WEBVTT NOTE This file was generated by Descript 00:00:13.518 --> 00:00:16.548 This is Self-Directed Research, our hosts, James and 00:00:16.548 --> 00:00:19.458 Amos get hyped about different topics and take turns each week, 00:00:19.458 --> 00:00:20.958 presenting their ideas to each other. 00:00:21.318 --> 00:00:23.748 You can check out the website, YouTube or Spotify to watch 00:00:23.748 --> 00:00:25.008 this episode's presentation. 00:00:25.308 --> 00:00:30.588 And visit sdr-podcast.com/episodes for previous episodes with presentations, 00:00:30.618 --> 00:00:32.418 videos, show notes and transcripts. 00:00:32.718 --> 00:00:34.698 New episodes are published every Wednesday. 00:00:35.688 --> 00:00:37.758 This episode is brought to you by CodeCrafters. 00:00:37.998 --> 00:00:40.788 There's more info at the end of the episode, or check out the link in our 00:00:40.788 --> 00:00:42.288 show notes to try it out for free. 00:00:42.618 --> 00:00:44.958 If you upgrade your plan, it supports this podcast. 00:00:45.118 --> 00:00:47.728 The costs can often be covered or reimbursed by your company, 00:00:47.728 --> 00:00:48.868 so it's worth asking about. 00:00:49.565 --> 00:00:53.250 Amos is up this week, exceeding the arbitrary time limit, with 45 slides 00:00:53.250 --> 00:00:55.050 covering the topic "target triples." 00:00:55.170 --> 00:00:55.950 Hope you enjoy! 00:01:03.324 --> 00:01:03.954 Oh, that reminds me. 00:01:04.364 --> 00:01:08.524 Did you, James, were you alive when- probably- people were passing around an 00:01:08.524 --> 00:01:11.284 ASCII art version of the Star Wars movie? 00:01:11.284 --> 00:01:13.534 Like you had to right size the notepad window and then you could 00:01:13.544 --> 00:01:15.214 hit page down, just like hold it. 00:01:15.214 --> 00:01:19.844 And then the repeat would make the- like a flip book, but in notepad for Windows 95. 00:01:19.894 --> 00:01:20.374 All right. 00:01:21.054 --> 00:01:21.844 I'm gonna share my screen. 00:01:22.312 --> 00:01:23.022 Is everybody ready? 00:01:23.260 --> 00:01:23.890 No? 00:01:24.817 --> 00:01:25.230 Yes. 00:01:25.550 --> 00:01:25.920 Yeah. 00:01:26.039 --> 00:01:30.559 You know, the fun thing is that every time I move my lips, it 00:01:30.559 --> 00:01:36.144 hurts because the interior of my lips rubs against very sharp metal bits. 00:01:36.434 --> 00:01:41.204 And so I've started correcting by just mumbling instead of pronouncing 00:01:41.204 --> 00:01:42.334 words because it hurts less. 00:01:42.354 --> 00:01:44.004 I'm optimizing for least pain. 00:01:44.174 --> 00:01:47.184 When we record the podcast, or when I record a video, I need to 00:01:47.184 --> 00:01:51.074 consciously reverse that and make it hurt so that it sounds good. 00:01:51.404 --> 00:01:52.054 It's fun. 00:01:52.404 --> 00:01:52.984 I love it. 00:01:54.042 --> 00:01:56.032 I did three or four years of braces too. 00:01:56.082 --> 00:01:59.042 So, I mean, I have not recently felt your pain, but I've absolutely 00:01:59.042 --> 00:02:00.202 felt your pain in the past. 00:02:00.342 --> 00:02:04.192 To be fair, I didn't have a podcast when I was in sixth grade, so... 00:02:04.332 --> 00:02:05.752 Maybe you should have, James. 00:02:05.862 --> 00:02:07.332 That was a blessing, I guess. 00:02:08.162 --> 00:02:10.552 I guess a blessing for both me and the world, I think, but... 00:02:10.557 --> 00:02:12.767 What would it have been about in sixth grade? 00:02:12.792 --> 00:02:15.172 I don't know, what was I into- probably video 00:02:15.172 --> 00:02:16.572 games or Pokemon at that point. 00:02:17.182 --> 00:02:18.622 Oh yeah, mine would have probably been 00:02:18.622 --> 00:02:20.636 like Spyro and Crash Bandicoot. 00:02:20.659 --> 00:02:21.439 Spyro? 00:02:21.659 --> 00:02:22.469 Sign me up! 00:02:22.469 --> 00:02:26.454 Mine would have been like, N64 games and Pokemon and... 00:02:26.484 --> 00:02:27.304 Anime, maybe? 00:02:27.304 --> 00:02:28.794 Like Dragon Ball Z and stuff like that? 00:02:29.172 --> 00:02:29.682 Okay. 00:02:29.822 --> 00:02:32.262 Well, unfortunately, today's podcast is about target triples. 00:02:33.142 --> 00:02:35.472 I have a subtitle here, because I'm contractually obligated 00:02:35.472 --> 00:02:37.962 to come up with subtitles for this podcast, against my will. 00:02:38.512 --> 00:02:41.242 And the subtitle is "Lies, damned lies, and target triples," 00:02:41.432 --> 00:02:42.322 which repeats the title. 00:02:42.322 --> 00:02:43.232 I don't know if that's legal. 00:02:43.302 --> 00:02:43.912 I have no idea. 00:02:44.147 --> 00:02:45.347 I accept it, it's fine. 00:02:45.661 --> 00:02:46.181 Thanks, Amanda. 00:02:46.181 --> 00:02:46.801 What's a target? 00:02:47.071 --> 00:02:48.991 It's a machine that code can run on. 00:02:49.291 --> 00:02:51.941 James, you have 15 seconds to disagree. 00:02:53.061 --> 00:02:57.371 It's more specific than that because one machine can have different 00:02:57.381 --> 00:03:00.461 operating systems, ABIs, configurations... 00:03:00.601 --> 00:03:02.411 Yeah, we're getting to that as you can imagine. 00:03:02.821 --> 00:03:06.347 . It's called a target triple, which has the prefix 'tri', so you would 00:03:06.347 --> 00:03:08.177 expect it to have three components. 00:03:08.537 --> 00:03:12.637 If you do check the target triple on the machine that I'm recording this from, 00:03:12.867 --> 00:03:20.800 which is a Mac Studio- If you do rustc -vV, like lowercase vV, which stands for 00:03:20.800 --> 00:03:25.110 'verbose version', then you get the full version of rustc, which includes 1.82.0. 00:03:25.130 --> 00:03:26.150 That's the current stable. 00:03:26.280 --> 00:03:32.300 You get the commit hash, the commit date, and the host shows aarch64-apple-darwin. 00:03:32.990 --> 00:03:35.900 So we're going to get into what each of those components mean, 00:03:36.274 --> 00:03:41.084 So in aarch64-apple-darwin: aarch64 is the architecture. 00:03:41.324 --> 00:03:45.524 In that case, it means 64 bit ARM- we'll get into ARM naming conventions. 00:03:45.574 --> 00:03:47.154 I have 45 slides. 00:03:47.254 --> 00:03:47.814 Get ready. 00:03:48.364 --> 00:03:50.334 The second segment in this case is the vendor. 00:03:50.334 --> 00:03:50.994 It's Apple. 00:03:51.254 --> 00:03:55.054 And the last segment in this case is the operating system, which is Darwin. 00:03:55.504 --> 00:03:58.834 Funny, I don't remember installing Darwin on this computer. 00:03:59.164 --> 00:04:03.584 Darwin is the name that Steve Jobs came up with when he was interim 00:04:03.584 --> 00:04:09.864 CEO of Apple in 1999 because, and I quote, "It's about evolution." 00:04:10.564 --> 00:04:12.434 I'm very glad that I found this slide. 00:04:12.474 --> 00:04:14.484 Again, if you're not looking at slides you're missing out, you can 00:04:14.484 --> 00:04:19.404 go to sdr-podcast.com to find the episode and find the slides for this. 00:04:19.440 --> 00:04:22.020 It's an article from Wired that says "Apple Opens OS Code." 00:04:22.695 --> 00:04:25.155 I'd like to interject for a second, what you've been referring 00:04:25.155 --> 00:04:30.645 to as Mac OSX is actually Mac OSX running on the Darwin kernel, or as I 00:04:30.645 --> 00:04:33.825 like to call it, Mac OSX plus Darwin. 00:04:33.895 --> 00:04:35.115 Yeah, well- 00:04:35.515 --> 00:04:36.795 Can I be an Apple neckbeard? 00:04:37.345 --> 00:04:38.665 You can, you can. 00:04:38.685 --> 00:04:41.855 I started looking into this, there's the whole history, like the lineage 00:04:41.855 --> 00:04:47.025 of like BSD and NeXTSTEP and OpenStep and Darwin was supposed to be this 00:04:47.075 --> 00:04:50.375 open thing where like, they said, " We believe that putting it out there will 00:04:50.375 --> 00:04:53.575 help us make it the best technology collectively we know how to make." 00:04:53.575 --> 00:04:55.745 So it was Apple saying, "Oh, we're open now. 00:04:55.745 --> 00:04:57.125 Everybody can come contribute." 00:04:57.545 --> 00:05:00.695 Nowadays it's like, yeah, they have an ftp server essentially that are legally 00:05:00.695 --> 00:05:05.085 required to have where they dump source code and you can look at it, I guess. 