Wrapping Up

[00:00:00]
>> Richard Feldman: So this concludes the workshop.
>> [APPLAUSE]
>> Richard Feldman: We just have one more section, just a little bit of like, if you wanna do a little further reading type of stuff, but congratulations, you made it. If there's one thing that you take away from this course, I think this is maybe the most conceptually useful one.

[00:00:16]
Is just that memory is basically a gigantic array of bytes. They're all these ones and zeros, and you can choose to interpret them however you want. As the C programmers or other low level languages, not only can you interpret these however you want, but as we've seen, there are actually specific benefits.

[00:00:32]
Especially when it comes to performance of making use of that understanding of the gigantic array of bytes, concept of memory. Granted, we are doing manual memory management there's the malloc and free and memory on the stack and buffer overruns and reading into garbage memory and all this initialized stuff.

[00:00:47]
There's a lot of performance to be had, but also a ton a ton of foot guns. If you've ever thought pointers were hard, hopefully the fact that we just talked about memory addresses and integers a lot [LAUGH] will make that at least a little bit easier. But if anyone ever says pointer, and you're like, what's a pointer?

[00:01:04]
It's a memory address. It's an integer that refers to where that gigantic array of bytes are we. We talked about what is C, what is it used for? It's one of the top 10 most popular programming languages, and now you hopefully have a more visceral sense of why, if you really, really want build an operating system where performance is really critical.

[00:01:23]
A database, a programming language compiler, or a programming language runtime, all of these things are things that can benefit from having all of the safety rails removed so that you can just make things as optimized as fast as you possibly can. We talked about where C comes from, [LAUGH] back in the PDP-11 days, just one step removed from punch cards, working on a teletype, which is a big ergonomics improvement compared to punch cards.

[00:01:49]
This is what they built this language in. And everything we did today is just an evolution of the same language that they made in 1972 on that teletype with this extremely limited amount of system resources. Now we're working on machines that are multiple orders of magnitude more powerful than this.

[00:02:05]
Why is C popular? Basically performance. I mean, this is the real thing. C++, you can get the same levels of performance. It is possible. Same thing with Rust, same thing with JAI, and Odin and Zig and all those. But it is much simpler than C++ and although there are new competing alternatives, they're much newer and less established than C.

[00:02:26]
C is 50 plus years old, Zig is, I don't know, five, maybe 10 years old, something like that. And yet, despite this, in the local static HTTP server Perf Olympics, what we've just built in the course of this workshop with no standard. No dependencies other than the standard library was able to outcompete these things that have been worked on way more than just this little workshop, and also which have millions in some cases of downloads and usage.

[00:02:53]
And this is again with my silly little benchmark of [LAUGH] frontendmasters.com. But you can reproduce this locally if you want. Okay, yeah, we talked about portable assembly and how it's like, well, I mean, yes, when it comes to the instructions we're compiling to, but there are still library differences with the OS.

[00:03:07]
So it's not like, it's all just complete roses, and you can just write your .c file and then just ship it out the door, and you don't have to think about anything. But at the same time, we did also look at some actual non portable assembly and hopefully we can all agree that for all of its problems, C is a significant ergonomics upgrade over all of that level of verbosity.

[00:03:27]
Zero overhead programming, pretty much true unless you wanna really micromanage your registers like an assembly language. But honestly, the compilers do tend to do a very good job of that. Almost zero safety programming. I mean, we had this slide at the beginning of the workshop. Yeah, I remember there were some questions about, hey, what does that mean in terms of memory safety?

[00:03:43]
Well, now we've seen a bunch of examples of specifically what memory unsafety means, and all of the fun consequences that can come from that. And it is also the language that other languages use to talk to each other. I mentioned at the very beginning that we would just talk very briefly about how you could use this if you are, a JavaScript programmer, and you wanna get some C into your JavaScript code.

