ineffectual neolibs and self-styled 'strong man' autocracts.
Factoring hasn't been shown to be NP-complete. It maybe the case that it's in P (and we just don't have an algo yet) or it maybe the case that it is an NP intermediate problem.
It's an approximation algorithm. It's not about the runtime. It's about the worst or average case approximation quality while remaining in a polynomial budget.
The 'general' TSP (unless P=NP) does not admit a worst-case constant-factor approximation, at least not on classical machines (we're still working on the quantum PCP theorem).
So while I'm not immediately familiar with the approximation algorithm you're suggesting (unless it is Christofides[1]), it seems unlikely that it would produce a good approximation for the TSP. It might still perform well in the average-case over random instances. I'll admit I haven't delved deeply into that about the TSP, but I don't believe there are currently any favorable results. There are certainly no Overlap Gap Property related results about the TSP. If I wanted to prove something in the average case about the TSP, I'd definitely be starting there.
[1]: If you were thinking about Christofides: Christofides is indeed a 3/2-approximation, but not for the general TSP. Instead, it approximates a problem like the TSP, known as the Metric TSP. The Metric TSP introduces additional symmetries that make it much, much easier to approximate. As such, Christofides is an approximation algorithm for an approximate version of the problem, which I think is pretty neat. If I'm remembering correctly, the inapproximability results for the Metric TSP are rather favorable.
In grad school, wrote a paper as a homework assignment for a course where reviewed a paper on the idea of using a minimium spanning tree for the traveling salesman problem. The paper did have some math addressing how close the idea was to optimality. That was a long time ago, and now don't have the reference or the homework! From a fast search, now there is
I think its more like trying to find the highest mountain in a range of mountains, except there's too much fog so you can't see how high other mountains are compared to the one you're on, so you can't tell if you're on the highest mountain until you've climbed them all.
A lot of the non-quantum heuristical approaches leave you isolated on one of a few mountains (local maxima), because simply climbing back down and hoping to find another peak is computationally expensive and you know it will lead to a lot of suboptimal solutions before finding something hopefully comparable or even better than your highest peak reached so far.
Mexico is the way it is because the people who live there by and large want to make it the way it is. It's not something inherent to the soil that makes it that way. Mexico can thrive if Mexicans choose to make it thrive.
Every single person who can effect change and publicly opposes the cartel gets killed. Every single one. There is a list [0] of the ones killed in 2024 alone. Even outside of politics, you can't even really joke about the cartel while in Mexico, no matter who you are. They torture and kill entire families over nothing.
I don't want this to be a defense of elon musk. Everything he is currently doing is unequivocally wrong and he does not deserve the benefit of the doubt.
That said, there have been multiple cases where 'unconstitutional' actions had good faith interpretations. The obvious one being the Underground Railroad. Less obvious was Lincoln suspending habius corpus in 1861 (I don't know if the term executive order existed back then, but this sure felt like one). It would be another two years before Habeas Corpus Suspension Act of 1863 was passed.
Overused? Sure, beyond belief. But dropping a "The Constitution is not a suicide pact" feels particularly appropriate right about now.
magically = "group action driven by a shared understanding of the common good. Buttressed by clear through-line that connects what each individual has to give up in terms of time/freedom/autonomy to the benefit of themselves and their community."
It's always fascinating when folks see the stick as the only vehicle for change. But hey, there is a weird honesty to it. The only way we can keep people 'in line' when their lot in life is all lemons ...is naked force.
Misquoting (in translation) the sophomoric military machinations of a person who never served but who's polices managed to kill millions isn't going to curry favor here.
Group action with clear purpose is compatible with capital. In fact, we have a plurality of names for it; businesses, clubs, linux kernel development...
Misquoting my arse. Mao didn't need to hold a gun if he could control those with it. The currently crumbling pax Americana wasn't held by just capital and group action. You think Xi would be fine if the people in his country controlling the guns decided they want someone else?
