Chris Lattner: Compilers, LLVM, Swift, TPU, and ML Accelerators

Chris Lattner is a senior director at Google working on several projects including CPU, GPU, TPU accelerators for TensorFlow, Swift for TensorFlow, and all kinds of machine learning compiler magic going on behind the scenes. He is one of the top experts in the world on compiler technologies, which means he deeply understands the intricacies of how hardware and software come together to create efficient code. He created the LLVM compiler infrastructure project and the CLang compiler. He led major engineering efforts at Apple, including the creation of the Swift programming language. He also briefly spent time at Tesla as VP of Autopilot Software during the transition from Autopilot hardware 1 to hardware 2, when Tesla essentially started from scratch to build an in-house software infrastructure for Autopilot. If you would like to get more information about this podcast go to https://lexfridman.com/ai or connect with @lexfridman on Twitter https://twitter.com/lexfridman, LinkedIn https://www.linkedin.com/in/lexfridman/, Facebook https://www.facebook.com/lexfridman, Medium https://medium.com/@lexfridman, or YouTube https://www.youtube.com/lexfridman where you can watch the video versions of these conversations.

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:00Lex Fridman?

  The following is a conversation with Chris Lattner. Currently, he's a senior director at Google, working on several projects, including CPU, GPU, TPU accelerators for TensorFlow, Swift for TensorFlow, and all kinds of machine learning compiler magic going on behind the scenes. He's one of the top experts in the world on compiler technologies, which means he deeply understands the intricacies of how hardware and software come together to create efficient code. He created the LLVM compiler infrastructure project and the Klang compiler. He led major engineering efforts at Apple, including the creation of the Swift programming language. He also briefly spent time at Tesla as vice president of autopilot software during the

• 0:44Lex Fridman?

  transition from autopilot hardware one to hardware two, when Tesla essentially started from scratch to build an in-house software infrastructure for autopilot. I could have easily talked to Chris for many more hours. Compiling code down across the levels of abstraction is one of the most fundamental and fascinating aspects of what computers do, and he is one of the world experts in this process. It's rigorous science and it's messy, beautiful art. This conversation is part of the Artificial Intelligence Podcast. If you enjoy it, subscribe on YouTube, iTunes, or simply connect with me on Twitter at LexFriedman, spelled F-R-I-D. And now, here's my conversation with Chris Latner.

• 1:46Lex Fridman?

  What was the first program you've ever written?

• 1:50Chris Lattner?

  My first program?

• 1:51Lex Fridman?

  Back. And when was it?

• 1:52Chris Lattner?

  I think I started as a kid, and my parents got a basic programming book. And so when I started, it was typing out programs from a book and seeing how they worked and then typing them in wrong and trying to figure out why they were not working right, that kind of stuff.

• 2:10Lex Fridman?

  So BASIC, what was the first language that you remember yourself maybe falling in love with, like really connecting with?

• 2:17Chris Lattner?

  I don't know. I mean, I feel like I've learned a lot along the way and each of them have a different special thing about them. So I started in BASIC and then went like GW BASIC, which was the thing back in the DOS days, and then upgraded to Q BASIC and eventually QUICK BASIC, which are all slightly more fancy versions of Microsoft BASIC, made the jump to Pascal and started doing machine language programming and assembly in Pascal, which was really cool. Turbo Pascal was amazing for its day. Eventually gone to C, C++ and then kind of did lots of other weird things.

• 2:50Lex Fridman?

  I feel like you took the dark path, which is the, you could have gone Lisp.

• 2:56Chris Lattner?

  Yeah.

• 2:57Lex Fridman?

  You could have gone higher level, sort of functional, philosophical hippie route. Instead, you went into like the dark arts of the.

• 3:05Chris Lattner?

  It was straight, straight into the machine.

• 3:06Lex Fridman?

  Straight to the machine.

• 3:07Chris Lattner?

  So started with BASIC, Pascal and then assembly and then wrote a lot of assembly. And.

• 3:13Lex Fridman?

  Why?

• 3:14Chris Lattner?

  Eventually, eventually did Smalltalk and other things like that, but that was not the starting point.

• 3:19Lex Fridman?

  But so what, what is this journey to C? Is that in high school? Is that in college?

• 3:24Chris Lattner?

  That was in high school. Yeah. So and then that was, it was really about trying to be able to do more powerful things than what Pascal could do and also to learn a different world. So he was really confusing to me with pointers and the syntax and everything and it took a while, but Pascal's much more principled in various ways, sees more, I mean, it has its historical roots, but it's, it's not as easy to learn.

• 3:52Lex Fridman?

  With pointers, there's this memory management thing that you have to become conscious of. Is that the first time you start to understand that there's resources that you're supposed to manage?

• 4:03Chris Lattner?

  Well, so you have that in Pascal as well, but in Pascal, they use like the carrot instead of the. There's some small differences like that, but it's not about pointer arithmetic. And see, you end up thinking about how things get laid out in memory a lot more. And so in Pascal, you have allocating and de-allocating and owning the memory, but just the programs are simpler and you don't have to. Well, for example, Pascal has a string type. And so you can think about a string instead of an array of characters which are consecutive in memory. So it's a little bit of a higher level abstraction.

• 4:37Lex Fridman?

  So let's get into it. Let's talk about LLVM, C lang and compilers.

• 4:42Chris Lattner?

  Sure.

• 4:43Lex Fridman?

  So can you tell me first what LLVM and C lang are? And how is it that you find yourself the creator and lead developer? One of the most powerful compiler optimization systems in use today?

• 4:57Chris Lattner?

  Sure. So I guess they're different things. So let's start with what is a compiler?

• 5:04Lex Fridman?

  It's a.

• 5:04Chris Lattner?

  Is that a good place to start?

• 5:06Lex Fridman?

  What are the phases of a compiler? Where are the parts? Yeah. What is it?

• 5:08Chris Lattner?

  So what is even a compiler used for? So the way, the way I look at this is you have a two sided problem of you have humans that need to write code. And then you have machines that need to run the program that the human wrote. And for lots of reasons, the humans don't want to be writing in binary and want to think about every piece of hardware. And so at the same time that you have lots of humans, you also have lots of kinds of hardware. And so compilers are the art of allowing humans to think at a level of abstraction that they want to think about. And then get that program, get the thing that they wrote to run on a specific piece of hardware. And the interesting and exciting part of all this is that there's now lots of different kinds of hardware. Chips like x86 and Power PC and ARM and things like that.

• 5:53Chris Lattner?

  But also high performance accelerators for machine learning and other things like that are also just different kinds of hardware. GPUs, these are new kinds of hardware. And at the same time, on the programming side of it, you have, you have basic, you have C, you have JavaScript, you have Python, you have Swift. You have like lots of other languages that are all trying to talk to the human in a different way to make them more expressive and capable and powerful. And so.

• 6:17Chris Lattner?

  Compilers are the thing that goes from one to the other now.

• 6:20Lex Fridman?

  End to end, from the very beginning to the very end.

• 6:22Chris Lattner?

  End to end. And so you go from what the human wrote and programming languages end up being about expressing intent, not just for the compiler and the hardware, but the programming language's job is really to to capture an expression of what the programmer wanted that then can be maintained and adapted and evolved by other humans, as well as by the interpreted by the compiler. So so when you look at this problem, you have on one hand, humans, which are complicated, and you have hardware, which is complicated. And so compilers typically work in multiple phases. And so the software engineering challenge that you have here is try to get maximum reuse out of the the amount of code that you write, because this these compilers are very complicated.

• 7:07Chris Lattner?

  And so the way it typically works out is that you have something called a front end or a parser that is language specific. And so you'll have a C parser, and that's what Clang is, or C++ or JavaScript or Python or whatever. That's the front end. Then you'll have a middle part, which is often the optimizer. And then you'll have a late part, which is hardware specific. And so compilers end up, there's many different layers often, but these three big groups are very common in compilers. And what LVM is trying to do is trying to standardize that middle and last part. And so one of the cool things about LVM is that there are a lot of different languages that compile through to it. And so things like Swift, but also Julia, Rust,

• 7:53Chris Lattner?

