Algorithms + Data Structures = Programs - Episode 62: 2021 Retro (Part 2)

Episode Date: January 28, 2022

In this episode, Bryce and Conor talk about the highlights of 2021 (part 2).Date Recorded: 2022-01-08Date Released: 2022-01-28RAPIDS decimal128 Github Pull RequestProgramming Languages Virtual MeetupCategory Theory for Progrmmers by Bartosz MilewskiConor’s other podcast: ArrayCastNorth Devner C++ MeetupC++ Russia ConferenceACCU Belfast ConferenceMeeting C++ ConferenceSuperComputing Conference“The Future of Conferences” by Cristina (Crista) Lopes at StrangeLoop 2021C++Now ConferenceReady Player One by Ernest ClineNVIDIA’s OmniverseSnow Crash by Neal StephensonNVIDIA Isaac Sim (Robot Simulation)RAPIDS AINVIDIA Thrust Parallel AlgorithmsAcutarial ScienceBernoulli DistributionBinomial DistributionWeibull DistributionC++23CppNorth Call For PapersAndrei Alexandrescu on TwitterBartosz Milewski on TwitterBetter Algorithm Intuition - Conor Hoekstra - Meeting C++ 2019Pactifc Northwest C++ Users’ GroupEric Niebler on TwitterC++ RangesIntro Song InfoMiss You by Sarah Jansen https://soundcloud.com/sarahjansenmusicCreative Commons — Attribution 3.0 Unported — CC BY 3.0Free Download / Stream: http://bit.ly/l-miss-youMusic promoted by Audio Library https://youtu.be/iYYxnasvfx8

I have done more matrix multiplying than you. That does not make you a machine learning. Why is this always a contest? Last time it was the blenders. This time I mentioned matrix multiply. Bryce's response, Connor, I'd just like you to know that I have done more matrix multiplication than you have in your lifetime. welcome to adsp the podcast episode 62 recorded on january 8th 2022 my name is connor and today with my co-host bryce we continue and finish part two of our 2021 retro. Yeah, well, we haven't, we've done my 2021 retrospective.

What about your 2021 retrospective? Mine is, mine's mostly outside of work. So my work highlight, I worked on a really big project on the Rapids ecosystem called basically Decimal 128. I'm not going to go into the details, but it's like fixed point arithmetic.

And it was a pretty big deal because there was compiler work that was done in order to support a bunch of work, a bunch of work that ended up that my team ended up having to do. Yeah. Yeah. Thanks. A 128 bit integers work. So thanks.

Thanks to Wesley on my team for making that happen. Yeah. There was compiler folks. There was sort of on the standard library side, a bunch of stuff. So yeah, that was an awesome project. But my top three, starting number three, there was a trip I took in August that went to both Vancouver and then also to Winnipeg, I believe, to visit some friends and the family. That was the first time I'd been on a plane in over a year and a half. And more importantly, the first time I'd seen my family. So that was a big

deal. Yeah, obviously, because of the pandemic, I basically didn't want to hop on a plane until I had my vaccines. Anyway, so that was awesome. Number two, what did I write down? Oh, yeah, both the podcasts and my book meetup. We finished category theory for programmers like this is like podcasting is like my substitute for conferences because you don't get i was just at like the the denver c++ meetup a couple days ago virtually he was not at person at the denver c++ meetup yeah virtually and um yeah man i just it's it's a very privileged statement but i just i miss i miss the conferences so much like you don't get to see like my my like socializing like my my most happy favorite socializing is like at those conferences well you know it's one of the things that i realized a few months into the

pandemic is and a couple months into the pandemic this was something that like I thought was a fear. It was something that was a fear of mine, but that I now recognize back at the at the you know 2011 to 2019 as this like golden age of tech conferences and this golden age of c++ conferences and that things never really went back to the way they were and i'm now fairly convinced that that's that that's the case so like you know like you and i it's sort of like peaked in 2019. You know, you and I both went. Where I went to the, that was my first conference. So I basically got the tail end of the golden age. Well, like you remember like November 2019, I went for like four weeks. I went to like, no, it was like six weeks.

