std_throwaway

Official video: https://www.youtube.com/watch?v=0m7bHdcq87o

Translated lyrics: https://lyricstranslate.com/en/mutter-der-mann-mit-dem-koks-...

Important to know:

* Koks: Coke as in coal but also Cocaine as in drug.

* Kohle: Coal but also Money.

The song reflects on the situation after WW2 where people were freezing in the winter and longed for warmth but had no money to pay. This is contrasted with the party generation of the 90s.

Both, they benefit from the regime but don't want their children to grow up under the circumstances they helped to create.

> it manages the system

It manages the system in a way that not everybody needs or even likes. To the people who hate it, it solves problems that other people have but brings new problems that they now have. It's not so much about if it's good or bad but more about whom it was made for.

> I think Apple knows this perfectly.

They research the market very well and optimize for maximum profit. Some people will feel left in the rain and some will think they made a steal. This is of no big concern for apple. Their overall profit is and in the last years it seemed to only go up and up.

An even more stealth method is to make the system work perfectly if everything is running fine. Like correctly reporting net frequency in the range 49-51 Hz but reporting false values when outside of that range. That way there will be no detectable errors until everything starts behaving erratically.

The best way for a society to prepare is to have a few small incidents that put the safety systems to the test on a regular basis.

C++ is perfectly able to interface and interact with C libraries. Also it doesn't necessarily produce larger or slower binaries. Virtually everything is optional. The only hard part about C++ is to identify the good parts (those that are useful for your project) and to not use the other parts. Then you can do "C with templates" or "C with simple classes that have proper RAII" but you don't need exceptions. The argument that I keep hearing is kind of the "slippery slope" that if one allows a little bit of C++ the rest will come creeping and with it will come all complexity. This shows that project management is weak and tries to deal with this weakness by selecting tools with limited capability. A method that seems to work reasonably well in cases like the Linux kernel.

The golden ratio is the most irrational number: https://www.youtube.com/watch?v=CaasbfdJdJg

This is why it creates some distinctive patterns when applied to circular/spiral shapes.

From experience I can tell you that the solution often is very very complex in the beginning. So complex even the researcher doesn't understand it. Publication comes when the solution can be expressed in a way that at least a handful of researchers can understand it. But this is not the end. Going back and looking at it one, three, ten, twenty years later people often find better, easier ways to express it.

Wouldn't modern C++ be better suited than Rust because it's closer to C?

I know that old C++ got some bad reputation in the past but I think it's time to reconsider that with the changes that started with C++11. The only area where Rust really is better is the management of lifetimes and its associated higher memory safety.

Copying is fast if the access pattern is a good fit for the CPU architecture.

If you need to copy only every N-th byte from a AoS it might be as inefficient as random access. So, copying could be expensive.

The article suggests striping your data in blocks but then you end up with the worst of both worlds in terms of program code complexity.

SolarCity is now part of Tesla.

Cache locality is kind of the whole point of it. Some algorithms benefit hugely if you choose a specific layout.

Other algorithms do some kind of random access to a few fields only and they don't benefit at all. Those algorithms can make up 90% of your code but only account for 10% of the computation. Therefore it would be easier to have your data look like a AoS in 90% of your code but actually be stored as a SoA to gain the speed in 90% of the computation.

FTA: "Another thing I might consider is to keep the data stored at AoS, but generate temporary SoA data for processing by some algorithm."

Both strategies work to some extent.

Some topics need a large investment to show anything at all.

Some topics show immediate results (good or bad).

Should we, for example, stop all activities in fusion research because so far nobody has shown that it will work and we already invested billions?

Most scientists work on topics that are quite niche. Most of those topics lead to nothing. A lot of good research took years to ripen enough to be of actual value. A lot of popular topics started out in a niche. Most of mathematics took dozens of years to fully come to fruitition. Can you decide beforehand which one will be the next big thing?

Today, most scientists go for the popular topics and whatever is on the government research plan to get funding.* Whenever the wind changes direction they change their topics because they need that funding.

*: This might seem to contradict with the statement that most scientists work on nice topics. But only on the surface. In order to get published you have to do something novel. So, you choose a popular topic and then research a rather unpopular side aspect on it like how a specific chemical behaves when applied to the popular topic. If you're successful you publish and continue. Citations come later or they don't but the next round of funding comes with publishing. After a few years without many citations you move on to the next thing.

On government plans often you need to publish and then it's done. The citations only matter long-term if at all. Most scientists don't achieve anything of greater value. They are happy if they can publish at all. If the institute has a few scientists with a high citation count it carries all the rest of them.