  Clang for C, C++, Objective C, like these are all very different languages and they can all use the same optimization infrastructure, which gets better performance, and the same code generation infrastructure for hardware support. And so LVM is really that that layer that is common, that all these different specific compilers can use.

• 8:12Lex Fridman?

  And is it is it a standard, like a specification, or is it literally an implementation?

• 8:18Chris Lattner?

  It's an implementation. And so it's, I think there's a couple of different ways of looking at it, right? Because it depends on which which angle you're looking at it from. LVM ends up being a bunch of code. Okay, so it's a bunch of code that people reuse and they build compilers with. We call it a compiler infrastructure because it's kind of the underlying platform that you build a concrete compiler on top of. But it's also a community and the LVM community is hundreds of people that all collaborate. And one of the most fascinating things about LVM over the course of time is that we've managed somehow to successfully get harsh competitors in the commercial space to collaborate on shared infrastructure.

• 8:58Chris Lattner?

  And so you have Google and Apple, you have AMD and Intel, you have Nvidia and AMD on the graphics side. You have Cray and everybody else doing these things. And like all these companies are collaborating together to make that shared infrastructure really, really great. And they do this not out of the goodness of their heart, but they do it because it's in their commercial interest of having really great infrastructure that they can build on top of. And facing the reality that it's so expensive that no one company, even the big companies, no one company really wants to implement it all themselves.

• 9:31Lex Fridman?

  Expensive or difficult?

• 9:33Chris Lattner?

  Both. That's a great point because it's also about the skill sets.

• 9:37Lex Fridman?

  Right.

• 9:37Chris Lattner?

  And these, the skill sets are very hard, hard to find.

• 9:43Lex Fridman?

  How big is the LVM? It always seems like with open source projects, the kind, and LVM is open source?

• 9:50Chris Lattner?

  Yes, it's open source. It's about, it's 19 years old now, so it's fairly old.

• 9:55Lex Fridman?

  It seems like the magic often happens within a very small circle of people.

• 10:00Chris Lattner?

  Yes.

• 10:00Lex Fridman?

  At like at least their early birth and whatever.

• 10:03Chris Lattner?

  Yes. So the LVM came from a university project. And so I was at the University of Illinois. And there it was myself, my advisor, and then a team of two or three research students in the research group. And we built many of the core pieces initially. I then graduated and went to Apple and Apple brought it to the products, first in the OpenGL graphics stack. But eventually to the C compiler realm and eventually built Clang and eventually built Swift and these things. Along the way, building a team of people that are really amazing compiler engineers that helped build a lot of that. And so as it was gaining momentum and as Apple was using it. Being open source and public and encouraging contribution, many others, for example, Google

• 10:46Chris Lattner?

  came in and started contributing. And in some cases, Google effectively owns Clang now because it cares so much about C++ and the evolution of that ecosystem. And so it's investing a lot in the C++ world and the tooling and things like that. And so likewise, Nvidia cares a lot about CUDA. And so CUDA uses Clang and uses LVM for graphics and GPGPU.

• 11:12Lex Fridman?

  And so when you first started as a master's project, I guess, did you think it's gonna go as far as it went? Were you crazy ambitious about it?

• 11:23Chris Lattner?

  No.

• 11:23Lex Fridman?

  It seems like a really difficult undertaking, a brave one.

• 11:26Chris Lattner?

  Yeah, no, no, no, it was nothing like that. So I mean, my, my goal when I went to the University of Illinois was to get in and out with a non thesis masters in a year and get back to work. So I was not, I was not planning to stay for five years and, and build this massive infrastructure. I got nerd sniped into staying. And a lot of it was because LVM was fun and I was building cool stuff and learning really interesting things and facing both software engineering challenges, but also learning how to work in a team and things like that. I had worked at many companies as interns before that. But it was really a different, a different thing to have a team of people that were working together and trying to collaborate in version control. And it was, it was just a little bit different.

• 12:09Lex Fridman?

  Like I said, I just talked to Don Knuth and he believes that 2% of the world population have something weird with their brain, that they're geeks. They understand computers, they're connected with computers. He put it exactly 2%. Okay, so.

• 12:22Chris Lattner?

  He's a specific guy.

• 12:23Lex Fridman?

  It's very specific. Well, he says, I can't prove it, but it's very empirically there.

• 12:30Lex Fridman?

  Is there something that attracts you to the idea of optimizing code? And it seems like that's one of the biggest coolest things about LVM.

• 12:38Chris Lattner?

  Yeah, that's one of the major things it does. So I got into that because of a person, actually. So when I was in my undergraduate, I had an advisor or a professor named Steve Wegdahl and he. I went to this little tiny private school. There were like seven or nine people in my computer science department, students in my, in my class. So it was a very tiny, uh, very, very small school. It was a kind of a wart on the side of the math department, kind of a thing at the time. I think it's evolved a lot in the many years since then, but, but Steve Wegdahl was a compiler guy and he was super passionate and he, um, his passion rubbed off on me.

• 13:19Chris Lattner?

  And one of the things I like about compilers is that they're large, complicated software pieces. And so one of the, um, culminating classes that many computer science departments, at least at the time, did was to say that you take algorithms and data structures and all these core classes. But then the compilers class was one of the last classes you take because it pulls everything together. And then you work on one piece of code over the entire semester. And, um, and so you keep building on your own work, which is really interesting. It's also very challenging because in many classes, if you don't get a project done, you just forget about it and move on to the next one and get your, you know, your B or whatever it is. But here you have to live with the decisions you make and continue to reinvest in it. And I really like that.

• 14:04Chris Lattner?

  And, uh, and so I did a extra study project with him the, the following semester and he was just really great. And he was also a great mentor in a lot of ways. And so from, from him and from his advice, he encouraged me to go to graduate school. I wasn't super excited about going grad school. I, I wanted the master's degree, but I didn't want to be an academic. And, um, but like I said, I kind of got tricked into saying, and was having a lot of fun and I definitely do not regret it.

• 14:31Lex Fridman?

  What aspects of compilers were the things you connected with? So LVM, there's also the other part. This is really interesting. If you're interested in languages is parsing and

• 14:43Chris Lattner?

  Yeah.

• 14:43Lex Fridman?

  just analyzing like, yeah, analyzing the language, breaking it down, parsing and so on. Was that interesting to you? Or were you more interested in optimization?

• 14:51Chris Lattner?

  For me, it was more, so I'm, I'm not really a math person. I could do math. I understand some bits of it when I get into it. But math is never the thing that, that attracted me. And so a lot of the parser part of the compiler has a lot of good formal theories that Dawn, for example, knows quite well. Still waiting for his book on that, but. But the, but I just like building a thing and, and seeing what it could do and exploring and getting it to do more things and then setting new goals and reaching for them. And, um, and with, in the case of, in the case of LVM, when I started working on that, my research advisor that I was working for was a compiler guy. And so I, he and I specifically found each other because we're both interested in compilers. And so I started working with them and taking his class.

• 15:36Chris Lattner?

  And a lot of LVM initially was it's fun implementing all the standard algorithms and all the, all the things that people had been talking about and were well known and they were in the, the curricula for advanced studies and compilers. And so just being able to build that was really fun. And I was learning a lot by instead of reading about it, just building. And so I enjoyed that.

• 15:57Lex Fridman?

  So you said compilers are these complicated systems. Can you even just with language try to describe, uh, you know, how you turn a C++ program?

• 16:08Chris Lattner?

  Yes.

• 16:08Lex Fridman?

  Into code, like what are the hard parts? Why is it so hard?

• 16:11Chris Lattner?

  So I'll give you examples of the hard parts along the way. So C++ is a very complicated programming language. It's something like 1,400 pages in the spec. So C++ by itself is crazy complicated.

• 16:23Lex Fridman?

  Can we just sorry, pause? What makes the language complicated in terms of what's syntactically like?

• 16:29Chris Lattner?

  Uh, so it's what they call syntax. So the actual how the characters are arranged. Yes, it's also semantics, how it behaves. It's also in the case of C++ there's a huge amount of history. C++ is built on top of C. You play that forward and then a bunch of suboptimal, in some cases, decisions were made and they compound. And then more and more and more things keep getting added to C++ and it will probably never stop. But the, the language is very complicated from that perspective. And so the interactions between subsystems is very complicated. There's just a lot there. And when you talk about the front end, one of the major challenges, which Clang as a project, the, the C C++ compiler that I built, I and many people built, um, well, the challenges we took on was.