I went to a committee meeting and then Belfast. Yeah, it was the Belfast committee meeting. Then there was ACCU Belfast. And then there was also C++ Russia. And then there was meeting C++. And then I, and then like, like halfway through the trip, I had to like rebook to fly to supercomputing because one of my VPs was like, you need to go to supercomputing um and so i had just like five or six weeks of back-to-back conferences um and it was just in in like that that summer too i was in prague and then like paris and then uh we were in germany for the committee meeting um and like earlier in that year there was a ton of travel too and like there was it was just that, that fall I spent like eight weeks at conferences.

Um, and then like COVID hit and then, uh, you know, it was all over. And I just like, I, I, I doubt that I'll, I'm sure I'll go to conferences again in the future. I, I sort of doubt that there will ever again be a situation where we'll have, you know, six or eight weeks of conferences back to back like that. Yeah, that's the thing is that I completely agree. There was a talk given at Strange Loop 2021 in October. I believe her name was Krista.

Actually, let me get her name because it's in a spreadsheet that's right here. Krista Lopes, Christina Lopes, although she goes by Krista Lopes. Wait, when you say spreadsheet, is it the spreadsheet? Yes. Portfolio of words on the videos tab. I watched that talk on December 30th, 2021. And yeah, she gave a talk. It was a keynote actually called the future of conferences. And it's a fantastic, it sounds like, oh, that's not going to be very exciting. It was one of my, it's a very good talk and talks about how there's trade-offs in that it increases the accessibility, reduces the, you know, carbon footprint if, you know, things are happening virtually, but there's also a ton of pros to, you know, having them in

person. And so I do think there still will be in-person conferences that are awesome, but yeah, it's, it's never going to go back to the way it was, I don't think. And that's the thing is I say it's like it's the best socializing or my favorite socializing. But like it's more than that. Like the intellectual growth or like the ideas that you come away with. And like that's the thing is this, you know, the Denver C++ meetup, it was great. We did this live code review. But then there were sort of two hours of casual chatting. Inevitably, it was like 80 minutes of that 120 minutes was just talking about people that know

people that got the virus and COVID. And it's like, it's not really technical in nature. Not that the conversation has to be technical, but like when're at C++ now, 80% of what you're talking about is just like you're nerding out on nerd stuff. You're not talking about, oh, I just got the virus or I'm going to get my booster vaccine. And so it's both socializing, but I learned so much every single time I was at some conference. And also, too, giving a talk, you always get a bunch of people afterwards that come up and are excited and want to talk about like, that doesn't happen as much at virtual conferences. Yeah, I mean, there's there's an atmosphere and a chemistry to, you know, in person versus, you know, virtual. I think like, like, I think it is a reality that there's like a loss of chemistry when things are, you know, solely virtual. And I don't think that that's something that you can ever really fully replace.

We'll see. We'll see with, you know, Ready Player 1, 2, 3, 4, Haptic Suits. Not that I want a Haptic Suit to go to a conference. As good as the metaverse is for our stock price. Omniverse. Not the metaverse. Omniverse. I'm very sorry, Jensen.

Very sorry, Jensen. We were first. We were first. We were first. I mean, metaverse, isn't that from like the snow catcher, piercer? I don't know where they got it from. But anyways, as good as Omniverse has been for the company's stock price. You know, I think it's, it's one of those technologies that, that is important and foundational and will be transformative, but because it's important and

transformational, it will be overhyped. Like I the thing about the Walmart VR shopping experience. It's like the whole idea of grocery delivery is that you don't have to walk through the store to pick out the groceries. You're paying a premium to just have a better shopping experience than walking through an actual storefront and find things. And so that's just clearly a wrong use of the technology. I do think that Omniverse is actually very cool. One of my Nvidia colleagues who's in the city works on Omniverse, and I've chatted