[00:04:01]
This is how you would do this. So one of the reasons that I wanted to make sure that we were actually doing some like reading docs of C APIs is that if you go to this website, you can actually learn how to build Node.js add-ons in C. In other words, you can take your Node.js program and you can literally say, I want to call a little bit of C code that I've written from my Node.js code.

[00:04:21]
This will tell you how to do that, and it's gonna be using exactly the same styles of docs and the types and all of that type of stuff that we learned about in this course. So if you want to go out and build your own Node.js add ons, let's say you're writing something like, I just really need some really high performance section here.

[00:04:35]
If I could make this one part of my NodeJs server a little bit faster, this is a way you can do that. You can just bust out into C and do that. Or if you're like, I wanna write it in Rust, you could use C as an intermediary, because the Node.js API works in terms of C.

[00:04:48]
They actually made their library compatible with C++, but C++ is a superset of C. So anything that works in C++, you can also do in C. I've actually gotten paid at work to write a Node.js add-on [LAUGH] before. So I can confirm that this definitely works with just C, you don't need to know C to use this.

[00:05:06]
And essentially, this is a way that you can just speed up your your Node.js programs, or if you have some particular C library that you wanna use, that you wanna really get in there, and nobody's made a wrapper for that, you can totally do that. And in fact, you can even then distribute it on NPM.

[00:05:21]
I did wanna note that if you're trying to get into a large C or C++ code base, such as like a Postgres or an operating system or something like that, building it can be a significant challenge. So for this course, we're dealing with one .c file, building it not difficult.

[00:05:35]
If you're dealing with a gigantic C code base or C++ code base, building it, far from trivial. Fortunately, I do have a recommendation of how to learn about how to build these things. Andrew Kelley, who created Zig, gave this awesome talk called how to build software from source.

[00:05:49]
And depending on what programming language you're used to using, you might think that's easy. Why would you give a whole talk about that? It's not so easy when you're talking about large C and C++ projects. But if you watch this talk, he gives a lot of really, really good practical tips for how to actually configure and then build these large projects from source.

[00:06:06]
This is from the conference software you can love in 2023. And here is a link if you wanna go check that out. I also wanted to plug this book. This is a very famous book. I actually bought on eBay a copy. This is a picture of someone else's book.

[00:06:21]
But if you ever hear what K&R, this is the book they're talking about. So this is Brian Curnahan and Dennis Ritchie. So Dennis Ritchie was a creative C, and then Brian is a very famous C user and writer and author and stuff like that. This is the first edition, which I did wanna make sure and get cuz it has some fun little anecdotes in here about really old hardware stuff that aren't really applicable anymore.

[00:06:43]
But this is also just kind of a seminal it's one of the classic books that people talk about, right up there with structure and interpretation of computer programs as just even beyond the C stuff. It's also got a lot of advice about philosophies around around writing programming, and as we seen in this workshop, philosophies around writing C code are pretty different from best practices in other languages.

[00:07:03]
So I recommend giving it a read, if for no other reason, then it's just the perspective is pretty different than what you're probably used to. If you wanna get further down the performance optimization rabbit hole, I have two recommendations. One is Agner Foggs website, agner.org I wanted to include a screenshot of this with the pale green and everything.

[00:07:21]
He's also got a little animated plane that flies across when you go to the homepage and the plane has a little banner behind it that says I hate animations. That's kind of the vibe [LAUGH] that we're dealing with here. The information is incredibly, incredibly good. It's also pretty dense.

[00:07:35]
I personally have read through the optimizing software in C++ one. Some of it in C++ specific, but a whole lot of it is really, really just like low level hardware. Like here's what your CPU is doing, here are some tips and tricks. Very, very, very detailed, pretty dense, but if you really wanna go down the rabbit hole on this stuff, and you thought that some of the crazy tricks that we did at the end of a part six were fun, you can find more stuff like that at this website.

[00:08:00]
If you really wanna go all the way down the rabbit hole, Casey Muratore, who you might have seen on like the primogen stream. He's a frequent guest on there. He is an old school game engine programmer, and he made this unbelievably detailed course called performance aware programming. It is many, many, many hours, I've watched through all that.