• 17:13Chris Lattner?

  We looked at GCC. Okay, GCC at the time was like a really good industry standardized compiler that had really consolidated a lot of the other compilers in the world and was a standard. But it wasn't really great for research. The design was very difficult to work with. And it was full of global variables and other things that made it very difficult to reuse in ways that it wasn't originally designed for. And so with Clang, one of the things that we wanted to do is push forward on better user interface. So make error messages that are just better than GCC's. And that's actually hard because you have to do a lot of bookkeeping in an efficient way to be able to do that. We want to make compile time better. And so compile time is about making it efficient, which is also really hard when you're keeping

• 17:56Chris Lattner?

  track of extra information. We wanted to make new tools available. So refactoring tools and other analysis tools that GCC never supported, also leveraging the extra information we kept, but enabling those new classes of tools that then get built into IDEs. And so that's been one of the areas that Clang has really helped push the world forward in is in the tooling for C and C++ and things like that. But C++ and the front-end piece is complicated and you have to build syntax trees and you have to check every rule in the spec and you have to turn that back into an error message to the human that the human can understand when they do something wrong. But then you start doing the, what's called lowering. So going

• 18:38Chris Lattner?

  from C++ in the way that it represents code down to the machine. And when you do that, there's many different phases you go through. Often, there are, I think LVM has something like 150 different, what are called passes in the compiler, that the code passes through. And these get organized in very complicated ways, which affect the generated code and the performance and compile time and many other things. What are they passing through? So after you do the Clang parsing, what's the graph? What does it look like? What's the data structure here? Yeah, so in the parser, it's usually a tree. And it's called an abstract syntax tree. And so the idea is you have a node for the plus that the human wrote in their code,

• 19:23Chris Lattner?

  or the function call, you'll have a node for call with the function that they call and the arguments they pass, things like that.

• 19:31Chris Lattner?

  This then gets lowered into what's called an intermediate representation. And intermediate representations are like LVM has one. And there it's a it's what's called a control flow graph. And so you represent each operation in the program as a very simple. Like this is going to add two numbers. This is going to multiply two things. Maybe we'll do a call, but then they get put in what are called blocks. And so you get blocks of these straight line operations where instead of being nested like in a tree, it's straight line operations. And so there's a sequence and ordering to these operations.

• 20:06Lex Fridman?

  So within the block or outside the block?

• 20:08Chris Lattner?

  That's within the block. And so it's a straight line sequence of operations within the block. And then you have branches, like conditional branches between blocks. And so when you write a loop, for example, in a syntax tree, you would have a for node, like for a for statement in a C like language, you'd have a for node and you have a pointer to the expression for the initializer, a pointer to the expression for the increment, a pointer to the expression for the comparison, a pointer to the body. Okay, and these are all nested underneath it. In a control flow graph, you get a block for the code that runs before the loop. So the initializer code, then you have a block for the body of the loop. And so the the body of the loop code goes in there, but also the increment and other things like that.

• 20:52Chris Lattner?

  And then you have a branch that goes back to the top and a comparison and a branch that goes out. And so it's more of an assembly level kind of representation. But the nice thing about this level of representation is it's much more language independent. And so there's lots of different kinds of languages with different kinds of, you know, JavaScript has a lot of different ideas of what is false, for example, and all that can stay in the front end, but then that middle part can be shared across all those.

• 21:21Lex Fridman?

  How close is that intermediate representation to neural networks? For example, are they? Because everything you describe is a kind of a closed neural network graph.

• 21:33Chris Lattner?

  Yeah.

• 21:33Lex Fridman?

  Are they neighbors or what?

• 21:35Chris Lattner?

  They're quite different in details, but they're very similar in idea. So one of the things that neural networks do is they learn representations for data at different levels of abstraction.

• 21:45Lex Fridman?

  Right.

• 21:46Chris Lattner?

  And then they transform those through layers. Right?

• 21:51Chris Lattner?

  So the compiler does very similar things, but one of the things the compiler does is it has relatively few different representations. Where a neural network often, as you get deeper, for example, you get many different representations in each, you know, layer or set of ops is transforming between these different representations. In a compiler, often you get one representation and they do many transformations to it. And these transformations are often applied iteratively. And for programmers, there are familiar types of things. For example, trying to find expressions inside of a loop and pulling them out of a loop. So they execute fewer times or find redundant computation or find

• 22:30Chris Lattner?

  constant folding or other simplifications, turning, you know, two times X into X shift left by one. And, and things like this are all, all, all the examples of the things that happen. But compilers end up getting a lot of theorem proving and other kinds of algorithms that try to find higher level properties of the program that then can be used by the optimizer.

• 22:49Lex Fridman?

  Cool. So what's like the biggest bang for the buck with optimization? What's a?

• 22:55Chris Lattner?

  Today? Yeah.

• 22:56Lex Fridman?

  Well, no, not even today. At the very beginning, the eighties, I don't know.

• 22:59Chris Lattner?

  But yeah, so for the eighties, a lot of it was things like register allocation. So the idea of in, in a modern, like a microprocessor, what you'll end up having is you'll end up having memory, which is relatively slow. And then you have registers that are relatively fast, but registers, you don't have very many of them. Okay. And so when you're writing a bunch of code, you're just saying like, compute this, put it in temporary variable, compute this, compute this, compute this, put in temporary variable. I have a loop, I have some other stuff going on. Well, now you're running on an x86, like a desktop PC or something. Well, it only has, in some cases, some modes, eight registers. Right? And so now the compiler has to choose what values get put in what registers at what points in the program. And this is actually a really big deal.

• 23:43Chris Lattner?

  So if you think about, you have a loop and then an inner loop that executes millions of times, maybe. If you're doing loads and stores inside that loop, then it's going to be really slow. But if you can somehow fit all the values inside that loop and registers, now it's really fast. And so. Getting that right requires a lot of work because there's many different ways to do that. And often what the compiler ends up doing is it ends up thinking about things in a different representation than what the human wrote. You wrote into X. Well, the compiler thinks about that as four different values, each which have different lifetimes across the function that it's in. And each of those could be put in a register or memory or different memory, or maybe in some parts of the code recomputed instead of stored and reloaded. And there are many of these different kinds of techniques that can be used.

• 24:28Lex Fridman?

  So it's adding almost like a time dimension to, it's trying to, it's trying to optimize across time. So it's considering when, when you're programming, you're not thinking in that way.

• 24:39Chris Lattner?

  Yeah, absolutely. And so the, the RISC era made things, so, so RISC chips, R I S C, the, the RISC, RISC chips, as opposed to CISC chips, the RISC chips made things more complicated for the compiler because what they ended up doing is ending up adding pipelines to the processor where the processor can do more than one thing at a time. But this means that the order of operations matters a lot. And so one of the classical compiler techniques that you use is called scheduling. And so moving the instructions around so that the processor can like keep its pipelines full instead of stalling and getting blocked. And so there's a lot of things like that that are kind of bread and butter compiler techniques that have been studied a lot over the course of decades now. But the engineering

• 25:24Chris Lattner?

  side of making them real is also still quite hard. And you talk about machine learning. This is, this is a huge opportunity for machine learning because many of these algorithms are full of these like hokey hand-rolled heuristics which work well on specific benchmarks, but don't generalize and full of magic numbers. And, you know, I hear there's some techniques that are good at handling that.

• 25:44Lex Fridman?

  So what would be the, if you were to apply machine learning to this, what's the thing you're trying to optimize? Is it ultimately the running time?

• 25:55Chris Lattner?

  Yeah, you can pick your metric and there's, there's running time, there's memory use, there's, there's lots of different things that you can optimize for. Code, code size is another one that some people care about in the embedded space.

• 26:05Lex Fridman?

  Is this like the thinking into the future or is somebody actually? Been crazy enough to try to have machine learning based parameter tuning for optimization of compilers.

• 26:18Chris Lattner?