about it a couple times, and I'm always very impressed with him. If you go and look at some of Jensen's GTC keynotes or some of the talks about Omniverse before, you know, Facebook made its, like, metaverse announcement, so, like, if you go look at, look at some of the talks from 2019, early 2020, I think that really does a good job of showcasing some of the use cases that can be really transformative. One of the things that I always thought was very clever was this idea that if you look at the self-driving car problem, to get the amount of data that you would need to really like train self-driving car software sufficiently to be, you know, road safe, you would need need orders of magnitude more hours on the actual road than you have today. And so it may just not be practical to go and gather that data, just like driving around on the road. And so one of the ideas of the Omniverse was, okay, well, what if we built a simulation environment where we could, you know, simulate the real world and then train the neural nets in that simulated environment?

And not just for self-driving cars. I think the actual use case that was shown in one of the keynotes was doing this for robotics of like, you know, training the robots in this virtual simulated reality. And that's just like, when I saw that, that's really when it clicked for me. And I just like, you know, I saw that and I was like, wow, this is really cool. And I think it's potential. And one of the reasons is that, you know, before I worked at NVIDIA, one of the things I did was computational science, which is more or less the same idea. You know, I worked on doing these simulations of white dwarf mergers. And one of the reasons why we want to do those simulations is we want to understand, we want to get more data than we can possibly get today because there are these very rare events

that happen very, very far away. And so one of the things that we do with these simulations is we want to run these simulations with our understanding of how the physics of them work and then compare the simulated results to actual observational results to make sure that the simulation actually reflects reality. And that's become the basis of a ton of science today.

I think that you could argue in some ways that a lot of HPC is in some ways some form of the omniverse. Now, the big difference with something like omniverse is that it's visual. For something like training self-driving cars or a robot that needs to see things, you need an actual visual, real-time environment. Yeah, I'll try and find a couple of the videos that show the robot simulation training. Because, yeah, there's a couple ones. There's one, like a golf one. There's like a mini robot and sort of a factory thing. And, yeah, it's very mind-op opening watching it and being like oh yeah that's

a thing yeah you know it's it's funny with things like machine learning and omniverse um in data science um you know like i think i feel like oftentimes some of us who are working in or adjacent to those fields, because there's so much hype around it, we sometimes have like a pessimistic outlook. It's like, oh, it's like, you know, everybody's excited about machine learning. Everybody thinks that machine learning is going to solve all their problems. Every company's got, you know, a data science and a machine learning division now. And, you know, it's a gold rush and, and, you know, everybody's overhyped, but like every now and then you, you get reminded of the fact that like the reason that these things are overhyped is because they are actually amazing transformational technologies, which are having a huge impact on the world.

Yeah. Yeah. Yeah, you say that, and it makes me think about the fact that technically RAPID sits under the AI infrastructure. And technically, I could go around saying that, yeah, I work on artificial intelligence infrastructure, which is like it's not false, but I never say that. Like when people ask me what I do, I just, you know, say, I work on the C++ backend for a accelerated Python library. Because I just feel like such a, like saying that I work on AI infrastructure. Like I know that there's an equivalent team at Facebook and they have jackets that say AI infra on them.

Like, you know, Patagonia jackets. And I just like, I just can't do it because I just like, I know that the semantic attachment that people have to AI, it's like, you know, the Hollywood movies and stuff. And I'm just like, it's not that sexy what I do. Like, it's awesome. It's really cool. But it's not, it's not what people think. Well, I think one of the reasons is that to some degree, the AI, or I really prefer to say machine learning, because I think AI is much more

of a public-facing term. Like, AI is a very broad field, and the majority of the AI that we're doing today is machine learning. So, you know, I like, you know, often say to people, like, you know, I know nothing about ML, which is not entirely untrue. And, like, I certainly wouldn't, like, say that, like, I work on, like, you know, core infrastructure for ML and AI, although that is kind of, like, true. But I think one of the reasons that there's perhaps some hesitance there is that it is true that the ML gold rush has,