I didn't just talk about the stick though. When you promote the idea that all cops are bastards, and make sure people who will police the way you want don't go into policing, you'll get bastards. Doesn't matter if your community votes DSA, Green or Communist.
If you don't have people that align with your ideology in areas that matter or aren't incentivised to you won't get the changes you desire.
How will you get change when the people who want the changes refuse to work or have their people in the organizations that they want change? And even disincentivize it.
Barring the use of force it's the functional elites, not the hoi polloi (that manage to have time for protests and such) that drive change.
You can tell from the biography of pretty much every activist or philosopher you know.
As for Linux, it takes people; individuals on the inside to drive change too. As anyone who follows development there will know. If the people that matter don't write or don't accept the code you desire, it's never going to be in-tree.
I am a little confused about what is happening now. What Mao did was wrong. Holding the gun or being in charge of those who do is a distinction without a difference. Law by violence is wrong. Whether it is Xi, Mao, or the mob, it's wrong. The same holds true for America's rise to power. My argument is normative; it is not a historical narrative.
Just because he did it "effectively" (by a measure of effectiveness I would take issue with) doesn't make it less wrong. If you want to stop the lion's share of crime, just imprison every young person aged 16–26 and release them on their 27th birthday. Effective beyond belief, crime will decrease in a way never before seen. It is still really wrong. Effectiveness is not a substitute for morality.
Asking people to join the police, as it is currently constituted, is asking people to go against their morality. It is that simple. It doesn't matter if you believe a police system can exist in its current form "done right", or if you think it needs restructuring, or if you think it doesn't need to exist. Right now it sucks, and a 10% increase in cops with some subtlety in their actions isn't fixing it.
People are in a big dialogue right now about how using social media is ruining people... what do you think going against your basic moral precepts will do to someone? You are not providing a clear avenue for change but still asking for a lot out of folks and wondering why they won't take part.
Regarding Linux development, having some hierarchy is not a negation of group action. Furthermore, analogizing a maintainer being pissy about a Rust PR and the police force killing people might as well be the platonic ideal of a false equivalence.
(The canonical and contemporaneous translation is (I _think_ from Edgar Snow): "Political power grows out of the barrel of a gun.")
To change things you have to hold power, analyzing and critiquing make no change while the internal culture and incentives don't change.
There's no false equivalence. The sort of change you want almost always requires a change of control. i.e people inside .
If you think you're above being inside, it's not any morality nonsense. It's a lack of desire to effect change. Like in Ukraine, the guns won't shoot themselves.
EM is probably just mad that he wasn't rich early enough in the life of the computer to have a dedicated IRS machine for his taxes [1].
In all seriousness, of all the things happening at the moment, this seems the most expected. Under-staffing the IRS was part of the first term [2], and a general move of right leaning executive administrations. Honestly, I would even be against it, just as long as they also made the tax code more straightforward and more bulletproof. But that second step never seems to happen.
From the article it seems like they are considering a 'creator garden' model like patreon.
Ex: you like podcast/streamer/whatever Y. You pay to support Y (via reddit). Y has a paid section of the 'official' subreddit for subscribers. If reddit could figure out a way to provide hosting for Y... or some other kinda value add... it isn't the worst idea? I mean why use patreon, if your fans already have a community on reddit? Plus less friction is always nice for the subscribers. A paying user who logs in can see the 'public' and 'paid' portions of the sub at the same time.
I'm not on either, but supposedly using the patreon app is kinda meh. Private discords seems better, but if your already have a userbase discussing your work on reddit, why not?
I honestly think the biggest issue is the reddit brand. Most people now see reddit as "that one place you get user generated answer from google". Or still see it as "where the insufferable folks are". But that isn't universal. I suppose reddit doesn't have to eat all of patreon's lunch, just some.
(I don't have reddit, patreon, or discord accounts, so...)
My only worry would be that the software model of compute that ROCm has is going to drastically different than the hardware model of compute the Fujitsu chip has. This would be like using one of those CUDA compilers that target AMD hardware. Sure you can produce something that will run, but depending on what you are compiling, it will run dog slow.
reply