  So this is something that is, I would say, research right now. There are a lot of research systems that have been applying search in various forms. And using reinforcement learning is one form, but also brute force search has been tried for quite a while. And usually, these are in small, small problem spaces. So find the optimal way to code generate a matrix multiply for a GPU. Right, something like that, where you say, there, there's a lot of design space of, do you unroll loops a lot? Do you execute multiple things in parallel? And there's many different confounding factors here because graphics cards have different numbers of threads and registers and execution ports and memory bandwidth and many different constraints to interact in nonlinear ways. And so search is very powerful for that.

• 27:03Chris Lattner?

  And it gets used in, in certain ways, but it's not very structured. This is something that we need, we as an industry need to fix.

• 27:11Lex Fridman?

  So you said eighties, but like, so have there been like big jumps in improvement and optimization?

• 27:18Chris Lattner?

  Yeah.

• 27:18Lex Fridman?

  Yeah, since then, what's?

• 27:21Chris Lattner?

  Yeah, so it's largely been driven by hardware. So hardware and software. So in the mid nineties, Java totally changed the world, right? And, and I'm still amazed by how much change was introduced by Java.

• 27:35Lex Fridman?

  Way or?

• 27:35Chris Lattner?

  In, in a good way. So like reflecting back, Java introduced things like, all at once, introduced things like JIT compilation. None of these were novel, but it pulled it together and made it mainstream and, and made people invest in it. JIT compilation, garbage collection, portable code, safe code, like memory safe code. Like a very dynamic dispatch execution model. Like many of these things, which had been done in research systems and had been done in small ways in various places, really came to the forefront and really changed how things worked. And therefore changed the way people thought about the problem. JavaScript was another major world change based on the way it works.

• 28:16Chris Lattner?

  But also on the hardware side of things, multi-core and vector instructions really change the problem space and are very, they don't remove any of the problems that compilers faced in the past, but they, they add new kinds of problems of how do you find enough work to keep a four wide vector busy, right? Or if you're doing a matrix multiplication. How do you do different columns out of that matrix in at the same time? And how do you maximally utilize the the arithmetic compute that one core has? And then how do you take it to multiple cores?

• 28:50Lex Fridman?

  How did the whole virtual machine thing change the compilation pipeline?

• 28:55Chris Lattner?

  Yeah, so what the Java virtual machine does is it splits, just like I was talking about before, where you have a front end that parses the code. And then you have an intermediate representation that gets transformed. What Java did was they said, we will parse the code and then compile to what's known as Java bytecode. And that bytecode is now a portable code representation that is industry standard and locked down and can't change. And then the the back part of the compiler that does optimization and code generation can now be built by different vendors. Okay, and Java bytecode can be shipped around across the wire. It's memory safe and relatively trusted.

• 29:34Lex Fridman?

  And because of that, it can run in the browser?

• 29:35Chris Lattner?

  And that's why it runs in the browser. Right? And so that way you can be in, you know, again, back in the day, you would write a Java applet and you'd use a as a little, as a web developer, you'd build this mini app that would run a webpage. Well, a user of that is running a web browser on their computer. You download that that Java bytecode, which can be trusted. And then you do all the compiler stuff on your machine so that you know that you trust that.

• 29:59Lex Fridman?

  Now, was that a good idea or a bad idea?

• 30:01Chris Lattner?

  It's a great idea. I mean, it's a great idea for certain problems. And I'm very much a believer that technology is itself neither good nor bad. It's how you apply it.

• 30:10Chris Lattner?

  You know, this would be a very, very bad thing for very low levels of the software stack. But but in terms of solving some of these software portability and transparency or portability problems, I think it's been really good. Now, Java ultimately didn't win out on the desktop. And like, there are good reasons for that. But it's been very successful on servers and in many places. It's been a very successful thing over over decades.

• 30:33Lex Fridman?

  So what has been L L V M's and C lang's improvements in optimization that throughout its history? What are some moments where you had setback and really proud of what's been accomplished?

• 30:50Chris Lattner?

  Yeah, I think that the interesting thing about L L V M is not the innovations in compiler research. It has very good implementations of various important algorithms, no doubt.

• 31:02Chris Lattner?

  And And a lot of really smart people have worked on it. But I think that the thing that's most profound about LVM is that through standardization, it made things possible that otherwise wouldn't have happened. Okay. And so interesting things that have happened with LVM. For example, Sony has picked up LVM and used it to do all the graphics compilation in their movie production pipeline. And so now they're able to have better special effects because of LVM. That's kind of cool. That's not what it was designed for, right? But that's, that's the sign of good infrastructure when it can be used in ways it was never designed for, because it has good layering and software engineering and it's composable and things like that.

• 31:40Lex Fridman?

  Which is where, as you said, it differs from GCC.

• 31:43Chris Lattner?

  Yes. GCC is also great in various ways, but it's not as good as a infrastructure technology. It's, it's, you know, it's really a C compiler or it's, or it's a Fortran compiler. It's not, it's not infrastructure in the same way.

• 31:56Lex Fridman?

  Is it, now you can tell I don't know what I'm talking about because I'm, I keep saying C lang. You can, you could always tell when a person is close, by the way, pronounce something. I'm, I don't think, have I ever used clang?

• 32:09Chris Lattner?

  Entirely possible. Have you, well, so you've used code, it's generated probably. So clang is, and LVM are used to compile all the apps on the iPhone effectively. And the OSes, it compiles Google's production server applications. It's used to build like GameCube games and PlayStation four and things like that.

• 32:33Lex Fridman?

  So as a user I have, but just everything I've done that I experienced for Linux has been, I believe, always GCC.

• 32:40Chris Lattner?

  Yeah, I think Linux still defaults to GCC.

• 32:43Lex Fridman?

  And is there a reason for that? Or is it, I mean, is there?

• 32:46Chris Lattner?

  It's, it's a combination of technical and social reasons. Many, many Linux developers do use, do use clang, but the distributions for lots of reasons have used GCC historically and they've not switched.

• 33:00Lex Fridman?

  Yeah. Because it's just anecdotally online, it seems that LVM has either reached the level GCC or superseded on different features or whatever.

• 33:10Chris Lattner?

  The way I would say it is that they're, they're so close, it doesn't matter.

• 33:13Lex Fridman?

  Yeah, exactly.

• 33:13Chris Lattner?

  Like they're slightly better in some ways, slightly worse in other ways, but. It doesn't actually really matter anymore at that level.

• 33:20Lex Fridman?

  So in terms of optimization breakthroughs, it's just been solid incremental work.

• 33:26Chris Lattner?

  Yeah, yeah, which, which is, which describes a lot of compilers. The hard, the hard thing about compilers in my experience is the engineering, the software engineering, making it so that you can have hundreds of people collaborating on really detailed, low level work and scaling that. And that's, that's really hard. And that's one of the things I think LVM has done well. And that kind of goes back to the original design goals with it to be modular and things like that. And incidentally, I don't want to take all the credit for this, right? I mean, some of the, the best parts about LVM is that it was designed to be modular. And when I started, I would write, for example, a register allocator. And then somebody much smarter than me would come in and pull it out and replace it with something else that they would come up with.

• 34:09Chris Lattner?

  And because it's modular, they were able to do that. And that's one of the challenges with, with GCC, for example, is replacing subsystems is incredibly difficult. It, it can be done, but it wasn't designed for that. And that's one of the reasons that LVM has been very successful in the research world as well.

• 34:25Lex Fridman?

  But in the, in the community sense, Guido van Rossum, right, from Python, just retired from, what is it? Benevolent Dictator for Life, right? So in managing this community of brilliant compiler folks, is there, did it, for a time at least, fall on you to, to approve things?

• 34:48Chris Lattner?

  Oh, yeah. So I, I mean, I still have something like an order of magnitude more patches in LVM than anybody else. And many of those I wrote myself, but.

• 35:00Lex Fridman?

  You still write, I mean, you still, you still close to the, to the, I don't know what the expression is, to the metal, you still write code.

• 35:08Chris Lattner?

