Kazushige Goto, a patent clerk who became one of the best assembly optimisers | Hacker News Reader

Kazushige Goto, a patent clerk who became one of the best assembly optimisers

twitter.com

83 pointsksec21 comments

Is this a case of nominative determinism [1]?

Haven't read his code, but direct jumps (gotos) are quite common in assembly and can certainly help optimize things some times.

[1]: https://en.wikipedia.org/wiki/Nominative_determinism

Not a very good one. The term "goto" is not used in Assembly, usually "jump" or "branch" is used instead. And "Goto" is just a very common Japanese name, and a somewhat mangled spelling in the first place.

In Japanese itself, the connection there is entirely nonexistent.

The GotoBLAS paper (“Anatomy of High-Performance Matrix Multiplication”, https://www.cs.utexas.edu/users/flame/pubs/GotoTOMS_final.pd...) is really a masterpiece.

It’s worth internalizing almost every single detail if you’re an engineer interested in writing high performance numerical codes on modern hardware.

For me, this is a little more... readable story:

https://www.nytimes.com/2005/11/28/technology/writing-the-fa...

Since I know some Japanese, I'd like to point out that "Goto" should be pronounced more like "Go-toe"

Dumb question: in Cryptonomicon, there's a character named "Goto Dengo". Is Goto used as both a first and last name, or is the notion of first and last name something that doesn't map cleanly between English and Japanese?

Japanese family names go first. e.g., GRAHAM Paul.

Respectfully, that's not such a good example, as both of those names can be first or last names. DOE John would be a better example.

The example is referring to a certain famous Graham Paul, for which people here are assumed to know what's his actual first and last name.

In Japanese, you write family name first. Sometimes when using a Japanese name in English, you flip them to the English order, and sometimes not.

What is it with theses patent clerks [1]?

[1] https://www.ige.ch/en/about-us/the-history-of-the-ipi/einste...

Former patent examiner here. The job of a patent examiner can be quite stressful, and I don't think it's particularly amenable to making breakthroughs. At the end of the day I rarely felt like working on personal projects as I was too drained. Einstein and Goto might have had a lot more energy than I do. Alternatively, it could simply be that the huge number of patent examiners (about 10,000 current USPTO patent examiners from what I understand) means that at least a few of them will do something astounding.

My wife has 28 cousins which I jokingly say means, they include one of everything. Such as airline pilots, or priests.

This type of talent is extremely valuable. They sometimes show it in the most direct fashion (HFTs, game engines, etc.) but when they are not most of the time companies don't even bother. IMHO We pay the price of that as a consumer every day: PCs take 10s just to start booting a system that takes 2s to load, webpages are just eating your battery life etc.

PC startup time crosses too many organizational and compatibility boundaries to be amenable to this kind of optimization.

If anyone else wondered about the mentioned LU decomposition, Wikipedia says:

> In numerical analysis and linear algebra, lower–upper (LU) decomposition or factorization factors a matrix as the product of a lower triangular matrix and an upper triangular matrix.

> Applications: Solving linear equations, Inverting a matrix, Computing the determinant

https://en.wikipedia.org/wiki/LU_decomposition

This is the first time I was liked to someones post who I was blocked by on Twitter. I wonder how that happened.

On the other hand, you've got a "coder" that doesn't know what LU decomposition is...

Different people have different ideas of the meaning of "understanding". Take simple concepts like "time" and ask a physicist.

That would be the norm, yes.

So, just like 99.99% of coders, including famous coders whose output and results run circles around many coders who do know what it is?

Reply on news.ycombinator.com