um, I'm thinking about how to say this nicely, has the ML Gold Rush has shifted priorities in ways that have left certain, certain domains of computational science and software engineering feeling like they've sort of been abandoned at the prom for another date. You know, like traditional HPC, traditional compute has sort of taken a backseat to the needs of ML and DL over the past few years. And certainly from like the HPC perspective, you know, a lot of the hardware innovation that gets done has been around accelerating ML. Now, a lot of that hardware acceleration can be leveraged by HPC

and a lot of traditional HPC is starting to use machine learning techniques. But I think it's a little bit natural for there to be some reaction of like, oh, that's like, oh, that's nonsense. And it's just shifting priorities away from like the real, you know, the real science or the real thing that we're doing. But I will say I am always proud to say that I work on enabling data science frameworks. Because, I don't know, something about data science feels more noble to me.

Yeah, I mean, it is weird to think. Didn't Jensen, wasn't he man of the year, time of the year person or whatever? Person of the year? And the whole profile was NVIDIA is the company leading the AI, whatever, frontier, something like that. So to me, one of the reasons why I'm happy to say that I work with people that build data science frameworks and have this gut reaction that data science is more noble is that I think a lot of people pick up ML and a lot of people and a lot of organizations have picked up ML and have like, sort of treated it as like a hammer. Um, like, oh yeah, we'll just like, we'll just throw some machine learning into like our thing and it'll be great. Um, and the reality is that like, not everybody necessarily has the amount of data, um, to, to, to like, you know, to do that. If it fits in a spreadsheet, guess what? It's not big data.

But, um, but you know, data science is really just this field like data science is just like the field of using like techniques and algorithms and systems to get insights out of data, you know, potentially noisy data. So like, it's not necessarily, you know, using machine learning. It's just like, we have data, like, you know, how do we extract knowledge out of that data? And that to me, like, that's the sort of hammer that I think is one that almost anybody can pick up and get something useful out of. Like if you have data, you know, almost any quantity of data, yeah, you can probably use data science to learn something from that data. Yeah. All right. Well, we just went on a huge, we went from podcasts and my meetup is number two to I

miss conferences to Omniverse to we work at nvidia and we know ai except we don't but yeah we do we do kind of know data science we know people that know stuff yeah we know we do know people that know stuff yeah mostly i just use thrust parallel algorithms and yeah we also don't know graphics like like the group of like engine like all the nvidia c++ people with a couple exceptions most of the nvidia c++ people don't know graphics and don't really know machine learning which is funny because like if you meet it like those are the two things that nvidia is kind of known for i do i mean i don't i don't know machine learning but I did do a lot

of matrix multiplication when I was studying to be an actuary. So I hear that's like 80% of it. So I have done more matrix multiplying than you. It does not make you a machine learner. Last time it was the blenders. This time I mentioned matrix multiply Bryce's response, Connor, I just liked you to know that I have done more matrix multiplication than you have in your lifetime. I have written optimized matrix multiplication algorithms. I feel like why are you saying you don't know machine learning then? That's like I've heard that's like basically all it is.

Like that's that's linear algebra. That's a different field. Like machine learning is like a specific field of like, of like applying these techniques. I think I probably took in, in my undergrad, my, my, my math education at LSU, the department was very, very theory focused, or at least the program I was in was very theory focused. It was called applied math, but it was very focused on like graph theory. It was not, not really a very computational program, which is actually why I liked it. I think I've said this on the podcast before. I did not get a computer science degree because at the time I knew I was a little shit and like a little arrogant shit. And I thought that if I was going to be in a computer science