  Yeah, I still write code, not as much as I was able to in grad school, but that's an important part of my identity. But the way that LVM has worked over time is that when I was a grad student, I could do all the work and steer everything and review every patch and make sure everything was done exactly the way my opinionated sense felt like it should be done. And that was fine, but as things scale, you can't do that. Right. And so what ends up happening is LVM has a hierarchical system of what's called code owners. These code owners are given the responsibility not to do all the work, not necessarily to review all the patches, but to make sure that the patches do get reviewed and make sure that the right thing's happening architecturally in their area.

• 35:49Chris Lattner?

  And so what you'll see is you'll see that, for example, hardware manufacturers end up owning the the hardware specific parts of their, their, their hardware. That's very common. Leaders in the community that have done really good work naturally become the de facto owner of something. And then usually somebody else is like, how about we make them the official code owner? And then, and then we'll have somebody to make sure that all the patches get reviewed in a timely manner. And then everybody's like, yes, that's obvious. And then it happens, right? And usually this is a very organic thing, which is great. And so I'm nominally the top of that stack still, but I don't spend a lot of time reviewing patches.

• 36:28Chris Lattner?

  What I do is I help negotiate a lot of the, the technical disagreements that end up happening and making sure that the community as a whole makes progress and is moving in the right direction and, and doing that. So we also started a nonprofit six years ago, seven years ago. It's time's gone away and the nonprofit, the, the elevator foundation nonprofit helps oversee all the business sides of things and make sure that the events that the elevator community has are funded and set up and run correctly and stuff like that. But the foundation is very much stays out of the technical side of where, where the project is going.

• 37:06Lex Fridman?

  Right. So it sounds like a lot of it is just organic, just.

• 37:10Chris Lattner?

  Yeah. Well, and this is LVM is almost 20 years old, which is hard to believe. Somebody pointed out to me recently that LVM is now older than GCC was when LVM started.

• 37:21Chris Lattner?

  Right. So time has a way of getting away from you, but the good thing about that is it has a really robust, really amazing community of people that are in their professional lives, spread across lots of different companies, but it's a, it's a community of people that are interested in similar kinds of problems and have been working together effectively for years and have a lot of trust and respect for each other. And even if they don't always agree that, you know, we're able to find a path forward.

• 37:48Lex Fridman?

  So then in a slightly different flavor of effort, you started at Apple in 2005 with the task of making, I guess, LVM production ready. And then eventually 2013 through 2017, leading the entire developer tools department. We're talking about LLVM, Xcode, Objective C to Swift. So in a quick overview of your time there, what were the challenges? First of all, leading such a huge group of developers, what was the big motivator, dream, mission behind creating Swift, the early birth of it from Objective C and so on and Xcode?

• 38:30Chris Lattner?

  Yeah. So, so these are different questions.

• 38:33Lex Fridman?

  Yeah, I know. But I want to talk about the other stuff.

• 38:36Chris Lattner?

  I'll stay, I'll stay on the technical side, then we could talk about the big team pieces.

• 38:40Lex Fridman?

  Got it.

• 38:40Chris Lattner?

  If that's okay.

• 38:41Lex Fridman?

  Sure.

• 38:41Chris Lattner?

  So it's to really oversimplify many years of hard work. LVM was started, joined Apple, became a thing, became successful and became deployed. But then there's a question about how, what, how do we actually parse the source code? So LVM is that back part, the optimizer and the code generator. And LVM is really good for Apple as it went through a couple of hardware transitions. I joined right at the time of the Intel transition, for example, and 64-bit transitions, and then the transition to ARM with the iPhone. And so LVM was very useful for some of these kinds of things. But at the same time, there's a lot of questions around developer experience. And so if you're a programmer pounding out at the time Objective C code,

• 39:21Chris Lattner?

  The error message you get, the compile time, the turnaround cycle, the tooling and the IDE were not great. Were not as good as they could be. And so,

• 39:33Chris Lattner?

  you know, as I occasionally do, I'm like, well, okay, how hard is it to try to see compiler? And so I'm not going to commit to anybody. I'm not going to tell anybody. I'm just going to just do it on nights and weekends and start working on it. And then, you know, I built up and see there's this thing called the preprocessor, which people don't like, but it's actually really hard and complicated. And includes a bunch of really weird things like trigraphs and other stuff like that, that are, that are really nasty. And it's the crux of a bunch of the performance issues in the compiler. Started working on the parser and kind of got to the point where I'm like, oh, you know what? We could actually do this. Everybody's saying that this is impossible to do, but it's actually just hard. It's not impossible. And

• 40:13Chris Lattner?

  eventually told my manager about it and he's like, oh, wow, this is great. We do need to solve this problem. Oh, this is great. We can like get you one other person to work with you on this, you know? And slowly a team is formed and it starts taking off. And C++ for example, huge complicated language. People always assume that it's impossible to implement and it's very nearly impossible, but it's just really, really hard. And the way to get there is to build it one piece at a time incrementally. And, and there, that was only possible because we were lucky to hire some really exceptional engineers that, that knew various parts of it very well. And, and could do great things. Swift was kind of a similar thing. So Swift came from,

• 40:54Chris Lattner?

  we were just finishing off the first version of C++ support in Clang. And C++ is a very formidable and very important language, but it's also ugly in lots of ways. And you can't influence C++ without thinking there has to be a better thing. Right? And so I started working on Swift again with no hope or ambition that would go anywhere. Just let's see what could be done. Let's play around with this thing. It was, you know, me in my spare time, not telling anybody about it, kind of a thing. And it made some good progress. I'm like, actually it would make sense to do this. At the same time, I started talking with the Senior VP of Software at the time, a guy named Bertrand Sorle, and Bertrand was very encouraging. He was like, well, you know, let's, let's have fun, let's talk about this.

• 41:39Chris Lattner?

  And he was a little bit of a language guy. And so he helped guide some of the, the early work and encouraged me and like got things off the ground and eventually told other, told like my manager and told other people and, and it started making progress. The, the complicating thing with Swift was that the idea of doing a new language is not obvious to anybody, including myself.

• 42:05Chris Lattner?

  And the tone at the time was that the iPhone was successful because of Objective C.

• 42:10Lex Fridman?

  Right?

• 42:10Chris Lattner?

  Oh, interesting.

• 42:11Lex Fridman?

  And not despite of or just because of.

• 42:14Chris Lattner?

  And, and you have to understand that at the time, Apple was hiring software people that loved Objective C. Right? And it wasn't that they came despite Objective C. It's they loved Objective C and that's why they got hired. And so you had a software team that the leadership in, in many cases went all the way back to next where Objective C really became real. And so they quote unquote, grew up writing Objective C. And many of the individual engineers all were hired because they loved Objective C. And so this notion of, okay, let's do new language was kind of heretical in many ways. Right? Meanwhile, my, my sense was that the outside community wasn't really in love with Objective C. Some people were and some of the most outspoken people were, but other people were hitting

• 42:59Chris Lattner?

  challenges because it has very sharp corners and it's difficult to learn. And so one of the challenges of making Swift happen that was totally non-technical is the, the social part of what do we do? Like if we do a new language, which at Apple, many things happen that don't ship. Right? So if we, if we ship it, what, what, what is the metrics of success? Why would we do this? Why wouldn't we make Objective C better? If Objective C has problems, let's file off those rough corners and edges. And one of the major things that became the reason to do this was this notion of safety, memory safety. And the way Objective C works is that a lot of the object system and everything else is built on top of pointers in C.

• 43:44Chris Lattner?

  Objective C is an extension on top of C. And so pointers are unsafe. And if you get rid of the pointers, it's not Objective C anymore. And so fundamentally, that was an issue that you could not fix safety or memory safety without.

• 44:01Chris Lattner?

  Fundamentally changing the language. And so, once we got through that part of the the mental process and the thought process, it became a design process of saying, okay, well, if we're going to do something new, what what is good? Like, how do we think about this? And what do we like? And what are we looking for? And and that that was a very different phase of it.

• 44:19Lex Fridman?

  So what what are some design choices early on in Swift? Like, we're talking about braces. Are you making a typed language or not? All those kinds of things.

• 44:30Chris Lattner?

  Yeah, so so some of those were obvious, given the context. So a typed language, for example, Objective C is a typed language. And going with an untyped language wasn't really seriously considered. We wanted, we wanted the performance and we wanted refactoring tools and other things like that to go with typed languages.