00:05:05.233 --> 00:05:07.073 I'm just here so I won't get fined. 00:05:07.728 --> 00:05:08.428 That's exactly it. 00:05:08.758 --> 00:05:10.448 So, uh, there's a whole history there. 00:05:10.448 --> 00:05:13.138 Which would make a fascinating other episode because this is only slide eight, 00:05:13.138 --> 00:05:14.618 so I had to cut the research short. 00:05:14.675 --> 00:05:16.675 People will write emails about this. 00:05:16.725 --> 00:05:21.015 It is fine I accept the responsibility on behalf of our producer Amanda who 00:05:21.015 --> 00:05:22.065 will actually have to look at those. 00:05:22.525 --> 00:05:24.905 So, there's other targets that look like it. 00:05:24.905 --> 00:05:31.215 Mine is aarch64-apple-darwin, but there's also x86_64-apple-darwin. 00:05:31.235 --> 00:05:34.265 I'm gonna be saying a lot of punctuation out loud so get used to it. 00:05:34.445 --> 00:05:35.625 Those were Intel Macs. 00:05:35.645 --> 00:05:39.185 Those we had before the Apple silicon Macs. 00:05:39.321 --> 00:05:41.701 So Apple M1, M2, M3, M4. 00:05:41.721 --> 00:05:42.741 Those are Apple silicon. 00:05:42.741 --> 00:05:44.131 They're aarch64. 00:05:44.331 --> 00:05:45.781 But before that we had Intel Macs. 00:05:45.801 --> 00:05:48.101 I still have one of them and you can recognize them very easily. 00:05:48.101 --> 00:05:50.673 The fan makes a noise when they turn on. 00:05:50.673 --> 00:05:50.763 Yes. 00:05:50.848 --> 00:05:51.958 They are loud and hot. 00:05:51.958 --> 00:05:52.374 Yes. 00:05:52.408 --> 00:05:53.138 Like me! 00:05:53.259 --> 00:05:54.448 Yeah, I was gonna say I was looking around. 00:05:54.668 --> 00:05:56.688 I have one in my house, but not in this room. 00:05:57.113 --> 00:06:00.723 Before Intel Macs, we had PowerPC Macs. 00:06:01.200 --> 00:06:04.221 James just is making big eyes at my slide with a picture of- 00:06:04.221 --> 00:06:05.801 Hell yeah, 68k Macs. 00:06:05.820 --> 00:06:08.290 A Mac Quadra 605. 00:06:08.290 --> 00:06:08.920 I don't actually know... 00:06:08.920 --> 00:06:09.090 No. 00:06:09.945 --> 00:06:10.275 Oh, sorry. 00:06:10.275 --> 00:06:11.505 No, this is powerpc. 00:06:11.585 --> 00:06:13.145 68k is before that. 00:06:13.185 --> 00:06:14.095 Yeah, yeah, yeah. 00:06:14.175 --> 00:06:16.275 Oh, we'll get to 68k, don't you worry. 00:06:18.625 --> 00:06:20.225 Siri just woke up for some reason. 00:06:22.655 --> 00:06:26.360 I talked about the Quadra 605 and Siri was like, "I will help." 00:06:26.366 --> 00:06:26.876 Friend! 00:06:26.976 --> 00:06:27.706 Friend! 00:06:29.136 --> 00:06:32.456 rustc, the Rust compiler, has a bunch of PowerPC targets. 00:06:32.626 --> 00:06:36.121 There's three levels of support, which we'll get into roughly what they mean. 00:06:36.351 --> 00:06:41.238 There's three power PC tier two targets in Rust: powerpc 00:06:41.238 --> 00:06:48.159 -unknown-linux-gnu, powerpc64 - the same thing, and powerpc64le for 00:06:48.159 --> 00:06:50.079 little endian dash the same thing. 00:06:50.289 --> 00:06:51.959 What devices are still PowerPC today? 00:06:51.959 --> 00:06:53.229 Because those are actually maintained. 00:06:53.299 --> 00:06:55.150 I was trying to find the answer and I tweeted about it. 00:06:55.200 --> 00:06:57.290 You might've seen it and answered something useful. 00:06:57.434 --> 00:06:59.534 Why is PowerPC still relevant today? 00:06:59.754 --> 00:07:03.440 A couple of answers: as recently as the PS3, the Sony PlayStation 00:07:03.440 --> 00:07:07.580 3 console, they were using PowerPC based processors on there. 00:07:07.630 --> 00:07:08.550 PS4 and PS5 changed. 00:07:08.550 --> 00:07:09.842 That was the Cell processor. 00:07:09.902 --> 00:07:11.402 Was the Cell PowerPC based? 00:07:11.412 --> 00:07:14.707 Cause Cell was its own weird architecture that only the PS3 used. 00:07:14.761 --> 00:07:16.511 I have speaker notes, I come prepared this time. 00:07:16.521 --> 00:07:21.201 The PS3 uses the Cell microprocessor, which is made up of one 3.2 GHz 00:07:21.201 --> 00:07:24.071 PowerPC-based power processing element and six accessible 00:07:24.081 --> 00:07:27.041 synergistic processing elements, SPEs. 00:07:27.050 --> 00:07:29.850 So yes, it is- at least part of it is PowerPC based. 00:07:30.384 --> 00:07:33.184 And then also, Apple used to make servers, remember that? 00:07:33.954 --> 00:07:37.879 They made the XServe G5, I think that's the last one they released? 00:07:38.409 --> 00:07:41.489 We have a picture of it here with typical Apple marketing, which 00:07:41.489 --> 00:07:44.919 means there's nice shadows and like little reflections on the image. 00:07:45.069 --> 00:07:48.015 It says, "More power, less energy, introducing the fastest, most 00:07:48.045 --> 00:07:49.635 energy efficient Xserve ever." 00:07:49.732 --> 00:07:52.450 It has several configurations: that come in single processor, 00:07:53.080 --> 00:07:54.050 dual processor, cluster node. 00:07:54.490 --> 00:07:59.680 And they're like roughly two- between two and 2.3 gigahertz per processor? 00:07:59.807 --> 00:08:00.617 Core, probably? 00:08:00.627 --> 00:08:01.377 Yeah... 00:08:01.677 --> 00:08:04.947 uses ECC memory with a 16 gigabytes maximum. 00:08:04.997 --> 00:08:07.427 I don't have the year here, which is a shame, but it is a 00:08:07.427 --> 00:08:09.747 server from another era for sure. 00:08:09.884 --> 00:08:11.014 But it was PowerPC. 00:08:11.074 --> 00:08:11.694 It still is. 00:08:11.704 --> 00:08:13.114 Some people are still running it, I imagine. 00:08:13.514 --> 00:08:17.737 IBM is probably the reason that PowerPC is still a tier two target 00:08:17.767 --> 00:08:23.147 in rustc because they actually forked PowerPC in the nineties. 00:08:23.247 --> 00:08:29.127 I was born, butterfly flapped its wing, IBM forks PowerPC into an architecture 00:08:29.127 --> 00:08:32.837 called POWER, which is not at all confusing, which is actually an acronym. 00:08:32.837 --> 00:08:35.567 It means "performance optimization with enhanced RISC." 00:08:37.537 --> 00:08:40.883 And by "enhanced RISC," they mean an instruction set that actually has a 00:08:40.883 --> 00:08:44.330 hundred plus instructions and that supposedly has fixed width encoding, 00:08:44.350 --> 00:08:47.800 which is supposed to be a RISC thing, but actually they do have prefix instruction 00:08:47.800 --> 00:08:51.910 because they have a lot of SIMD stuff, like single instruction, multiple data. 00:08:52.220 --> 00:08:57.390 So, POWER is still used today, as sort of a middle ground between like full 00:08:57.390 --> 00:08:59.853 on mainframe and like smaller things. 00:09:00.043 --> 00:09:01.493 Some people still have POWER hardware. 00:09:01.803 --> 00:09:02.783 Do you know where else they are? 00:09:03.019 --> 00:09:03.499 Banking? 00:09:03.819 --> 00:09:08.379 A lot of radiation hardened processors for satellites and Mars 00:09:08.379 --> 00:09:13.974 rovers and things like that are on CPUs like the RAD750 which is a larger 00:09:13.974 --> 00:09:20.208 radiation resistant power pc architecture designed to be resistant to the radiation 00:09:20.208 --> 00:09:22.418 you get in space or when there's not an atmosphere or magnetosphere. 00:09:22.418 --> 00:09:22.828 That's true! 00:09:23.953 --> 00:09:27.233 They run slow and but they are generally 00:09:27.233 --> 00:09:29.063 qualified to be rad resistant. 00:09:29.934 --> 00:09:33.364 They get used in either embedded systems including goofy rad-hardened 00:09:33.384 --> 00:09:36.274 embedded systems in space as well, still, to this day. 00:09:36.294 --> 00:09:38.580 Yeah, it's fun cause like, designing for satellites 00:09:38.580 --> 00:09:41.810 is completely different from designing for Earth, cause... 00:09:41.850 --> 00:09:43.290 yeah, you have to worry about these. 00:09:43.750 --> 00:09:46.630 Raptor Computing System came up when I asked people about 00:09:46.630 --> 00:09:48.050 PowerPC on social media. 