I assume that we want to access these arrays as "array of structs" for most functions but as "structure of arrays" for some calculation intensive functions. The article suggests storing it as array of structs and to make copies for those calculations but this seems inefficient to me. Modern C++ should provide a way to efficiently decouple the access model from the memory layout.

Can we hide the actual memory layout without big overhead using C++ inline/template functions/classes? Would that be the visitor pattern?

You need a source of trust in these systems. Journals used to have that role. They had high standards that were upheld by editors selecting only worthy publications. Today it seems that many journals aren't as trustworthy as they seemed to be in the past. It's also easier to spam the journals with your publication and to bullshit your way into publication. The incentives to publish a lot are also way higher now that your grant money is highly dependent on your citation count. Journals can publish more and easier and lower the standards for submission to earn more money. The system is basically eating itself and we haven't found a cure yet.

Filtering for self-citations is useful to identify the bubbles. But it is not sufficient to determine if those bubbles only contain hot air or if these scientists are actually working on something with substance in a narrow field where few others publish.

If you work in a very narrow field of science you're basically on your own and have to cite yourself because there's nobody else to cite.

AFAIK those benchmarks weren't specifically optimized for that processor architecture. Most programs are optimized with x86 systems in mind. The results for your specific application might come out differently.

With regards to your dog, you should look into chaotic systems and the butterfly effect. Chaos here doesn't mean that anything is possible or that each outcome is equally likely. It refers to that the outcome changes greatly for even minimal changes of the inputs. Since we don't know everything at one point in time, we can only make limited predictions about a future state. No matter how precise we measure the system it will still behave chaotically.

To some extent this is also true for the economy which contains many feedback loops and meta-stable states. Even if we have the perfect theory it might still behave chaotically and might be as unpredictable as the weather.

Mastery is when you not only know the rules but also know which rules you can ignore in your particular situation.

The emotions of your dog are not laws of nature.

The "true market" is like "no true Scotsman" or "a closed system" or "a spherical cow on a frictionless plane". It assumes something that doesn't exist. In economics the influence of the things it assumes wrongly can be very significant. In physics wrong assumptions can also make the results wrong and a lot of work (or rather most work) is done on checking if all the assumptions are sound and ensuring that the influence of the remaining gaps are negligible. That is something that is hard to do in economics when it involves interactions with the global market.

Yes, special theories can be precursors for more general theories.

Sure, there is randomness but in normal cases you can apply statistics to describe it reasonably well. The results are not 100% sure but 99% or 99.99999% or whatever you need. It's not that the laws of gravity stop working entirely over night. Newtons theory wasn't perfect and Einsteins theory isn't perfect but it describes what happens in nature very very well for billions of years.

Take a working idea, find out why it's working, find out how to make it better, try it out.

It doesn't make sense to only copy a working idea without understanding why it's working because your copy will be slightly worse. While you might have some success you won't take the big share. You must find a weakness in the original idea or the execution or the location or anything else. If you have a solid point where you can improve you have a chance. Don't be afraid of stealing ideas and making them better.

It's hard to test theories when the underlying system constantly changes like it does in economics. Something that works today might not work tomorrow. A theory that looks sound when applied to yesteryears data might stop working before the sun rises. When many others try to apply the same tricks that made you successful the cake on that table might suddenly be divided into so many pieces that everybody will go home hungry. You must find a new table with a new cake but old data and old theory doesn't help much with that because someone else is already there. You need to think fresh and unconventional or you need to squeeze blood out of stones harder than anybody else.

Physicists have it easy in comparison. One of the basic truths of natural science is that the laws of nature are independent of the observer and time and experiments can be repeated as many times as you like with the same results.

Is there a second independent manufacturer for Power processors?

Over copper it's easier to provide asymetric connections because the provider can better manage interference between different cables. When there's no interference problem it's only a question of allocating the bandwidth. Ideally the connection would auto-adjust and provide up to the full bandwidth in both directions instead of constant half/half or 0.9/0.1.

Could you get a dedicated fiber connection for your business? Those things cost a bit more.

> Everything has been designated to the cloud.

Because that's where the money and power lies.

You are a minority 1% user who doesn't really show up in the statistics. Most likely your upload is classified as file sharing and you are probably seen as abusing your connection. There is a reason why uploads are so limited for consumers. You should get a business connection if you really need it for conduction serious business.

If you do online backups you better get the backup plan that will be coming soon from a range of supported backup providers. /s