• 44:46Lex Fridman?

  Quick dumb question.

• 44:47Chris Lattner?

  Yeah.

• 44:48Lex Fridman?

  Was it obvious? I think this would be a dumb question, but was it obvious that the language has to be a compiled language? Not an.

• 44:57Chris Lattner?

  Yes, that's not a dumb question. Earlier, I think late 90s, Apple had seriously considered moving its development experience to Java.

• 45:06Chris Lattner?

  But Swift started in 2010, which was several years after the iPhone. It was when the iPhone was definitely on an upward trajectory. And the iPhone was still extremely and is still a bit memory constrained. Right? And so being able to compile the code and then ship it and then have having standalone code that is not compiled was is is a very big deal. And it's very much part of the Apple value system. Now, JavaScript is also a thing, right? I mean, it's not, it's not that this is exclusive and technologies are good, depending on how they're applied. Right? But in the design of Swift, saying like, how can we make Objective C better? Right? Objective C is statically compiled, and that was the contiguous, natural thing to do.

• 45:49Lex Fridman?

  Just skip ahead a little bit. Now, we'll right back just just as a question. As you think about today in 2019. In your work at Google, TensorFlow and so on is, again, compilations, static compilation.

• 46:04Chris Lattner?

  The right thing.

• 46:05Lex Fridman?

  Still the right thing.

• 46:06Chris Lattner?

  Yeah, so the funny thing after working on compilers for a really long time is that, and this is one of the things that LVM has helped with, is that I don't look at compilations being static or dynamic or interpreted or not. This is a spectrum. Okay. And one of the cool things about Swift is that Swift is not just statically compiled. It's actually dynamically compiled as well. And it can also be interpreted, though nobody's actually done that. And so what ends up happening when you use Swift in a workbook, for example, in Colab or in Jupiter, is it's actually dynamically compiling the statements as you execute them. And so this gets back to the software engineering problems, right?

• 46:50Chris Lattner?

  Where if you layer the stack properly, you can actually completely change how and when things get compiled because you have the right abstractions there. And so the way that a Colab workbook works with Swift is that when you start typing into it, it creates a process, a Unix process. And then each line of code you type in, it compiles it through the Swift compiler, the front end part, and then sends it through the optimizer, Jit compiles machine code, and then injects it into that process.

• 47:21Chris Lattner?

  And so as you're typing new stuff, it's like squirting in new code and overriding and replacing and updating code in place. And the fact that it can do this is not an accident, like Swift was designed for this. But it's an important part of how the language was set up and how it's layered. And this is a non-obvious piece. And one of the things with Swift that was, for me, a very strong design point is to make it so that you can learn it very quickly. And so from a language design perspective, the thing that I always come back to is this UI principle of progressive disclosure of complexity. And so in Swift, you can start by saying, print, quote, hello, world, quote, right? And there's no slash n, just like Python, one line of code, no main, no...

• 48:05Lex Fridman?

  No header files.

• 48:06Chris Lattner?

  No header files, no public static class void, blah, blah, blah, string, like Java has, right? It's one line of code, right? And you can teach that. Works great. Then you can say, well, let's introduce variables. And so you can declare a variable with var. So var x equals four. What is a variable? You can use x, x plus one. This is what it means. Then you can say, well, how about control flow? Well, this is what an if statement is. This is what a for statement is. This is what a while statement is.

• 48:32Chris Lattner?

  And then you can say, let's introduce functions, right? And, and many languages like Python have had this, this kind of notion of let's introduce small things and then you can add complexity. Then you can introduce classes and then you can add generics in the case of Swift. And then you can, and modules and build out in terms of the things that you're expressing. But this is not very typical for compiled languages. And so this was a very strong design point and one of the reasons that Swift in general is designed with this factoring of complexity in mind. So that the language can express powerful things. You can write firmware in Swift if you want to. But it has a very high level feel, which is really this perfect blend. Because often you have very advanced library writers that want to be able to use the, the nitty gritty details.

• 49:17Chris Lattner?

  But then other people just want to use the libraries and work at a higher abstraction level.

• 49:22Lex Fridman?

  It's kind of cool that I saw that you can just interparability. I don't think I pronounced that word enough, but you can just drag in Python.

• 49:32Lex Fridman?

  It's just strange. You can import like, I saw this in the demo, import number. Like, how do you make that happen?

• 49:37Chris Lattner?

  Yeah, well.

• 49:38Lex Fridman?

  What's, what's up with, is that as easy as it looks or is it?

• 49:42Chris Lattner?

  Yes, as easy as it looks. That's not, that's not a stage magic hack or anything like that.

• 49:46Lex Fridman?

  No, no, I don't mean from the user perspective. I mean from the implementation perspective to make it happen.

• 49:51Chris Lattner?

  So it's, it's easy. Once all the pieces are in place, the way it works. So if you think about a dynamically typed language like Python, right? You can think about it as in two different ways. You can say it has no types, right? Which is what most people would say, or you can say it has one type. Right? And you could say it has one type and it's like the Python object. Right? And the Python object is passed around and because there's only one type, it's implicit. Okay. And so what happens with Swift and Python talking to each other, Swift has lots of types. It has arrays and it has strings and all like classes and that kind of stuff. But it now has a Python object type. Right? So there is one Python object type. And so when you say import numpy, what you get is a Python object, which is the numpy module.

• 50:37Chris Lattner?

  And then you say np dot array. And it says, okay, hey, hey, Python object. I have no idea what you are. Give me your array member. Right? Okay, cool. And it just, it just uses dynamic stuff, talks to the Python interpreter and says, hey, Python, what's the dot array member in that Python object? It gives you back another Python object. And now you say parentheses for the call and the arguments you're going to pass. And so then it says, hey, a Python object that is the result of np dot array, call with these arguments, right? Again, calling into the Python interpreter to do that work. And so, um, right now, this is all really simple. And if you, if you dive into the code, what you'll see is that the, the Python module in Swift is something like 1,200 lines of code or something. It's written in pure Swift. It's super simple and it's,

• 51:21Chris Lattner?

  and it's built on top of the C interoperability because it just talks to the Python interpreter. But making that possible required us to add two major language features to Swift to be able to express these dynamic calls and the dynamic member lookups. And so what we've done over the last year is we've proposed, implement, standardized and contributed new language features to the Swift language in order to make it so it is really trivial. Right? And this is one of the things about Swift that is

• 51:50Chris Lattner?

  critical to the Swift or TensorFlow work, which is that we can actually add new language features. And the bar for adding those is high, but it's, it's what makes it possible.

• 51:59Lex Fridman?

  So you're now at Google doing incredible work on several things, including TensorFlow. So TensorFlow 2.0 or whatever, leading up to 2.0 has, uh, by default in 2.0 has eager execution. And yet in order to make code optimized for GPU or GPU or some of these systems, computation needs to be converted to a graph. So what's that process like? What are the challenges there?

• 52:26Chris Lattner?

  Yeah. So I, I'm tangentially involved in this, but the, the way that it works with autograph is that You mark your your function with a decorator. And when Python calls it, that decorator is invoked. And then it says, before I call this function, you can transform it. And so the way autograph works is, as far as I understand, is it actually uses the Python parser to go parse that, turn into a syntax tree, and now apply compiler techniques to, again, transform this down into TensorFlow graphs. And so you can think of it as saying, hey, I have an if statement. I'm going to create an if node in the graph, like you say, tf cond.

• 53:08Chris Lattner?

  You have a multiply, well, I'll turn that into a multiply node in the graph, and it becomes this tree transformation.

• 53:14Lex Fridman?

  So where does the Swift for TensorFlow come in? Which is, you know, parallels, you know, for one, Swift is an interface, like Python is an interface, TensorFlow. But it seems like there's a lot more going on in just a different language interface. There's optimization methodology.

• 53:32Chris Lattner?

  Yeah, so so the TensorFlow world has a couple of different, what I'd call, front end technologies. And so Swift and Python and Go and Rust and Julia and all these things share the TensorFlow graphs and all the runtime and everything that's later. Okay. And so Swift for TensorFlow is merely another front end for TensorFlow, just like any of these other systems are. There's a major difference between, I would say, three camps of technologies here. There's Python, which is a special case because the vast majority of the community efforts go into the Python interface. And Python has its own approaches for automatic differentiation, has its own APIs, and all this kind of stuff. There's Swift, which I'll talk about in a second.