degree, I would think that I was smarter than my professors and I would clash with them. And then I would be miserable. And also I hated school. And so I figured like if the thing that I loved, which is programming and computer science and software engineering, if that was like what I had to like study in school, that I was just going to be miserable and it would make me hate those things. So that's why I ended up getting a math degree. And I like intentionally picked the math course and the math degree program at lsu that was like the least

um it wasn't the least computational but it was like the second least computational which was the applied math program um uh it was like the one where you did not have to take any cs courses um like i intentionally picked that one um uh and uh i I don't know that we ever took any machine learning or neural net courses at the time because that was like 2011, 2014 and you know it wasn't as prominent as it is now there was definitely a lot of like applicable graph theory that I learned that was you know applicable to

machine learning techniques that are used today although most of that knowledge has drained out of my head there are still like a handful of like graph proofs that I had to like learn for like some test that I'm fairly certain that if under pressure and like presented with like the need to produce these proofs, I could probably like crank out like one of these like three page ridiculous proofs. Sounds like a challenge. Yeah. I also got, I, I, I, you know, it's funny.

Cause I was, I was somebody that was in, you know, HC, but I barely survived my, like, partial differential equations coursework. Like, that, like, differential equations and calculus never came naturally to me. Like, graph theory, you know, like, graph theory and, like, discrete math math um like that was all like very natural to me but like the calculus nope nope nope nope nope nope yeah i mean calculus calculus one and two um was fine but as soon as it hit multivariate and three i just yeah i hated the word problems um and i used to love word problems but I just remember there would be like some cone and water and water leaving at a certain rate and water being poured into the cone

and there was my problem was that like once you get to like partial differential equations like you to like solve these problems like you just had to have an intuition about like how to do it like literally my professors would tell me, like you just had to have an intuition about like how to do it.

Like literally my professors would tell me that, like you just have to have the right intuition about like how to approach this. And like, I don't know, I just lacked that intuition. Yeah. I never took PDE or ODE. I still have like, I have right here on this shelf, like my, these actually aren't even my coursework PDE books. These were like the graduate level PDE books that I had to pick up to like learn how to do my job. Like to learn the math that I needed to do the computational astrophysics and other stuff that I was doing.

All right. Well, enough with boring our listeners with, or boring the listener, sorry, with our mathematics, not even successes. It sounds like we both struggle towards the end. I mean, I think it's a good thing for people to know. Like, you know, like I, I think, I think I'm good at math. I actually think that I'm strong, my, my like, my verbal and my like communication skills are stronger. Like I remember when I took the SATs, which are like the, the college like entrance exams in the US.

I got like, like the math section was like always like hard for me, which was weird because like, I was like a science kid, you know, like, like, like my parents were like convinced I was going to like get, get a STEM degree and whatnot. Like that was always what interested me. Um, but like the math, like, um, just because I liked science doesn't mean it didn't mean I was good at math, which you need to be to like go far in science. But like the math section I struggled with i ended up doing pretty good but the like verbal section of the like sats like i didn't have to study for and i got like a perfect score um wow we're like opposites uh yeah yeah yeah but like like people don't expect that from like me um when they meet me they're like oh like you're a software engineer you have a math

degree like you know writing can't be your strong suit, but, like, the reality is, like, writing and communicating, like, that is, like, just natural to me, whereas math is something that I had to struggle with, and to some degree, that's why I chose to get a math degree, you know, as part of it was that sort of, like, reason, like, I didn of like reason, like, I didn't want to like, I didn't want to, um, get a CS degree because I thought I would argue with my professors. And cause I, I'd never enjoyed school and I enjoyed programming and I didn't want to like combine, I didn't want to come to hate programming through studying it in the school. But like the other reason was like, I thought about like, you know, I know how to program

right now. I don't think I need to go to school to learn that. But the thing that I don't know is math. And so I'm going to go to get a degree in math to hopefully, you know, learn more about it. So I guess that's maybe not something that a lot of people think to do. Like a lot of people think like I'm going to go to college in the field that I'm good at. I consciously chose to get a degree in the field that I'm bad at so that I could learn more about it. This is actually, man, it's, we, we have these back and forths on, uh,

