shash

The 30% spending is from government numbers in the last budget, so let’s get that out of the way.

How many people do you think could conceivably work in manufacturing? And who buys all those manufactured goods? The social problem isn’t solved by quotas.

Actually, legally, caste doesn’t exist. It’s been abolished, except for schemes for socially eradicating it (which are controversial but conceptually no different from any other affirmative action).

Not to say it isn’t a problem. But race is still a problem elsewhere too.

Thirty? Try more like 10, and that’s only for the advanced nodes. Everything else, we will have capacity coming online soon enough.

Design, the gap is nonexistent in terms of talent deployed (though we have some way to go to have homegrown product companies). Which puts us roughly where Europe is.

Also consider, a tiny percentage of the Indian population is a ridiculous number of people. Dwarfing pretty much any other country except China.

Your last point on welfare and corruption is also false; 35% of government spend is on interest payments (which is kind of normal for most countries?) and 30% is on infrastructure and other capex. Defence is 8% and central welfare schemes plus co-sponsored schemes with the states are only 26%. The point being, while corruption is definitely a problem, it doesn’t significantly come out of the central budget. And welfare is smaller than infrastructure spending even if it’s quite significant (and it better be, in a country like this).

It’s not like GC isn’t a massive part of RVA23 - those are the basic instructions that handle 80-90% of all uses (including most of what CPython needs).

I’m sure one could do some optimizations on RVA23, but is it really worth it?

We ran it 4 years ago (including numpy) and it wasn’t a massive uphill climb or anything. It’s gotten a little easier since then if anything.

For RISC-V that would functionally be RV64GC then. And you go incrementally from there as required.

Again, depends… I have appreciated my seniors’ notes in some textbooks - especially where the book itself isn’t clear.

Learning and figuring out the meter and rhythm of a new passage? If you don’t have someone who knows to chant it for you…

Besides, it’s a fun exercise, so why not?

Wealth is not finite, but it does have to obey physical limits and real world implications. If person A takes a piece of lumber and carves it into a dresser, and person B, seeing that, buys up lower quality lumber that cannot become a dresser, is that wealth generation? What then, of the guy who buys up lumber futures and dresser futures, on something that won't actually ever happen (say, those lumber futures are for trees that are completely not viable) and then claims to be creating wealth? And asking the stock market to value him on the basis of hype?

There's real wealth and there is all of the manipulation to drive up prices.

Of course. But,

1. So are most other encyclopaedias, and most don’t even give you references or citations for an article 2. We are literally decades from that point now. Surely you must agree that times have changed.

TBH, "Don't trust anything on Wikipedia" was always just stuck up posturing. Plenty of badly sourced, badly edited and biased articles in regular encyclopedias too, but because Wikipedia is contributed to by the unwashed masses, don't trust it.

"Don't use Wiki as a primary source" is much more to the point - it's not meant to be. Use it to find the primaries.

Also, I don't see this as "reducing breadth" so much as attempting to hold to a minimum standard of "has this been written up anywhere by someone other than the people who started it?" which isn't perfect, but it's been the stance since the beginning.

These were big in India until the mid-90s. I recall one book of "tales from the Baltic states" which we had at home. There were others, but I can't remember the titles very well. And for stuff coming from the Soviet era, they were remarkably non-judgy about the Tsareivichs and Tsarinas who inhabited the tales...

On the technical front, one book that I fondly recall, but I haven't seen since is Experiments Without Explosion by O.M.Olgin: https://archive.org/details/ExperimentsWithoutExplosions The title, as well as the content...

But all of those things are an agreement with consideration; a "license" is a permission to use some property. If not a patent, what is the "thing" that the farmer is licensing from the company.

It's fine to say "there was an agreement that the farmer would grant exclusive rights to the produce of the trees, in exchange for seeds, rootstalk or cash", but that's not a license. I don't understand that aspect of this at all. The articles consistently talk about a "license", not other consideration such as the aforementioned seeds or a guaranteed purchase.

The only "property" that could conceivably be licensed in such a manner is a patent - other forms of IP such as copyright or trademark wouldn't enter into it. And any other property doesn't make sense because... well... the only property here is the farm and the trees, which are the undisputed property of the farmer.