00:09:48.260 --> 00:09:54.780 They make workstations you can buy, starting at $9,977, so I ordered a couple 00:09:54.820 --> 00:09:56.490 just to get a feel for them, y'know? 00:09:57.436 --> 00:09:57.729 Business expense. 00:09:57.729 --> 00:09:58.771 Can we expense those, Amanda? 00:09:58.771 --> 00:09:59.201 Yes? 00:09:59.394 --> 00:10:02.425 Yeah, no, okay, no, sorry, I'll just return them, it's fine. 00:10:02.645 --> 00:10:05.705 But they make new POWER9 workstations, even though IBM's already doing 00:10:05.715 --> 00:10:08.975 POWER10, because POWER9 was the last one you could still, like, all the 00:10:08.975 --> 00:10:12.345 bits were free, so Power10 is still up in the air for licensing reasons. 00:10:13.285 --> 00:10:18.330 There is Debian, the Debian Linux distribution makes a survey, a Popularity 00:10:18.330 --> 00:10:21.220 Contest, they call it, of which architecture are you running right now? 00:10:21.430 --> 00:10:27.850 And so we are looking, and you should be too, go to sdr-podcast.com/episodes 00:10:27.870 --> 00:10:29.150 and find this episode. 00:10:29.230 --> 00:10:32.340 you can look at the slide that says "Number of submissions per architecture," 00:10:32.340 --> 00:10:36.850 and we can see, because this ranges from 2004 to 2024, so we got 20 years 00:10:36.850 --> 00:10:40.000 worth of like personal computing history in there and servers and whatnot. 00:10:40.058 --> 00:10:45.058 You can see the rise of 64 bit computing starting from zero in 00:10:45.058 --> 00:10:48.128 2004, going all the way to... 00:10:48.518 --> 00:10:52.614 well, it's just powers of twos was like 262,000 respondents. 00:10:52.664 --> 00:10:56.498 And then you can see the slow decline of 32 bit, in this 00:10:56.498 --> 00:10:58.478 case the i386 architecture. 00:10:58.688 --> 00:11:00.674 And you can see also the rise of ARM64. 00:11:00.961 --> 00:11:05.571 And then you cannot at all see PowerPC on there The total responses for the 00:11:05.571 --> 00:11:08.632 last survey were 112 out of like a lot. 00:11:08.948 --> 00:11:09.188 Yeah. 00:11:09.634 --> 00:11:11.710 I also got an actual response from someone who works on the 00:11:11.710 --> 00:11:14.300 Rust compiler: why is PowerPC 64 tier 2? 00:11:14.560 --> 00:11:17.400 The simple answer is because there are people from IBM that care about fixing the 00:11:17.410 --> 00:11:18.960 target and therefore it gets to be tier 2. 00:11:18.980 --> 00:11:22.676 Tier 2 requirements essentially boil down to: can we legally and without too 00:11:22.676 --> 00:11:25.236 much pain compile core and std in our CI? 00:11:25.986 --> 00:11:29.506 And two, are there people who we can call when it stops working? 00:11:29.776 --> 00:11:31.256 So that's all it takes to be a tier two target. 00:11:31.385 --> 00:11:33.085 Yeah, there's two different flavors of tier two. 00:11:33.085 --> 00:11:36.035 One is basically tier two hosted, which means that they can compile 00:11:36.035 --> 00:11:38.325 and run cargo and rustc on it. 00:11:38.345 --> 00:11:40.305 And then there's ones particularly for the embedded system. 00:11:40.305 --> 00:11:44.269 So like all the embedded systems targets- -are all or mostly tier two 00:11:44.269 --> 00:11:45.729 because they are non hosted ones. 00:11:45.939 --> 00:11:48.749 Cause you can't run cargo on a microcontroller, but you can compile 00:11:48.959 --> 00:11:50.649 to the microcontroller with cargo. 00:11:50.969 --> 00:11:51.279 Yes. 00:11:51.659 --> 00:11:55.969 There was an effort in July of 2023 to kind of sort out tier two targets 00:11:55.989 --> 00:11:56.939 because there were too many of them. 00:11:56.939 --> 00:11:59.029 And some of them didn't have documented maintainers. 00:11:59.029 --> 00:12:00.169 So it was like: Oh, this one broke. 00:12:00.179 --> 00:12:00.749 Who do we call? 00:12:00.749 --> 00:12:01.289 We don't even know. 00:12:01.289 --> 00:12:02.139 It's not really clear. 00:12:02.369 --> 00:12:03.479 People change jobs. 00:12:03.479 --> 00:12:05.189 they started documenting that better. 00:12:05.409 --> 00:12:10.009 The thing with tier two targets is that they can be either promoted to tier one 00:12:10.089 --> 00:12:11.709 or they can be demoted to tier three. 00:12:11.719 --> 00:12:15.249 So there's a lot of different quality of targets in there. 00:12:15.819 --> 00:12:19.707 If we're looking at those PowerPC targets again, you may notice that for target 00:12:19.707 --> 00:12:21.527 triples, they have four components. 00:12:21.727 --> 00:12:25.347 There's PowerPC, which is the arch, there's unknown, which is 00:12:25.377 --> 00:12:29.597 the vendor, Linux is the operating system, and then gnu is the ABI. 00:12:29.841 --> 00:12:32.891 Did you ever read the Hitchhiker's Guide to the Galaxies books? 00:12:33.209 --> 00:12:35.229 I pretended I did, but I never actually. 00:12:35.391 --> 00:12:35.861 That's fair. 00:12:36.101 --> 00:12:38.731 So they were writing the books and they originally built it as the 00:12:38.731 --> 00:12:41.061 Hitchhiker's Guide to the Galaxy Trilogy. 00:12:41.291 --> 00:12:42.601 And then they made a fourth book. 00:12:42.621 --> 00:12:46.071 And on the title, they said, "The fourth entry in the increasingly less 00:12:46.091 --> 00:12:50.411 accurately named Hitchhiker's Guide to the Galaxy Trilogy" or something like that. 00:12:50.411 --> 00:12:52.981 So this is- -exactly that experience again. 00:12:53.208 --> 00:12:54.108 This is exactly it. 00:12:54.218 --> 00:12:58.768 The ones we're looking at ends with gnu for glibc, the gnu C library, 00:12:58.938 --> 00:13:02.568 but there's also musl variants, which I say 'muzzle,' some people say 00:13:02.568 --> 00:13:07.237 'muscle,' but it's actually spelled M U S L And interestingly, the gnu 00:13:07.237 --> 00:13:13.307 target for x86_64-unknown-linux-gnu -that's the podcast you chose to 00:13:13.307 --> 00:13:18.107 listen, don't put the blame on us- is tier 1, but the musl one is tier 2. 00:13:18.490 --> 00:13:24.156 And in most people's heads, the musl target is there to serve as a fallback 00:13:24.216 --> 00:13:29.586 for when you download a binary that was linked against the glibc but somehow your 00:13:29.626 --> 00:13:33.296 Linux distribution has a different version of glibc than their Linux distribution 00:13:33.466 --> 00:13:37.786 so it doesn't actually run and so you just get the musl one instead which is 00:13:37.826 --> 00:13:41.636 statically linked we're looking at the slide that shows that the musl build 00:13:41.636 --> 00:13:44.886 of cargo dist for example is statically linked it doesn't have any dependencies 00:13:44.906 --> 00:13:49.712 so it should be able to run on any Linux system out there that supports LSB I 00:13:49.712 --> 00:13:51.062 guess, but that's been a thing forever. 00:13:51.452 --> 00:13:52.272 Where was I going with this? 00:13:52.732 --> 00:13:53.002 Right. 00:13:53.192 --> 00:13:55.422 That people assume that the musl target is statically linked, 00:13:55.442 --> 00:13:56.892 but actually it shouldn't be. 00:13:56.892 --> 00:14:01.541 And in 2021, the Rust compiler team accepted a major change that says: 00:14:01.641 --> 00:14:05.731 Oh, right now we default musl to be statically linked, but that was a mistake. 00:14:05.971 --> 00:14:08.421 And it turns out that there's people who actually care about 00:14:08.431 --> 00:14:10.281 dynamically linking against musl. 00:14:10.701 --> 00:14:14.131 And those people are the people who develop musl and musl based distributions 00:14:14.131 --> 00:14:17.774 like Alpine and Void Linux and you can do that on Gentoo as well, I believe. 00:14:17.893 --> 00:14:21.253 And so actually we should change the default which is going to 00:14:21.253 --> 00:14:24.393 break everyone else's assumptions of what the musl target is for. 00:14:24.514 --> 00:14:28.959 I have like a zoomed in part of the slide that just shows how you are 00:14:28.979 --> 00:14:32.282 now, after that change, supposed to actually enable static linking and 00:14:32.282 --> 00:14:39.272 it's -Ctarget-feature=+crt-static. 00:14:39.582 --> 00:14:43.632 And the way you're supposed to disable it is to do =-crt-static or minus 00:14:43.632 --> 00:14:46.022 crt-static, I guess, in that sense. 