If this interests you, you should give her phd thesis a read. Easily one of the craziest kernels out there; a 1980s optimizing JIT kernel that goes to crazy lengths even when viewed against today.

http://valerieaurora.org/synthesis/SynthesisOS/index.html

Here's a PDF version: http://citeseerx.ist.psu.edu/viewdoc/summary?doi=10.1.1.29.4...

Oh my God, the end of the article is absolute gold! What a lovely tradition. I'm going to start offering piggyback rides to everyone I meet. Here's an excerpt of Alexia's experiences offering piggyback rides to various notable people:

> The Mighty Piggybacker

Marvin Minsky Lofty pioneer in artificial intelligence. "Great view! I could almost touch the bottoms of the traffic lights!"

__Ken Thompson and Dennis Ritchie __Creators of Unix. "I looked up to them as an undergrad at Cooper Union. Who would have imagined I'd actually get to meet them, let alone offer piggybacks! And afterward I got to fly with Ken in his plane!"

Zvi Galil Dean, School of Engineering, Columbia University. "Very fit. Jogs umpteen miles a day. Carries well."

__Penn & Teller __Comedians. "I carried Teller. But Penn (the big one) could not be persuaded that his weight was not an issue."

Calton Pu Massalin's PhD adviser. "Excellent piggybacks. Carried me 30 city blocks once! That's my longest piggyback ever from a single person."

__Neil Reynolds __Fellow grad student at Columbia. "At 6' 6", 350-plus pounds, the heaviest person I've carried. His rides are spectacular. He hardly notices it when I jump on."

__Renate Valencia __Massalin's girlfriend. "Much more than piggyback rides: also in-her-arms carries, over-her-shoulder carries, assorted spins and tosses. She can grab me by my armpits and twirl me round in the air!"

__Mort Meyerson __CEO, Perot Systems. "Not only accepted my offer, but did so with much enthusiasm. Restored my faith in humanity."

Richard Stallman Founder of the Free Software Foundation. "Impolite shrug; I did not even finish offering the piggyback when it became clear that further conversation would not have been appreciated."

__Douglas Adams __Author of The Hitchhiker's Guide to the Galaxy. "Unceremoniously ignored an offer of a piggyback, as if I was crazy or something :-)"

__Nicholas Negroponte __Director of the MIT Media Lab. "Very politely declined several offers."

Many years ago, I was in a group at Apple that interviewed Massalin. It was decided not to hire, as, while brilliant, there would be a net drop in productivity, as we would all spend too much time discussing the latest oddity. As one might expect, there was a stuffed koala toy on the table during Massalin's job talk. Brilliant work on the super-optimizer.

That seems like a really sad reason for no-hire.

It’s hard not to think that the diversity of thought and personality would have been a net positive.

Though I suppose it all depends on the other team personalities as well.

On John Regehr’s blog there was some interesting musings about the Synthesis kernel ideas in 2019 [0], discussed on HN at the time here [1].

[0]: https://blog.regehr.org/archives/1676

[1]: https://news.ycombinator.com/item?id=20337231

> Then Massalin discovered that he could turn it around and use the superoptimizer to discover and design better machine languages.

Any reference on this?

Thanks, I'm not so sure that superoptimizers have been used to design ISA, it seems more like "what could I do with this tool?".

I suspect it was a misinterpretation on the part of the interviewer. If you make the (to a nontechnical writer) minor rephrasing of "design better machine languages" as "write better machine language" you get something better describing what the superoptimizer helps one do.

I think the interviewer interpreted Massalin’s comment correctly.

She was (is?) involved in the design of the MicroUnity ISA.

You can find her name on quite a few of their patents.

She also posted to comp.compilers[0] about an improved version of the superoptimizer.

Their product was targeted at media codecs and signal processing so it would be natural to use a superoptimizer to identify whether an instruction was worth including in the ISA.

AFAIK MicroUnity never shipped a product but filed a lot of early patents on SIMD. They sued a number of tech companies for patent infringement so they may still exist.

[0] https://compilers.iecc.com/comparch/article/97-07-088