[00:08:20]
I don't think I'm quite caught up right now. But it goes into things like the very beginning of the course is you go through and build a simulator for a really old CPU. So you can actually see literally what is the CPU doing when it gets the ones and zeros in.

[00:08:34]
We're gonna write software that deals with those ones and zeros and simulates actually doing what the CPU does. So you really can absolutely understand what all is going on. And then from there, he's like, that was an old school CPU that was much simpler. Let's talk about modern CPUs.

[00:08:47]
And he just goes through all the different features that modern CPUs have added and all the complexities they've added. There's also a cool moment at the beginning of the course where he shows you a Python program and is like, I can make this, what was it like? 10,000 times faster in C or something like that?

[00:09:01]
And he actually goes through a debugger and steps into showing you all the layers of overhead that Python is adding, and how if you just rewrite it in C, you can get it to be [LAUGH] massively, massively faster. Really, really cool stuff, extremely detailed, not for the faint of heart, but definitely, highly recommend that check this out if you're interested in really going down the performance rabbit hole.

[00:09:24]
And I also wanted to mention, yeah, there was a question?
>> Speaker 2: Just on a similar vein, Mishko Avery, the creator of Angular, has a course on Frontend Masters called Bare Metal JavaScript. And he builds a CPU register and-.
>> Richard Feldman: Nice.
>> Speaker 2: Memory and all that kind of stuff from scratch.

[00:09:42]
So, something to check out if you're already on Frontend Masters.
>> Richard Feldman: There you go. Yeah, nice.
>> Speaker 3: Does C-sharp have any relation here?
>> Richard Feldman: Does C-sharp have a relation here? Only in name, really. So C-sharp is basically, how do I say this? Java came out and then C-sharp came out and then people were like, this looks a lot like Java.

[00:10:03]
Other than the name C, which I'm sure was like a pun on C and C++, it almost has nothing in common with C and a lot, a lot in common with Java. That's, I guess, a quick summary of it. [LAUGH] Yeah. Yeah, cool. So some low overhead C alternatives.

[00:10:20]
We mentioned a couple of these, like C++ definitely way off the complexity into the spectrum, probably the most complicated popular language on the planet right now. Russ, strong contender, also extremely complicated language, but has way more guarantees and a lot of ergonomics benefits compared to C++. Also has the benefit of not having been created in 1980s so [LAUGH] got to benefit from a lot of advances in various things.

[00:10:44]
The D programming language is also pretty old. D is notable for like anytime you mention C++ or Russ or Zig on Hacker News, someone will be like, and D, what about D? But it's just not like really picking up as much steam. It's been around for a long time.

[00:10:57]
It's very powerful, has a lot of things, but it doesn't really seem, if I'm being honest, like it's gonna take off anytime soon. Zig is definitely like having a significant uptick in popularity by basically being a reaction to these sort of more complex C alternatives, like C++ and Rust and D and then Odin and JAI are two significantly less popular.

[00:11:19]
But also, kind of taking the same basic approach of, yeah, let's try to improve on C, but in a way where we're not adding so much complexity. JAI, I should mention, is a working name. This is created by Jonathan Blow, who's a friend of Casey Muratori and a former collaborator, you might know him as the creator of the really popular games Braid and The Witness.

[00:11:38]
He's made this programming language. JAI is specifically designed to be for games, and it's based on his frustrations with C++. And it has some significant differences between Zig and Odin, but I haven't actually used it, so I can't really comment on it in depth other than to say that right now on the simplest C alternatives, definitely Zig of these three is by far the most popular right now.

[00:11:58]
And time will tell how those things shaken. Also all of these low overhead C Alternatives they all support C and Rob just like most mainstream languages do because C is the language that other languages use talk to each other as previously discussed. Like I mentioned earlier, I mean C is basically the most universal language in the world today and now that you know C you know the most universal in the world.

[00:12:22]
Thanks very much.
>> [APPLAUSE]