00:14:46.717 --> 00:14:49.757 Which is the most confusing naming scheme ever and I hate it personally. 00:14:49.880 --> 00:14:51.730 But yeah, the change was approved in 2021. 00:14:51.730 --> 00:14:52.780 what happened since? 00:14:53.000 --> 00:14:55.640 it's really hard to actually make that change happen without 00:14:55.640 --> 00:14:57.260 making everyone instantly angry. 00:14:57.480 --> 00:14:58.150 And so... 00:14:58.190 --> 00:15:02.390 I think there's other like intermediate changes that they want to do first, but... 00:15:02.825 --> 00:15:03.275 Yeah. 00:15:03.325 --> 00:15:04.555 The real life is messy. 00:15:05.305 --> 00:15:07.645 We're roughly halfway through my slides I would say. 00:15:08.041 --> 00:15:11.361 We're looking now at i686-pc-windows-gnu. 00:15:11.422 --> 00:15:12.442 That's kind of a weird change. 00:15:12.472 --> 00:15:17.892 On the Debian graph earlier we saw i386 as a placeholder for 32-bit processors, 00:15:18.013 --> 00:15:19.853 and this is chosen to be i686. 00:15:19.873 --> 00:15:20.323 Why? 00:15:20.743 --> 00:15:23.758 Because if you look at the history of Intel processors, you 00:15:23.758 --> 00:15:28.838 start in 1971 with the 4004, and then you have the 8008 in '72. 00:15:29.048 --> 00:15:33.233 Then you have the 8080, I'm going to call it, the uh, 8 bit successor, Then 00:15:33.233 --> 00:15:38.893 you have the 8086, the first in the x86 series, like the 186, if we had a 186, 00:15:38.893 --> 00:15:43.833 and then you have the 286 in '82, which was 16 bit, and then the 386, which 00:15:43.833 --> 00:15:45.573 started the era of 32 bit processing. 00:15:45.573 --> 00:15:47.938 So that's why Debian calls it 386. 00:15:47.938 --> 00:15:51.418 It's because it is the first 32 bit processor in the Intel line, at least, 00:15:51.418 --> 00:15:53.258 I don't know about any other vendors. 00:15:53.332 --> 00:15:57.902 Then you have the 486, then you have the 586, which didn't actually get 00:15:57.932 --> 00:15:59.972 called 586, but it is like the Pentium. 00:16:00.152 --> 00:16:03.522 Then the 686 is the Pentium Pro and then later the Pentium 2 and 00:16:03.522 --> 00:16:05.887 Pentium 3 also belong to that. 00:16:06.217 --> 00:16:10.537 And so actually in the Rust world, when you're looking at an i686 target, 00:16:11.107 --> 00:16:14.338 what you're talking about is: it has streaming SIMD extensions. 00:16:14.357 --> 00:16:15.797 It has SSE support. 00:16:15.797 --> 00:16:21.293 So there's also i586 targets, which I don't know who uses, but it's 00:16:21.303 --> 00:16:26.033 32 bit Linux without SSE, without those streaming SIMD extensions. 00:16:26.190 --> 00:16:27.680 Yeah, I'm trying to think of another analog of this. 00:16:27.788 --> 00:16:31.128 It's sort of like, I know on Mac, it's basically setting, like, your 00:16:31.138 --> 00:16:36.488 enablement list of that of, like: ah, this is SDK version whatever and above. 00:16:36.488 --> 00:16:40.128 Where, like, 386 means you only support the subset of features 00:16:40.138 --> 00:16:44.308 that that does, where, like, 586 or 686 gets you a larger superset. 00:16:44.598 --> 00:16:47.868 Or, it's also the reason why a lot of people, when they're compiling for 00:16:47.868 --> 00:16:49.158 high performance stuff, they do... 00:16:49.198 --> 00:16:51.839 like native or whatever because your, processor is probably like 00:16:51.867 --> 00:16:54.733 it's got a whole suite of things that the Rust compiler is not going 00:16:54.733 --> 00:16:58.353 to turn on because you told it like pretend that I'm compiling for 00:16:58.353 --> 00:17:04.033 a machine from 2011 and not your wonderful machine in 2024 or whatever. 00:17:04.233 --> 00:17:06.393 But then if you do that- like you built something with 00:17:06.393 --> 00:17:10.501 -march=native and -mtune=native and whatnot - and then you deploy it to like 00:17:10.581 --> 00:17:12.501 a Hetzner server you got on the auction. 00:17:12.691 --> 00:17:18.232 And then it actually is a processor from like 2011 Then you might notice SIGILL, 00:17:18.232 --> 00:17:20.557 I think, just illegal instruction, which is like: I don't have this 00:17:20.557 --> 00:17:23.115 extension, I don't have SSE 4.2, so, no. 00:17:23.350 --> 00:17:24.520 So that's why this target is there, 00:17:24.520 --> 00:17:27.570 There's like two really enthusiastic retro computer folks who 00:17:27.570 --> 00:17:31.430 are just very excited that they can run their- I think you can run Rust on like 00:17:31.430 --> 00:17:35.530 Windows 95 or Windows 98 and I think that's what that 586 target is for is 00:17:35.540 --> 00:17:37.522 for running it on like, Windows 98. 00:17:37.575 --> 00:17:38.355 Speaking of- 00:17:38.647 --> 00:17:39.167 Oh cool 00:17:39.412 --> 00:17:42.992 If you're making Windows binaries, I beg of you, please use 00:17:42.992 --> 00:17:47.552 the dash msvc ABI, because that's what you should do on Windows, 00:17:47.572 --> 00:17:49.812 because it's like, not using Xcode. 00:17:49.842 --> 00:17:53.802 It's not using Clang to build binaries for macOS, you could, but why would you? 00:17:54.092 --> 00:17:55.492 Just use the native toolchain, okay? 00:17:55.492 --> 00:17:56.452 Make me- make me happy. 00:17:56.452 --> 00:17:59.073 But if you don't want to, if you want to make me sad, there's a 00:17:59.073 --> 00:18:04.313 gnu ABI thing, which then doesn't correspond really to glibc. 00:18:04.313 --> 00:18:09.801 In this case, like it means you're using MinGW or MinGW-w64, which 00:18:09.821 --> 00:18:12.391 also supports 32 bit and 64 bit. 00:18:12.538 --> 00:18:17.678 And so we have those windows-gnu targets, which honestly is never what 00:18:17.678 --> 00:18:20.475 you want, unless you know already, and then don't listen to me. 00:18:20.655 --> 00:18:27.855 But then also usually the Windows targets are like x86_64-pc-windows-gnu 00:18:27.875 --> 00:18:32.233 or msvc, but actually in the target list for rustc, I find UWP, which is 00:18:32.233 --> 00:18:37.333 Universal Windows Platform, which doesn't have all the traditional Win32 stuff. 00:18:37.333 --> 00:18:41.523 all the calls to like open files, start processes, wait on objects 00:18:41.523 --> 00:18:42.553 and everything, none of that. 00:18:42.969 --> 00:18:45.569 Now everything is through WinRT, the new system. 00:18:45.779 --> 00:18:48.837 I don't know what's the current status of UWP to be completely 00:18:48.837 --> 00:18:50.597 honest, but we have a target for that! 00:18:51.215 --> 00:18:55.225 There's a specific target for Windows 7 as well, which I believe was dropped from the 00:18:55.225 --> 00:18:56.885 rest, but there's still a target for it. 00:18:56.925 --> 00:18:57.695 I don't know if it works. 00:18:57.735 --> 00:18:58.275 No idea. 00:18:58.325 --> 00:18:59.045 Didn't even try. 00:19:00.150 --> 00:19:05.619 Let's now talk about the way ARM naming conventions work, which is they don't. 00:19:05.999 --> 00:19:07.769 We have already seen aarch64. 00:19:07.789 --> 00:19:08.739 We have Apple Darwin. 00:19:08.739 --> 00:19:11.029 That's again, all the Apple silicon processors. 00:19:11.221 --> 00:19:15.275 We have aarch64 unknown Linux gnu, which you would get if you 00:19:15.275 --> 00:19:18.887 install Asahi Linux, for example, on your M1 MacBook or something. 00:19:18.995 --> 00:19:22.644 But also, some laptops that are made specifically to run Linux on 00:19:22.644 --> 00:19:26.655 ARM64, like the Pinebook pro, the HP EliteBook and everything. 00:19:26.755 --> 00:19:30.215 We have aarch64 PC Windows MSVC because Microsoft's also been 00:19:30.215 --> 00:19:32.525 trying to introduce ARM 64 laptops. 00:19:32.825 --> 00:19:35.435 The Surface Pro X is an ARM 64 machine. 00:19:35.735 --> 00:19:38.945 They introduced support in like Windows 10 and they improved it in Windows 11 00:19:38.975 --> 00:19:42.080 'cause it lacked x86_64 bit emulation. 00:19:42.080 --> 00:19:45.500 So there was this weird moment where you could have ARM 64 and it could 00:19:45.500 --> 00:19:50.840 emulate 32 bit x86, but not 64 bit x86. 