## Tuesday, June 11th, 2019

* TOC
{: toc }

### Reflection of Week of June 3, 2019

I spent 4.5/3 hours on curriculum which was good, 2/2 hours on WoofJS. I prepped for the Intel meeting for 2/5, because I ended up needing a nap the day before. I caught up on my French "debt" (and at this moment actually have a 30 min French surplus). On Friday, I finally published the podcast episode with Cyrus Omar. 

#### Intel Brainstorm

The meeting at Intel was interesting. I got a lot of time with some Urbit folks, which was useful. And it was really interesting to hear about the sorts of things CPU designers talk about, what acronyms they use, etc. Here are some of my takeaways:

* They use the acronym SOC a lot. I had never heard it before. Stands for "system on a chip." One theme of the meeting was that what used to just be a CPU now includes multiple CPU cores, as well as many other kinds of accelerators, such a GPU. If you generalize, maybe the CPU itself is just an accelerator for a certain kind of sequential code with mutable memory.
* Another theme was that when you have a system on a chip, it's really a distributed system, despite the parts being within millimeters of each other. You still have to reason about synchronization and other guarantees.
* One theme the Intel engineers were pushing was that there's a trade-off between uniform interfaces on one hand and performance/efficiency on the other hand. Generalized vs specialized compute. 
* They also used the term FPGA a lot, which Omar Rizwan talks about, and I also heard about from Conal Elliot, because he used to work for a cool company Tabula that made chips that could reconfigure themselves quickly.
* It was also interesting to hear that there's a strong division at Intel between the scientists who fabricate the smaller circuits, and the chip designers who design them, and it really could be two separate companies.
* All us software people got a kick out of the fact that the Intel engineers didn't know much about managed memory, nor functional programming. One of them stood up and explained that "real programming is sorting lists" at one point.
* I also learned a tiny bit about the memory cache. It sounds like they speed up languages by caching memory reads, but then they have all these cache invalidation problems... But I got the feeling that these could be solved with a better programming model, such as immutability.
* One other thing I heard said a couple times was "there's nothing more distributed (or parallel) than reality." And it made me think about how spread out in space everything is, even on a tiny chip, and how a big problem in all computation is coordinating computation over space.
* One thing I learned was that chips don't go past a certain speed sometimes for heat and power reasons, not just capacity reasons.

### Week of June 10th, 2019

Before my flight from London to NY yesterday, I put the audio of all 4 podcast episodes I've recorded on my iPad for editing. I'm over halfway done with the rntz episode. I'd like to make some podcast progress this week -- particularly because I'm not feeling like research and am not currently on other deadlines. 

Today I have 2 more hours, tomorrow I have ~4 hours, Thursday I have ~4 hours, and Friday I may only have ~2 hours, so just 12 more hours this week. Short week because travel yesterday, and I have my mom's birthday in DC Thursday, and my cousin's bar mitzvah in Maryland Friday/Saturday/Sunday. 

Here's what I'd like to get done:

1. Publish rntz episode (6 hours)
2. Meditate on future of my research, including Alan Kay reading list, category theory group, trading with Sam Loncar (2 hours)
3. Slack old search (2 hours)
4. Get off c9 (2 hours)
5. Misc #priority1 #research in workflowy (2 hours)
6. French (1 hour)
7. Curriculum (2 hours, already subtracted from time)

That's 3 extra hours, so we'll see how much I have time for the end of this week and this weekend... Next week is super open, so maybe I'll do some research! Maybe publish another podcast. Maybe research alternative platforms to Slack.

Author: Steve Krouse