how similar we are except for furnishing our apartments, which is where we really differ. But it's actually, that's crazy that we've talked about university quite a bit on this podcast or like our past experiences, but I basically had the exact same decision. Like, so I, I always had a, whatever predisposition towards math. But at one point in my third semester, I had taken two computer science courses, the two intro ones. And because they had projects with bonus marks and stuff, I got a hundred percent above a hundred percent, um, in both of those classes. And I loved it. And I, I was already working towards the actuarial math. And so I had a choice. And at the time I, and I consider this a mistake at the time, I didn't know that like computer science, I didn't know anything about software engineering and Google

and stuff. And that it was like a, you know, growing industry. My view was just that like, Oh, computer science is this super fun thing. I love coding in Python, but I decided to continue with the actuarial math thing. Cause I figured I was going to get bored with computer science is this super fun thing. I love coding in Python, but I decided to continue with the actuarial math thing because I figured I was going to get bored with computer science because it was just so easy. At least, I mean, that was a naive, ignorant view because I didn't know, you know, that it's going to get harder at the 200, 300, 400 level. I just figured I was going to continue to be able to do well. Whereas I knew actuarial math, even though I'm good at math, it was an extremely competitive and difficult, like the introductory actuarial course that you have to take in order to get into the program. I think I got like a B minus or a C plus and it was a very, very,

actually, no, I got higher than a C plus because you needed a certain minimum grade. But like I had studied harder for that course than any other course up to that point in university and done very, very poorly. And the course had like a 50% dropout rate. So the point being is I was like, oh, I'll do this. It'll be more challenging. In hindsight, I think it was a mistake. But it's just interesting that we both made the same decision to try something that was more challenging. I actually – I'm going to argue that I don't think that it was a mistake, Connor. And, and I mean, you know,

I think this is probably not applicable to every field. There's probably other fields outside of tech, you know, maybe even in tech where you really do need to have, you know, a degree in a particular field to get a job in that field. But, you know, in tech, like, in tech, you don't need to have a particular degree type to become a software engineer. That's not true for civil engineering. If you want to become a civil engineer, you need to go get a civil engineering degree. Ditto for other, if you want to be an accountant or a lawyer, you need to go to law school, you need to get an accounting degree. But like, I think that if you're planning on going into a field, or even if like, you don't know what field you're going is something to be said for seeking out the courses and the education and the degree program that does not come naturally to you and that challenges you, but that is like something that

you think that you'll need in your career. So like, you know, like I think that it probably has helped you more than you think about it, because I think that those like actuarial skills that you have, they give you this, this background in like math and statistics and, and to some degree, like data science that sort of like make you like unique as a software engineer and give you like a unique perspective on things. So like, I think, I think that you, I think that you would not have as diverse a skillset if you had switched programs. switch programs i think you're definitely correct to an extent and i agree that like my like i consider myself a statistician because i took so many stats courses and you know actual math you have to memorize all these distributions bernoulli binomial whatever weibel etc or you don't have to

memorize them but you need to know how to use them. And so I learned a ton, but the one regret is that taking that path, I spent, and this is not an exaggeration, thousands of hours studying for actuarial exams because there's eight of them. You fail some of them sometimes. I failed, I think, three, four, maybe five.

And each time you have to study, you know, four or 500 hours. I spent so much time spinning my wheels on these exams. And they're terribly designed exams. They're not focused on like, you know, how much you comprehend, they give you 2000 pages of material, and then they only test 10% of it. So it's just like, it's luck of the draw half the time. And I just think that there was like, there was a number of years that I was working towards something that like, as soon as I finished it, I switched. Although I will say, out of that, arguably, one of the most valuable things that I got was like an insane work ethic or I shouldn't use the word insane. Just like a very, very, my, my, my,

it's the knowledge that like I can work 40 hours a week or 50 hours a week. And then on top of that also study another 30 or 40 hours. And like, like back when I studied, I used to wake up at like 6am and I would track 15 minute intervals, like a consultant and like down to like showering, eating, exercise, like everything had a category in order to maximize like how much time I was studying. So anyways, you're not wrong. But there's a few thousand hours I could get back. Yeah.