00:19:50.840 --> 00:19:51.132 Yeah. 00:19:51.202 --> 00:19:52.412 Cause it's all emulation now. 00:19:52.646 --> 00:19:57.546 But if we look at ARM outside of desktop and laptop computers, then 00:19:57.916 --> 00:19:59.236 that's- count with me, all right? 00:19:59.236 --> 00:20:03.629 So arm just means ARM version six, and then you have armv7, 00:20:03.649 --> 00:20:05.059 which means ARM version seven. 00:20:05.349 --> 00:20:08.069 And then you have aarch64, which means ARM version eight. 00:20:08.189 --> 00:20:08.489 Of course. 00:20:08.489 --> 00:20:08.819 James: disagree. 00:20:10.044 --> 00:20:10.284 . What? 00:20:10.484 --> 00:20:11.954 No, yeah, there's even more. 00:20:12.603 --> 00:20:16.434 Even in supported in Rust there's armv5te, there's armv4. 00:20:16.464 --> 00:20:18.064 If you wanna run on like... 00:20:18.094 --> 00:20:21.014 what is it, like a Nintendo 3DS or something? 00:20:21.014 --> 00:20:23.084 Is still using like armv4 or something? 00:20:23.114 --> 00:20:23.804 Oh, we'll get there? 00:20:23.804 --> 00:20:24.074 Okay, cool. 00:20:24.074 --> 00:20:24.514 We'll get there. 00:20:24.514 --> 00:20:24.944 Don't worry. 00:20:25.084 --> 00:20:27.184 We're starting to get into like- I get into a lot 00:20:27.184 --> 00:20:28.244 of this with embedded stuff. 00:20:28.244 --> 00:20:30.806 This is not the single, the singular ARM slide. 00:20:31.126 --> 00:20:35.686 There's also armv7a, armv7r, 'cause there's different profiles. 00:20:35.686 --> 00:20:36.526 I don't know what those mean. 00:20:36.526 --> 00:20:36.736 There's- 00:20:36.743 --> 00:20:37.257 Oh, okay! 00:20:37.276 --> 00:20:40.906 So real quick: ARM has three major classes of processors now. 00:20:40.916 --> 00:20:43.946 They've got application processors, which are 'a': those are all 00:20:43.946 --> 00:20:46.376 of your mobile phones, your desktops and things like that. 00:20:46.676 --> 00:20:51.556 There's Cortex-R, which are the real time processors, which are used in, 00:20:51.556 --> 00:20:53.996 like, vehicles and real time systems. 00:20:53.996 --> 00:20:57.666 And then you have the microcontroller profile, which is what, whenever 00:20:57.666 --> 00:21:00.646 I'm talking about developing for, like, USB devices and whatever, 00:21:00.646 --> 00:21:01.896 I'm talking about the M profile. 00:21:02.026 --> 00:21:04.886 So ARM has three profiles: A, R, and M. 00:21:05.186 --> 00:21:08.866 And they are specified in the ARM Reference Manual: A, R, M- 00:21:08.901 --> 00:21:09.831 Mind blown. 00:21:09.941 --> 00:21:13.261 and the compact architecture used so when you want 00:21:13.261 --> 00:21:17.566 to use 16 bit instructions instead of 32 or 64 on these target is thumb! 00:21:17.726 --> 00:21:19.376 That's the tiny version of ARM. 00:21:19.644 --> 00:21:20.634 That's my next slide! 00:21:20.634 --> 00:21:21.774 That's thumb mode. 00:21:21.993 --> 00:21:22.489 Yeah. 00:21:22.710 --> 00:21:24.090 James, I made this slide deck for you. 00:21:24.821 --> 00:21:26.881 I've had to go look up the arm arm before! 00:21:27.096 --> 00:21:30.476 Yeah, so Thumb Mode, I discovered, I thought it was 00:21:30.476 --> 00:21:31.866 like, "What the frick is that? 00:21:31.866 --> 00:21:33.356 Is it the different class of devices?" 00:21:33.356 --> 00:21:35.876 No, it's just the mode, so you can switch between like Thumb 00:21:35.876 --> 00:21:36.936 Mode and regular mode, right? 00:21:37.071 --> 00:21:37.491 Well... 00:21:37.666 --> 00:21:38.646 Or just the startup? 00:21:38.646 --> 00:21:41.055 A class profiles can do either. 00:21:41.165 --> 00:21:43.235 So before armv8, you could switch. 00:21:43.235 --> 00:21:46.475 They actually had an instruction that would switch the CPU from 00:21:46.475 --> 00:21:50.321 interpreting large ARM instructions to small thumb instructions. 00:21:50.321 --> 00:21:53.691 So you could swap back and forth on a Cortex-A if you wanted 00:21:53.691 --> 00:21:55.141 compact instruction encoding. 00:21:55.461 --> 00:21:58.411 That's actually the feature of the Cortex-M devices though, is they 00:21:58.586 --> 00:22:01.006 only support thumb instructions. 00:22:01.016 --> 00:22:03.096 They cannot support ARM instructions. 00:22:03.286 --> 00:22:07.196 So they actually, you could run the same instructions on your mobile phone and 00:22:07.196 --> 00:22:11.726 your microcontroller, as long as you only use the thumb subset, which are the 00:22:11.736 --> 00:22:13.616 like sort of compressed instruction mode. 00:22:13.836 --> 00:22:14.046 Yeah. 00:22:14.046 --> 00:22:18.306 The arm mode instructions are 32 bits and in Thumb Mode they're 16 bits, 00:22:18.306 --> 00:22:19.686 so the, the whole program ends up- 00:22:19.756 --> 00:22:21.086 Except for when they aren't, but yeah. 00:22:21.206 --> 00:22:23.156 I assume there's prefixed instructions as well. 00:22:23.476 --> 00:22:24.676 Yeah, there are more. 00:22:24.736 --> 00:22:27.176 RISC instruction sets are not very RISC anymore. 00:22:27.271 --> 00:22:27.751 Yeah. 00:22:28.341 --> 00:22:32.641 So in terms of how they show up, I'm showing just the arch 00:22:33.011 --> 00:22:35.116 prefix of the relevant targets. 00:22:35.126 --> 00:22:41.616 We have thumb v6m, thumb v7m, thumb v7em, I don't know what the e's doing there. 00:22:41.736 --> 00:22:45.516 So V six- V six is Cortex M zero, cortex M zero plus. 00:22:45.516 --> 00:22:47.556 V seven is Cortex M three. 00:22:47.916 --> 00:22:53.296 Cortex V seven EM is Cortex M four, M seven and thumb V eight M are all the 00:22:53.296 --> 00:22:57.238 two digits, like Cortex 20 something Cortex 30 something processors. 00:22:57.293 --> 00:23:00.160 That's the first time I've seen a period in any of these. 00:23:00.172 --> 00:23:01.549 They got weird with the naming on this- 00:23:01.640 --> 00:23:04.320 Thumb v8m dot base? 00:23:04.320 --> 00:23:07.645 It broke all of our tooling when they did that, like both 00:23:07.655 --> 00:23:09.605 in rustc and in the embedded Rust world. 00:23:09.825 --> 00:23:12.355 There was like a couple of weeks where we just had to rewrite- 00:23:12.355 --> 00:23:13.980 I mean, it's not technically forbidden. 00:23:13.990 --> 00:23:15.780 There's no rules for target triples. 00:23:15.880 --> 00:23:16.800 They're not triples. 00:23:16.860 --> 00:23:17.760 They're not targets. 00:23:18.800 --> 00:23:21.000 Of course you can put dots in there, who cares, like... 00:23:21.270 --> 00:23:24.120 but there was an understanding, like, before that, all of them are, like, 00:23:24.130 --> 00:23:26.860 A to Z and 0 to 9, and that's it. 00:23:27.630 --> 00:23:27.970 Yeah. 00:23:28.150 --> 00:23:36.270 Speaking of saving memory, there's a very interesting arm64_32-apple-watchos. 00:23:36.440 --> 00:23:37.430 And I learned about this one. 00:23:37.470 --> 00:23:42.434 It's arm64 but 32 bit pointers, and it's used for the Apple Watch because they 00:23:42.454 --> 00:23:43.854 have lots of code in some applications. 00:23:44.344 --> 00:23:47.864 64 bit pointers actually use up a bunch of memory, just the whole, 00:23:47.874 --> 00:23:49.374 the address of other things. 00:23:49.789 --> 00:23:51.719 And there's not that much memory on the Apple Watch anyway. 00:23:52.019 --> 00:23:55.069 But you still want ARM64 because like, I guess that's what you're 00:23:55.069 --> 00:23:58.409 manufacturing, and also you get more efficient instructions for like 00:23:58.419 --> 00:24:01.609 SIMD and stuff, cause you do signal processing on there, you do like voice 00:24:01.609 --> 00:24:02.829 recognition, you do a lot of things. 00:24:03.139 --> 00:24:04.419 So you get the best of both worlds. 00:24:04.439 --> 00:24:06.269 You get 32 bit pointers on ARM64. 00:24:06.519 --> 00:24:09.114 There's something similar on a Raspberry Pi, where 00:24:09.114 --> 00:24:12.754 they have a 64 bit kernel, but user space is 32 bit, and they've- 00:24:13.144 --> 00:24:14.814 That's my next slide, James! 00:24:14.814 --> 00:24:15.964 You're so predictable! 00:24:15.964 --> 00:24:24.224 It's aarch64_be for big endian -unknown-linux-gnu_ilp32. 