• 54:17Chris Lattner?

  And then there's kind of everything else. And so the everything else are effectively language bindings. So they they call into the TensorFlow runtime, but they're not, they usually don't have automatic differentiation, or they usually don't provide anything other than APIs that call the C APIs in TensorFlow. And so they're kind of wrappers for that. Swift is really kind of special, and it's a very different approach. Swift for TensorFlow, that is, is a very different approach because there we're saying, let's look at all the problems that need to be solved in the full stack of.

• 54:49Chris Lattner?

  The TensorFlow compilation process, if you think about it that way, because TensorFlow is fundamentally a compiler. It takes models and then it makes them go fast on hardware. That's what a compiler does. And it has a front end, it has an optimizer and it has many backends. And so if you think about it the right way, or if you look at it in a particular way, like it is a compiler. Okay. And so Swift is merely another front end. But it's saying, and the design principle is saying, let's look at all the problems that we face as machine learning practitioners. And what is the best possible way we can do that? Given the fact that we can change literally anything in this entire stack.

• 55:33Chris Lattner?

  And Python, for example, where the vast majority of the engineering and effort has gone into, is constrained by being the best possible thing you can do with a Python library. Like there are no Python language features that are added because of machine learning that I'm aware of. They added a matrix multiplication operator with that, but that's as close as you get. And so with Swift, it's hard, but you can add language features to the language. And there's a community process for that. And so we look at these things and say, well, what is the right division of labor between the human programmer and the compiler? And Swift has a number of things that shift that balance.

• 56:12Chris Lattner?

  So because it has a type system, for example, it makes certain things possible for analysis of the code. And the compiler can automatically build graphs for you without you thinking about them. Like that's a big deal for a programmer. You just get free performance, you get clustering and fusion and optimization, things like that, without you as a programmer having to manually do it, because the compiler can do it for you. Automatic differentiation is another big deal. And it's, I think, one of the key contributions of the Swift for TensorFlow project. Is that there's this entire body of work on automatic differentiation that dates back to the Fortran days. People doing a tremendous amount of numerical computing in Fortran

• 56:54Chris Lattner?

  used to write these, what they call source to source translators, where you, where you take a bunch of code, shove it into a mini compiler and it would push out more Fortran code, but it would generate the backwards passes for your functions for you, the derivatives. And so in that work in the seventies, a tremendous number of optimizations, a tremendous number of techniques for fixing numerical instability and other, other kinds of problems were developed. But they're very difficult to port into a world where in eager execution, you get an op by op at a time. Like you need to be able to look at an entire function and be able to reason about what's going on. And so when you have a language integrated automatic differentiation,

• 57:36Chris Lattner?

  which is one of the things that the Swift project is focusing on, you can open, open all these techniques and reuse them in familiar ways. But the language integration piece has a bunch of design room in it and it's also complicated.

• 57:50Lex Fridman?

  The other piece of the puzzle here that's kind of interesting is TPUs at Google.

• 57:54Chris Lattner?

  Yes.

• 57:54Lex Fridman?

  So, you know, we're in a new world with deep learning. It's constantly is changing. And I imagine without disclosing anything, I imagine, you know, you're still innovating on the TPU front too.

• 58:05Chris Lattner?

  Indeed.

• 58:06Lex Fridman?

  So how much sort of interplays there between software and hardware in trying to figure out how to gather, move towards an optimized solution?

• 58:13Chris Lattner?

  There's an incredible amount. So we're on a third generation of TPUs, which are now 100 petaflops in a very large liquid cooled box, virtual box with no cover. And as you might imagine, we're not out of ideas yet. The, the great thing about TPUs is that they're a perfect example of hardware software co-design. And so it's about, it's about saying, what hardware do we build to solve certain classes of machine learning problems? Well, the algorithms are changing. Like the hardware takes, you know, some cases years to produce, right? And so you have to make bets and decide what is going to happen. And so, and what is the best way to spend the transistors to get the maximum?

• 58:55Chris Lattner?

  Um, you know, performance per watt or area per cost or like whatever it is that you're optimizing for. And so one of the amazing things about TPUs is this numeric format called B float 16. B float 16 is a compressed 16 bit floating point format, but it puts the bits in different places. And in numeric terms, it has a smaller mantissa and a larger exponent. That means that it's less precise, but it can represent larger ranges of values, which in the machine learning context is really important and useful because sometimes you have very small gradients you want to accumulate and very, very small numbers that are important to move things as you're learning. But sometimes you have very large magnitude numbers as well.

• 59:40Chris Lattner?

  And B float 16 is not as precise, the mantissa is small. But it turns out the machine learning algorithms actually want to generalize. And so there's, you know, theories that this actually increases the ability for the network to generalize across data sets. And regardless of whether it's good or bad, it's much cheaper at the hardware level to implement because the area and time of a multiplier is n squared in the number of bits in the mantissa, but it's linear with size of the exponent.

• 1:00:10Lex Fridman?

  And you're connected to both efforts here, both on the hardware and the software side?

• 1:00:14Chris Lattner?

  Yeah. And so that was a breakthrough coming from the research side and people working on optimizing network transport of weights across the network originally and trying to find ways to compress that. But then it got burned into silicon and it's a key part of what makes TPU performance so amazing and, and, and great. Now, TPUs have many different aspects that are important, but the, the co-design between the low level compiler bits and the software bits and the algorithms is all super important. And it's this amazing trifecta that only Google can do.

• 1:00:50Lex Fridman?

  Yeah, that's super exciting. So can you tell me about MLIR project, previously the secretive one?

• 1:00:58Chris Lattner?

  Yeah. So MLIR is a project that we announced at a compiler conference. Three weeks ago or something at the compilers for machine learning conference. Basically, if again, if you look at TensorFlow as a compiler stack, it has a number of compiler algorithms within it. It also has a number of compilers that get embedded into it and they're made by different vendors. For example, Google has XLA, which is a great compiler system. Nvidia has Tensor RT, Intel has NGraph. There's a number of these different compiler systems and they're very hardware specific and they're trying to solve different parts of the problems. But they're all kind of similar in a sense of they want to integrate with TensorFlow. Now TensorFlow has an optimizer and it has these different code generation technologies

• 1:01:42Chris Lattner?

  built in. The idea of MLIR is to build a common infrastructure to support all these different subsystems. And initially, it's to be able to make it so that they all plug in together and they can share a lot more code and can be reusable. But over time, we hope that the industry will start collaborating and sharing code. And instead of reinventing the same things over and over again, that we can actually foster some of that, that, you know, working together to solve common problem energy that has been useful in the compiler field before. Beyond that, MLIR is, some people have joked that it's kind of LVM2. It learns a lot about what LVM has been good and what LVM has done wrong. And it's a chance to fix that. And also, there are challenges in the LVM ecosystem

• 1:02:26Chris Lattner?

  as well, where LVM is very good at the thing it was designed to do, but you know, 20 years later, the world has changed and people are trying to solve higher level problems and we need, we need some new technology.

• 1:02:37Lex Fridman?

  And what's the future of open source in this context?

• 1:02:41Chris Lattner?

  Very soon. So it is not yet open source, but it will be hopefully.

• 1:02:45Lex Fridman?

  So you still believe in the value of open source in these kinds of.

• 1:02:48Chris Lattner?

  Oh, yeah, absolutely. And I think that the TensorFlow community at large fully believes in open source.

• 1:02:54Lex Fridman?

  So I mean, that's, there is a difference between Apple, where you were previously, and Google now in spirit and culture. And I would say the open sourcing of TensorFlow was a seminal moment in the history of software because here's this large company releasing a very large code base. It's open sourcing. What are your thoughts on that? Happy or not, were you to see that kind of degree of open sourcing?

• 1:03:20Chris Lattner?

  So between the two, I prefer the Google approach, if that's what you're saying.

• 1:03:25Chris Lattner?