If I wasn't doing that. All right. We got to wrap. I think we're past the hour mark on episode 61. I have no idea how I'm going to cut this up, and I haven't even gotten to the number one thing of 2021, which very quickly, and we're deferring 2022 outlook to another episode

because we've run out of time, is my running. Running was definitely the highlight of the, what was probably, I'd say 2021 was the worst year of my running. Running was definitely the highlight of the what was probably, I'd say 2021 was the worst year of my life. Why was it the worst year of your life? I mean, you know.

Okay, yeah. A lot of stuff happened. Plus too, 2021 was the first year in I don't know how many years that I haven't left the country. The last time, yeah, February 2020 was the last time I went to a different country. I was in New York.

I'm trying to get Strava to tell me how many miles or how many kilometers I biked last year. Have you biked more than I have run? That is the question. Well, so the problem is it has year to date, which is not applicable for last year. And it has all time. So I don't know whether I can actually get out the information that I would want. And I haven't really biked since like October because it got cold. But I started biking in like November, 2020, or I, I, I, I biked when I was at LSU and then like, I took like a two or three year when I moved to California, I like stopped

biking for three or four years. Um, and then I started again in like late 2020. Um, uh, I think Ryan, if you're listening, you you're like you and i like went on a bike ride together like in like november 2020 and that was like what motivated me to get back into biking so you you get credit for that but um but since then uh strava says i've ridden uh 4156 kilometers um this is pretty good this is from when to when this is from uh november 2020 to now oh damn so it depends how much did you bike in november and december of 2020 i don't think it was like that much um i think it was maybe, hang on, we'll see if I can find that. The numbers are going to be super close.

But this does not count because this does not count everything. Because when I spin indoors, that's not counted here, which is all that I've been doing recently. Okay, so you've definitely biked all that I've been doing recently. Okay. So you've definitely biked more than I was going to say. My 2021 stats are I ran 3,890 kilometers. Okay. That makes me feel bad, though, because I should have done a lot more.

A lot more. You still beat me. You still beat me. Anyways, 2022, episode 62. How much are you running a day normally? It depends. I mean, I think my best month was 512 kilometers in August. So that works out to like 2018 or something, 17.

I don't know, 13. I don't see all that. Oh, wait, wait. But do you run every single day uh basically in august basically i probably took off one or two days in august yeah okay so so i i do yoga with my mom uh or yoga do a 40 K once or twice a week. Um, but, but I, I haven't been biking since October, um, outside cause it's been cold and I definitely had some periods where I was not biking. Well, you know, towards the start of the year, I wasn't, I was

biking like 18 kilometers. It was, I only started doing like 25 K a day and I only started doing 40 Ks like in the spring. So probably like, like probably next year, my, I will probably bike a lot more than, uh, than you. I'll have to figure out a way to get the data for how long for, for my, um, my biking, my, my, my spin, uh, bike, which I'm going to count towards my, my kilometers. Yeah. Well, we'll see if we can do a contest. We'll see who, uh, who,

if you bike more than I run. I feel like I should probably, I feel like it probably should be, you know, like two times. If it's like about the same amount of exercise, I should probably be aiming to like twice the distance that you're running. Okay, we can do that. We can have a bet. How fast do you run normally?