00:24:24.554 --> 00:24:25.394 You didn't have a stroke. 00:24:25.394 --> 00:24:26.164 That is the entire name. 00:24:26.164 --> 00:24:29.174 I had to shrink the font just so it fits in the slide. 00:24:29.244 --> 00:24:30.594 I guess it's still four pieces, 00:24:30.654 --> 00:24:32.634 It is four pieces but there's underscores. 00:24:32.634 --> 00:24:32.974 Yeah. 00:24:33.314 --> 00:24:36.784 ILP32 is "Integer Long and Pointer as 32 bits." 00:24:36.824 --> 00:24:41.274 And it's the exact same idea, but now instead of being part of the architecture, 00:24:41.294 --> 00:24:45.634 like arm64_32 instead of being at the beginning of the triple, it's just at 00:24:45.634 --> 00:24:49.204 the end with the ABI like gnu_ilp32. 00:24:49.320 --> 00:24:50.530 But yes, it's the exact same idea. 00:24:50.560 --> 00:24:53.120 It just takes too much memory to have 64 bit pointers. 00:24:53.380 --> 00:24:55.643 We probably don't have more than four giga bytes of RAM anyway, 00:24:55.643 --> 00:24:57.293 so let's just do 32 bit pointers. 00:24:57.293 --> 00:25:01.263 We are finally getting to the part of the episode that's about more or less 68000. 00:25:01.625 --> 00:25:02.805 it used to be used everywhere. 00:25:03.325 --> 00:25:03.845 Pretty much. 00:25:04.135 --> 00:25:06.455 Very popular retro computing target, yeah. 00:25:06.458 --> 00:25:08.645 The Mega Drive was a 68000. 00:25:09.481 --> 00:25:13.001 I had a graphing calculator, the TI-89, and some other models 00:25:13.001 --> 00:25:15.561 of TI calculators were 68000. 00:25:15.561 --> 00:25:17.131 That's how I first learned about it as a kid. 00:25:17.131 --> 00:25:20.961 I printed the entire manual of like 68000 assembler just to learn it. 00:25:20.991 --> 00:25:21.981 It didn't go anywhere. 00:25:22.061 --> 00:25:22.891 I was too young. 00:25:22.931 --> 00:25:24.021 I was like 12. 00:25:24.051 --> 00:25:26.221 I couldn't follow, but I could play Mario on it. 00:25:26.231 --> 00:25:26.741 That was fun. 00:25:26.961 --> 00:25:28.221 This has been a long recording. 00:25:28.221 --> 00:25:29.031 We're almost done. 00:25:29.031 --> 00:25:29.531 Don't worry. 00:25:30.391 --> 00:25:32.571 Next up we have, uh, the China slide. 00:25:32.731 --> 00:25:36.001 There's an architecture called loongarch64. 00:25:36.021 --> 00:25:37.501 I'm not sure that's how you say it. 00:25:37.910 --> 00:25:39.990 It's not loooooong arch?? 00:25:40.158 --> 00:25:41.188 It's not looong. 00:25:41.208 --> 00:25:43.748 At first I was like, "Oh, they made aarch64 longer!" 00:25:43.844 --> 00:25:44.534 Darn it. 00:25:45.203 --> 00:25:48.202 It's actually a Chinese word, I don't know if it's loong, but it's 00:25:48.202 --> 00:25:50.682 definitely like not just loooong arch. 00:25:50.885 --> 00:25:56.325 It started in 2001 as Godson, and it was renamed in 2010 to Loongson. 00:25:56.535 --> 00:25:58.675 at first I was like, "Oh, cool, a new architecture!" 00:25:58.685 --> 00:25:59.655 And then I was like... 00:26:00.055 --> 00:26:03.325 "in March 2023, the United States Department of Commerce added Loongson 00:26:03.345 --> 00:26:05.885 to the Bureau of Industry and Security's entity list for acquisition 00:26:05.885 --> 00:26:08.445 of American technology to support the People's Liberation Army." 00:26:08.615 --> 00:26:09.865 And then you're like, "Oh! 00:26:09.905 --> 00:26:13.445 I accidentally opened the Cold War between China and the U. 00:26:13.445 --> 00:26:13.705 S. 00:26:14.580 --> 00:26:17.650 it's pretty much like China being progressively cut 00:26:17.650 --> 00:26:19.150 off from using American IP. 00:26:19.420 --> 00:26:23.110 So they had the contingency plan like: oh, this is actually our own architecture. 00:26:23.110 --> 00:26:24.780 You cannot prevent us from doing that. 00:26:25.080 --> 00:26:29.890 It is a spiritual successor to like MIPS and RISC V- that they fully 00:26:29.890 --> 00:26:32.030 own so they can just do whatever. 00:26:32.050 --> 00:26:37.835 In November 2023 they debuted the 3a6000 processor fabricated using a 00:26:37.835 --> 00:26:39.875 14 nanometer or 12 nanometer process. 00:26:39.905 --> 00:26:41.885 I love when Wikipedia is like: we don't- we're not really sure. 00:26:41.885 --> 00:26:43.465 We didn't send someone to China to check, 00:26:43.705 --> 00:26:46.785 a lot of the stuff that I've seen is more focused on RISC-V 00:26:46.855 --> 00:26:49.965 because RISC-V also has 64 bit modes and it seems like there's a lot more 00:26:49.975 --> 00:26:53.845 interest in around that, but maybe this is more targeted at the classic x86 00:26:53.845 --> 00:26:58.560 domain where RISC-V is like a different take on ARM and things like that. 00:26:58.705 --> 00:27:00.868 So the Loongson thing it was a research project. 00:27:00.878 --> 00:27:04.308 It was like part of China's policy "Made in China 2025," and it was this 00:27:04.308 --> 00:27:07.658 long plan of like: we're gonna stop just manufacturing other people's 00:27:07.688 --> 00:27:09.828 design, we're gonna have our own designs like the whole thing is gonna 00:27:09.828 --> 00:27:13.868 be like designed and manufactured and assembled and everything in China and 00:27:13.868 --> 00:27:15.898 so that was their thing for processors. 00:27:16.238 --> 00:27:20.138 Next up we have seen target triples with 3 components, as triple 00:27:20.138 --> 00:27:22.348 would suggest, or 4 components- 00:27:22.443 --> 00:27:24.853 I'm very interested to see how you pronounce this. 00:27:24.988 --> 00:27:27.608 But we also have ones with 2 components. 00:27:27.792 --> 00:27:29.002 Especially being in Germany. 00:27:29.132 --> 00:27:30.002 You say it. 00:27:30.422 --> 00:27:32.361 Why should I take the risks, James. 00:27:32.851 --> 00:27:33.631 It's fuchsia. 00:27:33.661 --> 00:27:33.951 Okay. 00:27:34.004 --> 00:27:36.382 In American English, I would say "fyoo-shuh" 00:27:36.782 --> 00:27:37.712 "Fyoo-shuh". 00:27:37.932 --> 00:27:39.662 In German, I'd probably say "fooks-i-a". 00:27:39.927 --> 00:27:41.117 Fuchsia, sure. 00:27:41.247 --> 00:27:42.557 Well, I'm going to spell it out. 00:27:42.607 --> 00:27:49.104 It's f u- f u c- this isn't going where you think it is: f u c h s i a. 00:27:49.354 --> 00:27:54.204 It's Google's permanently in a state of research operating system. 00:27:54.214 --> 00:27:55.144 It is in deployment. 00:27:55.234 --> 00:27:56.774 It's in quite a few devices, 00:27:57.019 --> 00:27:57.645 It is. 00:27:57.655 --> 00:28:00.585 In, some Google Nest Hub devices, according to my research. 00:28:00.597 --> 00:28:01.772 That they've talked about publicly. 00:28:02.077 --> 00:28:03.617 Yeah, yeah, that's true. 00:28:03.707 --> 00:28:05.327 They could be running it more internally. 00:28:05.727 --> 00:28:07.407 And I guess, if you're big enough, it's always good to 00:28:07.417 --> 00:28:08.867 have your own everything, right? 00:28:08.917 --> 00:28:10.689 Including your own operating system. 00:28:10.703 --> 00:28:13.523 But yeah, very annoyingly their targets are only two components. 00:28:13.523 --> 00:28:15.303 They have like aarch64-fuchsia. 00:28:16.083 --> 00:28:18.643 I'm just going to say it the French way because that's who I am. 00:28:19.102 --> 00:28:21.347 And underscore 64-fuchsia... 00:28:21.347 --> 00:28:23.780 I- - I have nothing more to say about this. 00:28:23.780 --> 00:28:24.330 Next up! 00:28:24.530 --> 00:28:28.090 We have other arches, when you have three components, you're 00:28:28.100 --> 00:28:29.020 like: which one did they omit? 00:28:29.030 --> 00:28:30.210 Did they omit the operating system? 00:28:30.210 --> 00:28:31.326 Did they omit the ABI? 00:28:31.536 --> 00:28:35.637 . So UEFI is the thing that like, has replaced the BIOS, I would say. 00:28:35.637 --> 00:28:39.907 The thing when your computer starts up, it's running in UEFI mode, you 00:28:39.907 --> 00:28:42.803 can adjust the speed of your fans and everything, you can build for that. 00:28:42.843 --> 00:28:44.853 You can write Rust code that will run in that mode. 00:28:45.193 --> 00:28:48.453 There's Redox, which is the kernel written in Rust. 00:28:48.728 --> 00:28:51.948 There's aarch64-nintendo-switch-freestanding, so 00:28:51.978 --> 00:28:54.068 that's not an ABI, that's the lack of one. 00:28:54.068 --> 00:28:56.698 It's like, you don't have a libc, you have nothing, good luck. 00:28:57.903 --> 00:28:58.243 Yeah. 00:28:58.348 --> 00:29:00.377 It's fun that there's a Nintendo Switch target. 