boulos

I appreciated the Sherlock Holmes joke:

> $ regex-cli find match pikevm --no-table -p '\b\w+\b' -y 'Σέρλοκ Χολμς'

0:0:12:Σέρλοκ

0:13:23:Χολμς

It's roughly the top 1% globally: https://www.credit-suisse.com/media/assets/corporate/docs/ab...

If you don't care for the global measure, you can focus on your country or region. But for a global measure, like for this map, it's relevant.

It's been over a decade since I looked at this, but in real-time Monte Carlo path tracing the speed of your RNG is material. In simple enough scenes, probably something like 5-10% IIRC when using MT.

I appreciate that you had a note about performance, but you were pretty quick to dismiss that there's a large quality difference between rand() and any modern PRNG. Plenty of these are fine enough and paying for a CSPRNG with 4x or more the perf hit may be a mistake.

I do agree that if you don't know what you're going to use them for or don't know if you have a performance problem that you should just use a CSPRNG. The potential for someone to accidentally cause a security bug is not worth it. If they decide to do so intentionally, that's on them.

tl;dr: maybe "Need a PRNG? Start with one that's cryptographically secure"?

Presumably you terrified them when you came to GCP :).

As folks are saying, there are already commercially developed panels at 24.7% [1] that should ship within the next year. The existing fleet of Maxeon 3 are apparently going to be 24% this quarter, rather than the 22.7% in the actively shipping parts [2].

Like cache miss ratios, it's hard to remember that 25% is a 10% improvement in efficiency over that 22.7% (and 4% over the 24% claim). But as jacquesm mentions, what matters for system design is either $/watt (installed!), total generation, or whatever other metric. Efficiency might get you more panels on a roof, but so could different panel sizing! One more panel because the sizing works out could easily be 5%.

tl;dr: it's interesting research, but shouldn't affect purchasing decisions.

[1] https://mediaroom.maxeon.com/2023-06-01-Maxeon-Solar-Technol...

[2] https://maxeon.com/us/sites/default/files/2023-05/sp_mst_112...

https://en.m.wikipedia.org/wiki/Sky_burial For those that don't know the term

I think anyone can get on the Top 500 as long as they run a verified FP64 LINPACK result. At the time, this would have been comparing against Perlmutter:

https://www.top500.org/lists/top500/2021/06/

So 6k A100s at just under 10 TFlops of FP64 => 60 PFlops. I don't remember if they said what the interconnect was, as that can hold you back from achieving those 60-ish PFlops.

It's too bad NVIDIA's blog referred to the FP16 value as "flops" in the same breath, as most people at NVIDIA know that Top500 focuses on achieved FP64 LINPACK results. I swear there was some talk of using MLperf or similar to make a Top500-for-ML/reduced precision, but can't find it now.

The Architecture of Open Source Applications has some great lessens learned. It's also free: https://aosabook.org/en/ .

I don't think there are lots of examples of the distributed systems discussion though, since most of the writing is about how to structure the source code of a single program. That's not unrelated, since most of what you're asking about is how different interfaces or modules should interact. In a distributed system, that happens over RPCs, but it can just as easily happen within a program with separate threads.

Note: they've added some other books, which show up at the top of the page, but I've only read select chapters from AOSA Volume 1.

Aww, I just spent time trying to reply in good faith :).

You phrased this poorly, but it's a worthwhile discussion to have.

People recall Trump as having said something about "forest management" as contributing to the California wildfires. That part is strictly true! The actual wording of his claims though, weren't for prescribed burns but rather [1]:

> "I said, you gotta clean your floors, you gotta clean your forests — there are many, many years of leaves and broken trees and they’re like, like, so flammable, you touch them and it goes up."

That quote could be read in good faith to include prescribed burns, because it's not explicitly saying how to "clean" the forests. The article goes on though:

> Trump’s suggestions have prompted head-scratching from experts who say his prescriptions — more raking, less water released into the ocean for environmental purposes — suggest he does not understand the science of wildfires. Critics also point out that most of California’s wildlands are federally managed.

And re: the Paradise fires:

> Back then, he pointed to Finland, claiming its leader said the European “forest nation” had “spent a lot of time on raking and cleaning and doing things. And they don’t have any problem.”.

> Finnish President Sauli Niinistö later said he told Trump that Finland takes care of its forests but did not say anything about raking.

So he wasn't wrong that forest management matters. But he wasn't supporting controlled burns either. The combination of his control over federal forest land and this quote:

> “Maybe we’re just going to have to make them pay for it because they don’t listen to us,” he added.

is what rankled people.