  The Apple approach makes sense, given the historical context that Apple came from, but that's been 35 years ago. And I think that Apple is definitely adapting. And the way I look at it is that there's different kinds of concerns in this space, right? It is very rational for a business to care about making money. That fundamentally is what a business is about, right? But I think it's also incredibly realistic to say, it's not your string library that's the thing that's going to make you money. It's going to be the amazing UI product differentiating features and other things like that that you build on top of your string library. And so keeping your string library proprietary and secret and things like that isn't maybe

• 1:04:08Chris Lattner?

  not the important thing anymore, right? Where before, platforms were different, right? And even 15 years ago, things were a little bit different. But the world is changing. So Google strikes a very good balance, I think. And I think that TensorFlow being open source really changed the entire machine learning field and it caused a revolution in its own right. And so I think it's amazing for amazingly forward looking because I could have imagined, and I wasn't at Google at the time, but I could imagine a different context and a different world where a company says, machine learning is critical to what we're doing. We're not going to give it to other people, right? And so that decision is a profound, profoundly brilliant insight that I think has really

• 1:04:53Chris Lattner?

  led to the world being better and better for Google as well.

• 1:04:56Lex Fridman?

  And has all kinds of ripple effects. I think it is really, I mean, you can't understate Google deciding that, how profound that is for software is awesome.

• 1:05:07Chris Lattner?

  Well, and it's been, and again, I can understand the concern about if we release our machine learning software, our competitors could go faster. But on the other hand, I think that open sourcing TensorFlow has been fantastic for Google. And I'm sure that obvious was that that decision was very non-obvious at the time, but I think it's worked out very well.

• 1:05:28Lex Fridman?

  So let's try this real quick.

• 1:05:30Chris Lattner?

  Yeah.

• 1:05:30Lex Fridman?

  You were at Tesla for five months as the VP of Autopilot Software. You led the team during the transition from H hardware one to hardware two. I have a couple of questions. So one, first of all, to me, that's one of the bravest engineering decisions undertaking. So like undertaking really ever in the automotive industry to me, software wise, starting from scratch. It's a really brave engineering decision. So my one question is there is what was that like? What was the challenge of that?

• 1:06:01Chris Lattner?

  Do you mean the career decision of jumping from a comfortable, good job into the unknown? Or?

• 1:06:06Lex Fridman?

  That combined, so the, at the individual level, you making that decision. And then when you show up, you know, it's a really hard engineering problem.

• 1:06:15Chris Lattner?

  Yes.

• 1:06:15Lex Fridman?

  So you could just stay, maybe slow down, say hardware one, or that those kinds of decisions. Just taking it full on, let's, let's do this from scratch. What was that like?

• 1:06:28Chris Lattner?

  Well, so, I mean, I don't think Tesla has a culture of taking things slow and seeing how it goes. So, and one of the things that attracted me about Tesla is it's very much a gung ho, let's change the world. Let's figure it out kind of a place. And so I have a huge amount of respect for that. Tesla has done very smart things with hardware one in particular. And the hardware one design was originally designed to be very simple automation features in the car for like traffic aware cruise control and things like that. And the fact that they were able to effectively feature creep it into lane holding and, and a very useful driver assistance feature is, is pretty astounding, particularly given the details of the hardware. Hardware two built on that a lot of ways.

• 1:07:11Chris Lattner?

  And the challenge there was that they were transitioning from a third party provided vision stack to an in-house built vision stack. And so for the first step, which I mostly helped with was getting onto that new vision stack. And that was very challenging. And there, it was time critical for various reasons and it was a big leap, but it was fortunate that it built on a lot of the knowledge and expertise and the team that had built hardware ones driver assistance features.

• 1:07:40Lex Fridman?

  So you spoke in a collected and kind way about your time at Tesla, but it was ultimately not a good fit. Elon Musk, we've talked on this podcast, several guests at the course. Elon Musk continues to do some of the most bold and innovative engineering work in the world, at times at the cost to some of the members of the Tesla team. What did you learn about this, working in this chaotic world with Elon?

• 1:08:03Chris Lattner?

  Yeah, so I guess I would say that when I was at Tesla, I experienced and saw the highest degree of turnover I'd ever seen in the company, which was a bit of a shock. But one of the things I learned and I came to respect is that Elon's able to attract amazing talent because he has a very clear vision of the future and he can get people to buy into it because they want that future to happen. Right? And the power of vision is something that I have a tremendous amount of respect for. And I think that Elon is fairly singular in the world in terms of the things he's able to get people to believe in. And it's, it's a very, it's very, there may be people that stand on the street corner and say, ah, we're going to go to Mars, right?

• 1:08:46Chris Lattner?

  But then, but then there are a few people that can get other, others to buy into it and believe in, build the path and make it happen. And so I respect that. I don't respect all of his methods, but, but I have a huge amount of respect for that.

• 1:09:02Lex Fridman?

  You've mentioned in a few places, including in this context, working hard. What does it mean to work hard? And when you look back at your life, what were some of the most brutal periods of having to really sort of put everything you have into something?

• 1:09:20Chris Lattner?

  Yeah, good question. So working hard can be defined a lot of different ways. So a lot of hours. And so that's, that is true. The thing to me that's the hardest is both being short-term focused on delivering and executing and making a thing happen, while also thinking about the longer term and trying to balance that, right? Because if you are myopically focused on solving a task and getting that done, and only think about that incremental next step, you will miss the next big hill you should jump over to, right? And so I've been really fortunate that I've been able to kind of oscillate between the two and

• 1:10:01Chris Lattner?

  Historically at Apple, for example, that was made possible because I was able to work with some really amazing people and build up teams and leadership structures and allow them to grow in their careers and take on responsibility, thereby freeing up me to be a little bit crazy and thinking about the next thing. And so it's a lot of that. But it's also about, you know, with experience, you make connections that other people don't necessarily make. And so I think that is, that's a big part as well. But the bedrock is just a lot of hours. And, you know, that's, that's okay with me. There's different theories on work-life balance. And my theory for myself, which I do not project onto the team, but my theory for myself is that, you

• 1:10:45Chris Lattner?

  know, I, I want to love what I'm doing and work really hard. And my purpose, I feel like, my goal is to change the world and make it a better place. And that's, that's what I'm really motivated to do.

• 1:10:57Lex Fridman?

  So, last question. L L V M logo is a dragon.

• 1:11:01Chris Lattner?

  Yeah.

• 1:11:01Lex Fridman?

  You explain that this is because dragons have connotations of power, speed, intelligence. It can also be sleek, elegant, and modular, though you removed the modular part. What is your favorite dragon-related character from fiction, video, or movies?

• 1:11:18Chris Lattner?

  So those are all very kind ways of explaining it. Do you want to know the real reason it's a dragon?

• 1:11:22Lex Fridman?

  Well, yeah.

• 1:11:24Chris Lattner?

  Is that better? So there's a seminal book on compiler design called The Dragon Book. And so this is a really old now, book on compilers. And so the dragon logo for L L V M came about because at Apple, we kept talking about L L V M-related technologies and there's no logo to put on a slide. And so we're like, what, what do we do? And somebody's like, well, what kind of logo should a compiler technology have? And I'm like, I don't know. I mean, the dragons are, the dragon is the best thing that, that we've got. And You know, Apple somehow magically came up with the logo and it was a great thing and the whole community rallied around it and then it got better as other graphic designers got involved, but that's originally where it came from.

• 1:12:06Chris Lattner?

  Story. Is there dragons from fiction that you connect with? That Game of Thrones, Lord of the Rings, that kind of thing? Lord of the Rings is great. I also like role playing games and things like computer role playing games and so dragons often show up in there, but it really comes back to to To the book. Oh, no, we need, we need a thing. Yeah. And hilariously, one of the, one of the funny things about LVM is that my wife, who's amazing, runs the LVM Foundation and she goes to Grace Hopper and is trying to get more women involved in the, she's also compiler engineer, so she's trying to get other, other women to get interested in compilers and things like this. And so she hands out the stickers and people like the LVM sticker because of Game of Thrones.

• 1:12:51Chris Lattner?

  And so sometimes culture has this helpful effect to like get the next generation of compiler engineers engaged with the cause. Okay, awesome. Chris, thanks so much for talking. It's been great talking with you.