00:29:00.546 --> 00:29:01.506 It's basically none. 00:29:01.536 --> 00:29:03.746 I don't know why they went with freestanding instead of none, but 00:29:03.746 --> 00:29:04.726 I'm sure they have a good reason. 00:29:04.868 --> 00:29:05.728 I don't know. 00:29:05.728 --> 00:29:07.038 There's another Nintendo one. 00:29:07.038 --> 00:29:11.238 It's armv6k-nintendo-3ds and then no ABI. 00:29:11.728 --> 00:29:19.218 In terms of game consoles, there's also armv7-sony-vita-newlibeabihf. 00:29:19.918 --> 00:29:23.958 We'll get into what that means right now, actually, because if you look at the ARM 00:29:23.958 --> 00:29:26.468 target, so that would be armv6, right? 00:29:26.468 --> 00:29:29.858 You have arm-unknown-linux-gnueabi. 00:29:30.158 --> 00:29:31.168 And then the same thing. 00:29:31.473 --> 00:29:35.083 But E A B I H F, and the H F stands for: James? 00:29:35.713 --> 00:29:36.243 Hard float. 00:29:36.243 --> 00:29:39.383 Hard float, which means that the one that doesn't have hard float 00:29:39.513 --> 00:29:42.743 has software-emulated floating points. 00:29:42.873 --> 00:29:47.098 So whenever you use F32, actually doing all the multiplications and divisions 00:29:47.098 --> 00:29:50.298 and everything, like, in software instead of just using a CPU instruction. 00:29:50.568 --> 00:29:52.308 That's true, but the reason why it shows up in the 00:29:52.338 --> 00:29:57.328 ABI is because those processors have floating point registers. 00:29:57.568 --> 00:30:01.058 So like you normally have integer registers that your processor uses. 00:30:01.238 --> 00:30:04.768 You have an additional set of floating point registers on there, which you don't 00:30:04.778 --> 00:30:08.498 necessarily have to use for floating point, but the ABI changes because when 00:30:08.498 --> 00:30:14.538 you pass, for example, a float by value in a function call, on a soft float target, 00:30:14.548 --> 00:30:16.458 you would put that in normal registers. 00:30:16.708 --> 00:30:19.988 If you had a hard float target, you would put that in there. 00:30:19.988 --> 00:30:24.608 So if you had a function that took a U32 and an F32, the ABI on 00:30:24.608 --> 00:30:28.528 the soft float target would be to put those in regular registers. 00:30:28.933 --> 00:30:31.278 On the hard float target, it would be to put the integer in the 00:30:31.278 --> 00:30:34.918 regular registers and the float in the floating point registers. 00:30:35.138 --> 00:30:38.348 So that's why it's in the ABI field, because you're right: you have to also do 00:30:38.348 --> 00:30:42.928 all the software-emulated floating point, but the reason that it shows up in the ABI 00:30:42.928 --> 00:30:48.198 field is because you get basically a whole bonus set of registers available there. 00:30:48.321 --> 00:30:51.634 Right, I'm tempted to ask why even have floating 00:30:51.634 --> 00:30:55.094 point specific registers as opposed to just having general purpose. 00:30:55.144 --> 00:30:57.484 I'm imagining it's like the simplicity of the design 00:30:57.574 --> 00:30:58.104 Yeah, I don't know. 00:30:58.104 --> 00:31:01.584 All of the floating point instructions will draw from the floating point 00:31:01.594 --> 00:31:04.214 registers and they also have some capabilities of like... 00:31:04.264 --> 00:31:08.284 how normally if you have 32 bit registers and you have a 64 bit value, you put 00:31:08.284 --> 00:31:12.894 them in two neighboring registers and floats are the same way for F32 and F64s. 00:31:13.444 --> 00:31:13.994 Why? 00:31:14.034 --> 00:31:14.774 I don't know. 00:31:14.824 --> 00:31:17.204 Probably because it's easier on the silicon... 00:31:17.204 --> 00:31:18.154 but yeah, I don't know. 00:31:18.184 --> 00:31:21.934 Yeah, cause like on ARM64, if you've reinterpret the bits of a float 00:31:21.934 --> 00:31:24.564 as an integer, that's a no op pretty much. 00:31:24.564 --> 00:31:27.945 It's like, it's just a type change from your compiler's perspective? 00:31:27.945 --> 00:31:30.884 I think it copies it from the float registers to the... 00:31:30.900 --> 00:31:31.895 and the integer registers. 00:31:31.904 --> 00:31:32.094 Yeah. 00:31:32.094 --> 00:31:34.914 On those processors, it would have to copy it to a different register. 00:31:35.178 --> 00:31:36.878 It's probably lazy, like until you actually 00:31:36.878 --> 00:31:38.138 do something integer-y with it- 00:31:38.138 --> 00:31:38.388 Probably. 00:31:38.388 --> 00:31:38.898 - it does nothing. 00:31:38.898 --> 00:31:43.468 But if you want to do integer ops on it, I think you have to move it from the float. 00:31:43.573 --> 00:31:47.137 Because the instruction encoding, you only have a couple bits to describe 00:31:47.137 --> 00:31:48.517 what register you're talking about. 00:31:48.757 --> 00:31:52.117 So if you had just more registers, you'd have to have more bits to encode 00:31:52.127 --> 00:31:53.787 which register you're sourcing from. 00:31:53.930 --> 00:31:55.120 That makes sense. 00:31:55.330 --> 00:31:58.340 So that's the hard float part, the hf part of those targets. 00:31:58.340 --> 00:31:59.500 And then there's also E ABI. 00:31:59.500 --> 00:32:00.760 I was curious what the E stands for. 00:32:00.760 --> 00:32:02.220 It stands for Embedded ABI. 00:32:02.508 --> 00:32:07.633 It was the successor of O ABI, which has been retronamed the Old ABI? 00:32:08.053 --> 00:32:10.923 Surely when they made it, they didn't call it Old ABI. 00:32:11.143 --> 00:32:12.193 What happened there? 00:32:12.420 --> 00:32:14.667 It's been gone for more than a decade, but yeah. 00:32:14.918 --> 00:32:19.058 This is specified in the AAPCS, which is the ARM Architecture 00:32:19.058 --> 00:32:21.378 Procedure Call Standard 32. 00:32:21.593 --> 00:32:23.963 And my next episode, we'll be going through every page of 00:32:23.963 --> 00:32:27.203 the AAPCS and recording for 17 hours. 00:32:27.274 --> 00:32:28.004 It's a good doc. 00:32:28.014 --> 00:32:30.054 It's actually on GitHub as a Markdown doc. 00:32:30.064 --> 00:32:30.604 It's good. 00:32:30.829 --> 00:32:31.729 I'm sure it is. 00:32:32.349 --> 00:32:36.689 next up is nvptx64-nvidia-cuda. 00:32:36.949 --> 00:32:39.159 I don't know if that's still a real target. 00:32:39.699 --> 00:32:43.669 For a while there, it was something you could build to, so that you could run 00:32:43.669 --> 00:32:45.519 code directly on your graphics card. 00:32:45.519 --> 00:32:49.779 But the thing is GPUs don't really look like CPUs. 00:32:49.819 --> 00:32:53.559 They have assembly, they have instructions, they have pipelines, but... 00:32:54.049 --> 00:32:58.149 they have like way too many registers, they do everything in parallel... 00:32:58.819 --> 00:32:59.429 it's complicated. 00:32:59.429 --> 00:33:03.569 I know there's definitely ways to write Rust code for the GPU, but I don't 00:33:03.579 --> 00:33:05.239 think it's through that target anymore. 00:33:05.239 --> 00:33:07.139 I might be mistaken. 00:33:07.139 --> 00:33:08.139 James, you're nodding. 00:33:08.199 --> 00:33:08.859 I have no idea. 00:33:08.859 --> 00:33:11.729 I know that like shader compilation is a thing and I know some folks like the 00:33:11.729 --> 00:33:16.180 Rust folks and stuff were working on that but that's totally out of my wheelhouse. 00:33:16.247 --> 00:33:19.741 I don't have the notes here, but readme for a project related to the 00:33:19.741 --> 00:33:25.201 nvptx64 target says: we got it working one day and then three days later they added 00:33:25.385 --> 00:33:30.265 support for 128 bit wide integers and that broke the targets, but unfortunately 00:33:30.265 --> 00:33:33.185 they did not release a nightly rustc version between those two dates. 