It just feels like a very loose usage of the word "license".

What I don’t understand is, _how_ it doesn’t depend on a patent right. How do you license something that doesn’t depend on a licensable right?

If it isn’t a patent right, exactly _what_ is he purchasing a license to? If it’s purely an agreement of exclusivity to Giumarra, it can’t be called a “sub license” and the consequence is just a breach of contract (which is what it sounds like). But in that case I’m sure there are mutual termination agreements in the exclusivity contract.

Also, if it’s _not_ a patent, what exactly does the company bring to the table? For what consideration does the farmer give them an exclusive access to his trees?

Comparing nodes across foundries is a kind of a coin toss. At least, from the name. You actually need to go into the specifics of the pdk and process to understand what features there are. Can’t rely on the name for anything.

Correction: it’s not stacking. They do things like FinFET (turning the gates in the third dimension) and gate-all-around which increases the density of transistors per unit area. But they don’t have layers of transistors. At least in the logic and analog processes.

It’s not merely a “gaming vs data center“. There’s so many other places DRAM and NVM are needed - mobile, automotive, other consumer electronics,… the current situation is that _all_ of that is deprived of the memory that it needs. And much of this is critical to the real economy.

Yes. It provides a prior for Bayesian analysis if nothing else.

Yes, actually. This is similar to having a 100 year flood five years in a row. It doesn’t mean that the flood occurs only once in 100 years. _On average_ it’s 1/100 probability of occurring in any given year.

But then, Apollo 1 was after all the first mission on the Saturn V. I think we should assess even its pre-launch risk much higher than the rest of them. Similarly Artemis II has a much higher risk than the subsequent ones will have.

Of course they didn’t. The delta-v needed to land the rockets is better expended in pushing the craft further. Reusable rockets isn’t always the best choice.

Is 12 enough of a sample size to make a statistical judgement? What if there were 20 more which didn’t have a loss of life? Is it then 1/30? What if there were 20 more?

The risk factor is calculated _per mission_ from what I understand. You can have three accidents in a row and nothing for decades but the risk itself can still be 1 in 30.

How are you/they instructing those agents? If you are writing detailed spec.md and reviewing those results, you are _still_ programming. Just in pseudocode effectively. I’ve seen enough session transcripts and detailed prompts that would have been easier to just write the code instead!

Remember that the defender has home team advantage. That’s precisely what you see happening both in Iran and Ukraine. That advantage exists with Taiwan. There’s a reason that China hasn’t made a move in all these years, and the US is only one part of that equation.

That analysis requires discovering what the US’s objectives were. Not sure we can…

The key here is “could be”. But most four (or in my case, six) year olds can’t really grasp the abstract concepts of what JWST is or the data it’s sending back. For that matter most 40 year olds can’t.

A manned mission on the other hand is tangible in a way a probe isn’t. “See the big round thing in the night sky? There are four people going around it in a spacecraft”.

It isn’t a _complete_ argument in favour of manned missions- that has to account for the risk of the endeavour and reward of the science potential of having people there to react in ways robots can’t. But it’s hard to pretend that the inspiration pretty much everyone feels when they see manned missions is somehow achievable purely by robotic ones.

They’ve changed it so III isn’t landing. That will be IV apparently.

Hi, CEO of Mindgrove here!

Good to see that we're being noticed. But we all still need to deploy (in scale) to make anything worthwhile.

> The name, yes, but going by name is a bad idea as the V in AVX also stands for Vector.

Now I get your point after reading more of the linked page. Yes. It is very implementation specific.

One of the things about RVV (and in general any vector ISA) is that the data path can be different enough between different implementations such that specific rules of thumb for hand tuning most probably won’t carry over. As you say it is true of even sufficiently advanced SIMD architectures like AVX.

As sibling said, stripped down in the sense it doesn’t have dedicated registers. In terms of supported functions it’s somewhere close to MMX.

I don’t personally like it because it still ends up with all the headache of building most of a vector subsystem (data path, functional units,…) while _only_ pretty much reducing one special vector file.

It _is_ RISC-V Vector extensions, so a very specific ISA in mind at the very least. There's another extension (not ratified I think) called Packed SIMD for RISC-V, but this isn't about that.