It depends. So when I'm like yesterday, I went for – yesterday was – well, let's figure out. It was minus 8 but felt like minus 16. Minus 16 Celsius in Fahrenheit, which for all of you Americans was three degrees Fahrenheit. Oh my. So it was very cold. I thought it was cold here because there's snow on the ground and it is um it is 26 degrees

which is cold so yeah it was very cold and i went for a 23k run and i ran at a um 4 45 so four minutes and 45 seconds per kilometer so if we change that to miles pace, it's like seven and a half minutes per mile, which is like that's like it's not fast, but it's also not slow. It's just like a decent pace. Yeah. Anyways, we'll wrap up 2021, 2022. It's going to be a better year for sure. Hopefully we'll have in-person conferences. Hopefully Bryce and I will be able to see each other in person at some point.

I'm certain that'll happen. Not so certain about in-person conferences. And I mean, C++23 is going to be finalized this year, which I think is going to be very exciting. And I think you and I both have some exciting... There's various different exciting work projects that we've got going on. We'll preview what we can preview in the next episode of 2022. And then also,

we got to have Dave Abrahams back on to finish up the rest of his story. And we got to have Sean Parent back on. Yeah. I think we got to expand our guest repertoire. Well, so that's actually my last thing was we mentioned CPP North. We talked about BQN.

And my last note is bringing on other PL people. So I think because we're on our quest to become number one in Slovenia. And in order to grow the podcast, I think what we need to do is start bringing someone. We technically already had Dave Abrahams from Swift. But let's bring on someone from Rust, someone from Go. We can even go to JavaScript and Scala and Ruby. We should bring on Andre, of course, someone from Go. We can even go to JavaScript and Scala and Ruby. We should bring on Andre, of course, at some point.

Alexandrescu? Yeah, yeah, yeah. He's one of the creators of D. And he's just an amazing and charismatic storyteller. Oh, yeah. We'd have a good time with him. And we should also bring on Bartush. Maluski? Yeah.

Oh, yeah. That'd be awesome. Why do we should also bring on Bartosz. Maluski? Yeah. Oh, yeah. That'd be awesome. Why do you want to bring on Bartosz? Same thing. Just like I think he has the same quality like Sean and Andre. He's going to have – he's a great storyteller. He's just a great storyteller. And then – well, wait.

So there was like four of them that made up like the Northwest C++ user group. Who was the other horseman? So for those of you that don't know what the heck Bryce is talking about, in my better algorithm intuition talk that I first gave back in 2019 on that little European tour that we both went on, although admittedly yours was a lot longer than mine, I had talked to Eric Niebler about sort of his history. And he talked about this era where they were in the Pacific Northwest C++ user group in Seattle, or maybe it was just

outside Seattle. And he, so himself, Eric, and then three other people would always, you know, after these C++ meetups, they'd go up and they'd go out and, you know, have a couple drinks and sometimes meet together on the weekends. And the four of them would always talk about, like, functional programming and all this awesome, you know, category theory stuff. And then they all sort of went in different directions. So those four people were Eric Niebler, Bartosz Maluski, who went and did category theory for programmers, and then Andrzej Alexandrescu. And the fourth one is Walter Bright, who is the main creator of the D language. And so basically, Walter and Andrzej went and did D. Eric went and did C++ ranges.

And then Bartosz went and did category theory and functional programming. And yeah, we could potentially get them all on at the same time that might be a bit crazy that would be fun that would be fun and uh we gotta have patricia and uh chandler back at some point there's a lot well we obviously at some point we're gonna have titus on this podcast oh yeah yeah like i don't know that titus or the titus doesn't yet but at some point at some point we're gonna have yeah we also have to have Allison. We've got to do the thrust. We've got to have Allison on.

Although I've since discovered that the thrust CUDA backend has an optimized partition path, and it's pretty clever. I haven't fully figured out how it works yet. And also I think I've determined that you need to have it done at least two. Well, no, you can't do it in one pass, but it's... Let's save it. Let's save it for the future episode where we have a dedicated time to talk about this. Anyways, lots of guests.

Yep. Happy New Year. Happy New Year to everybody. And we will preview 2022 next time. Thanks for listening. We hope you enjoyed and have a great day.