Finally, while forest management contributed to the California wildfires of the last 5 years, PG&E's mismanagement of power lines and power line maintenance played a large role as well. Either way, having people rake the leaves wouldn't have done much :).

[1] https://www.politico.com/states/california/story/2020/08/20/...

https://www.wiltshiremuseum.org.uk/?exhibition=wiltshire-tha... which is the exhibition seems to have it, too.

Touché! I didn't even notice :).

Hmm, the BBC coverage is actually more informative: https://www.bbc.com/news/uk-england-wiltshire-67336495 but I can't edit the submission URL.

I felt the first season primarily mimicked The Soul of a New Machine. Taken from that angle, I found it quite good.

Amusingly, when I read that book, I mistook it for fiction. Kidder's narrative non-fiction style was so detailed I assumed it couldn't have been from notes or interviews.

Long ago, in https://graphics.stanford.edu/~boulos/papers/brdftog.pdf we tried to trade off energy conservation and data reproduction at the cost of reciprocity.

In the end, nobody actually cared about automatically fitting material data from BRDFs or if they do, they actually prefer a symmetric and conserving-enough BSDF, which is what most future research improved. In particular, breaking symmetry means making bidirectional path tracing and other techniques kinda weird.

> Independents?

Good one :).

The author even mentioned him in the notes. Do you know any of the other designers names?

As Bryan mentions in https://news.ycombinator.com/item?id=37880669, I would read Matt's post-Intel and post-Google write up of why Intel didn't really invest.

https://pharr.org/matt/blog/2018/04/30/ispc-all if you don't want to follow the comment link again :).

Sure, but you should still keep the concepts disjoint. "I took whatever I could get" isn't the same as "spot instances are inconsistent".

Optical switching is about training at scale and replaces classical physical networking systems (people will say things like "electrical", "copper", etc.). Running models at edge devices is almost by definition without networking.

> but they obviously provide inconsistent performance

No. Unless you're using a "burstable" (overcommitted) instance family, this shouldn't be the case. The performance will be consistent, but you may get preempted before you finish your work. I would keep those concepts separate.

Objects have three fields for this: Version, Generation, and Metageneration. There's also a checksum. You can be sure that you were the writer / winner by checking these.

For extremely rare emergency comms?

I had (have?) a side project to build a self-hostable Mint/PersonalCapital/whatever on top of ofxtools. The basic idea would be full control of your credentials, so you don't have to trust a third-party like Plaid or Yodlee. Self-hosted doesn't necessarily mean not supporting SaaS-ish behavior though, it's fairly easy to do "dump all my transactions to BigQuery in my own project".

I feel like self-hostable asset and expense tracking is something a few people would care about. I believe tech forward folks are faced with a choice between:

- Command-line tools like gnucash or ofxtools that are pretty bare bones

- Spreadsheets that are manually updated

- Give in, and hope that these companies don't have data breaches

Those options kinda suck :). Does anyone want this? I made it part of the way, until I got infuriated by the sad state of "open banking" in Europe (a friend was interested, but is in Europe, where OFX isn't popular).

> Protobuf is another library to handle cross-language serialization, but wouldn’t offer us the flexibility to add all the other features we now have — plus, it was a binary serialization when I last used it.

Protobuf has a well-defined text format, which is the top search result for "protobuf text" [1]. Proto3 also has a clearly defined set of rules for supporting JSON.

[1] https://protobuf.dev/reference/protobuf/textformat-spec/

Yes, at 1% margin on those sales, that's more like $125M in revenue. It's important to remember that things like Prime Day are basically marketing that results in revenue outside the event.

Try Mama Ji's, Mister Jiu's (upscale) or Mamahuhu (casual). The mapo tofu at Mamahuhu is great, but be prepared for the spice.

Edit: looks like the author included Mister Jiu's :).

Yep. This was going to be the next phase: there isn't anything special about the SI system, but it is consistent. The poster didn't seem prepared for "or any other velocity units you want".

For the original poster, you could equally use c in miles per hour, mass in "pounds", and get whatever that produces for energy (I'm sure there's a name for it).

This is also why Imperial units produce such amusing things as Foot-Pounds (for torque). The math all works out in the end, but you get some amusing numbers along the way.

Do you have a more specific example? Xiaomi's site, Mi.com, for instance doesn't seem to cover every EU or European country (e.g., Bulgaria, Croatia, Norway, Switzerland, etc.). Are you referring to OPPO or OnePlus as a "cheap no name Chinese phone"? They account for about 10% of the global sales (as does Xiaomi).

Assuming this is in good faith, the units are:

- c: speed of light in meters per second

- m: mass in kilograms

- energy: joules