00:33:33.445 --> 00:33:36.305 So there is actually zero versions of rustc that actually 00:33:36.305 --> 00:33:37.545 still builds this target. 00:33:37.691 --> 00:33:40.101 I don't know if that's still accurate, but I thought that was very funny. 00:33:40.414 --> 00:33:43.304 We have s390x-unknown-linux-gnu. 00:33:43.474 --> 00:33:45.694 That's IBM System/390. 00:33:45.714 --> 00:33:48.654 I know very little about this because it's mainframe, and I've 00:33:48.654 --> 00:33:50.094 never worked with mainframes. 00:33:50.094 --> 00:33:52.014 But it's a computer the size of a room? 00:33:52.024 --> 00:33:53.124 Not anymore, I guess. 00:33:53.144 --> 00:33:55.291 It's just the size of a big, like... 00:33:55.406 --> 00:33:56.356 where do you put your clothes? 00:33:56.866 --> 00:33:58.416 A big closet, right? 00:33:58.786 --> 00:34:00.110 Yeah, very pricey. 00:34:00.220 --> 00:34:01.760 Yeah, I ordered a couple as well. 00:34:02.190 --> 00:34:06.810 And, uh, you can run System Z, which is IBM's operating system for mainframes, 00:34:06.820 --> 00:34:08.790 or you can run Linux on there. 00:34:09.130 --> 00:34:10.710 And so that's what this target is for. 00:34:11.445 --> 00:34:13.115 We have the WebAssembly targets. 00:34:13.325 --> 00:34:16.205 The first one we saw was wasm32-unknown-unknown, 00:34:16.258 --> 00:34:17.548 uh, which has nothing. 00:34:17.788 --> 00:34:18.818 It has no operating system. 00:34:18.818 --> 00:34:19.618 It has no ABI. 00:34:19.628 --> 00:34:20.448 It has nothing. 00:34:20.458 --> 00:34:21.728 It can compute things. 00:34:21.755 --> 00:34:22.787 And that's all it can do. 00:34:22.817 --> 00:34:25.280 This is our none or freestanding target, basically. 00:34:25.280 --> 00:34:25.510 Yeah. 00:34:25.574 --> 00:34:30.438 So if you want to generate random numbers, you have no random number source. 00:34:30.448 --> 00:34:31.068 So good luck. 00:34:31.088 --> 00:34:34.308 If you want to time something, you have no timers, you have nothing. 00:34:34.668 --> 00:34:38.518 So what you do is because you're probably calling this from JavaScript. 00:34:38.648 --> 00:34:41.638 You go back and forth between your JavaScript world and the WebAssembly 00:34:41.638 --> 00:34:45.495 world, which results in your WebAssembly powered thing probably 00:34:45.495 --> 00:34:47.705 not being any faster because you're doing all those translations. 00:34:47.705 --> 00:34:49.205 You're crossing the bridge all the time. 00:34:49.544 --> 00:34:52.764 And so they made a different thing called wasm32-wasi. 00:34:52.784 --> 00:34:56.196 We have the return of triplets with only two components. 00:34:56.371 --> 00:34:57.891 And that one is also versioned. 00:34:57.901 --> 00:35:02.251 I did not include all the other ones because I predicted we would be tired 00:35:02.251 --> 00:35:06.181 by this point, but there's preview one preview two there's there's so many 00:35:06.181 --> 00:35:11.487 different things happening but they are adding apis for file io network io, 00:35:11.497 --> 00:35:14.547 timers, random number generation, all you would expect from like a kernel. 00:35:14.987 --> 00:35:18.607 they're adding syscalls essentially to WebAssembly so that you can 00:35:18.607 --> 00:35:19.787 actually do useful things with them. 00:35:20.487 --> 00:35:22.367 We are nearing the end of the presentation. 00:35:22.527 --> 00:35:23.507 I want to reassure you. 00:35:23.717 --> 00:35:27.027 we've seen arm64_32-apple-watchos. 00:35:27.047 --> 00:35:28.837 There's a couple other WatchOS targets. 00:35:28.847 --> 00:35:32.807 There's aarch64-apple-watchos, which is what I have on my 00:35:32.807 --> 00:35:34.027 wrist, Which you cannot see. 00:35:34.027 --> 00:35:38.250 Even if you go to sdr-podcast.com/episodes, There's 00:35:38.250 --> 00:35:43.400 aarch64-apple-watchos- sim, which is what you get if you run the WatchOS 00:35:43.410 --> 00:35:46.270 simulator on an Apple silicon Mac. 00:35:46.480 --> 00:35:52.470 And then there's x86_64-apple-watchos-sim, which is what you get if you 00:35:52.470 --> 00:35:56.440 run the WatchOS simulator on an Intel Mac, I would imagine. 00:35:56.460 --> 00:35:58.470 I didn't check, but that's the only thing that makes sense to me. 00:35:58.698 --> 00:36:01.118 And then we have our final target, which I love very much, 00:36:01.118 --> 00:36:06.018 which is x86_64-apple-ios-macabi. 00:36:06.628 --> 00:36:07.888 James, do you have any idea what that is? 00:36:08.086 --> 00:36:09.124 I have no idea what 00:36:10.056 --> 00:36:12.616 not Maccabi, I said it wrong on purpose to be funny. 00:36:12.866 --> 00:36:13.886 It's MacABI. 00:36:14.047 --> 00:36:14.982 Does that help a little bit? 00:36:15.062 --> 00:36:18.715 I can imagine, but I have no idea why they have, like, seven 00:36:18.715 --> 00:36:23.215 flavors of ABI's that they'd like to play with, just to make programmers sad. 00:36:23.555 --> 00:36:27.138 No, it's to run iPad apps on MacOS. 00:36:27.373 --> 00:36:27.828 Okay. 00:36:27.850 --> 00:36:30.820 Remember a while ago when Twitter made a new app back when it 00:36:30.820 --> 00:36:33.058 was still called Twitter and they were like, "We have a desktop app now!" 00:36:33.058 --> 00:36:34.828 And it wasn't like a Mac app. 00:36:34.868 --> 00:36:39.408 It was the iPad app And that's a technology Apple calls Catalyst. 00:36:39.718 --> 00:36:41.278 And that's the target for that. 00:36:41.474 --> 00:36:41.904 Gotcha. 00:36:41.956 --> 00:36:42.886 we are finally done. 00:36:42.956 --> 00:36:46.376 Thank you for following me on this journey through Target Triples. 00:36:46.566 --> 00:36:48.296 We didn't even talk about some of the funny ones. 00:36:48.426 --> 00:36:51.525 There's QNX in there somewhere, there's AIX, there's A lot of things happened. 00:36:51.535 --> 00:36:53.685 It's fascinating to me because it's history because you look at them 00:36:53.685 --> 00:36:54.725 and you're like, "Oh, that's right! 00:36:54.725 --> 00:36:55.465 We used to do that. 00:36:55.465 --> 00:36:56.095 And now we don't. 00:36:56.215 --> 00:36:57.135 And it's better now." 00:36:57.217 --> 00:36:58.967 I want aarch64 all the things now. 00:36:58.997 --> 00:37:00.097 I don't, you know... 00:37:00.211 --> 00:37:03.271 Every target triple made sense when it was written, the problem 00:37:03.271 --> 00:37:06.201 is that it's been written over 40 years, which means the things that 00:37:06.201 --> 00:37:08.301 made sense then, no longer make sense. 00:37:08.301 --> 00:37:11.242 Because we talked about Thumb Mode and it reminded me of the fact that 00:37:11.242 --> 00:37:14.892 even today my desktop computer, when it wakes up, is in 16 bit mode and then it 00:37:14.892 --> 00:37:18.772 needs to switch to 32 bit mode and then to 64 bit mode and then maybe it's awake. 00:37:18.822 --> 00:37:19.842 ARM's gone the opposite way. 00:37:19.842 --> 00:37:24.372 Newer versions of aarch64 can't run thumb instructions anymore to simplify the core. 00:37:24.432 --> 00:37:26.092 They can only support ARM instructions. 00:37:26.102 --> 00:37:29.712 So Apple's actually going the opposite way in that they're removing 00:37:30.152 --> 00:37:32.892 legacy support because otherwise the cool thing about ARM is it's kind 00:37:32.892 --> 00:37:35.532 of like the 386, 586, 686 stuff. 00:37:35.582 --> 00:37:39.872 you can run Arm V four on V five or six or seven or eight, and I think V 00:37:39.872 --> 00:37:42.932 eight is the first one that really broke something, in that it just said you 00:37:42.932 --> 00:37:45.482 can't run thumb instructions anymore. 00:37:45.492 --> 00:37:47.372 So that's target triples for you. 00:37:47.782 --> 00:37:48.682 They're not triples. 00:37:49.182 --> 00:37:50.722 Don't let them tell you that. 00:37:51.116 --> 00:37:52.273 This is a good conclusion, right? 00:37:52.313 --> 00:37:53.493 This is what podcasts are about? 00:37:53.673 --> 00:37:54.033 All right. 00:37:54.403 --> 00:37:54.973 Bye everyone. 00:38:00.916 --> 00:38:03.196 This episode is sponsored by CodeCrafters. 00:38:03.384 --> 00:38:06.654 CodeCrafters is a service for learning programming skills by doing. 00:38:07.324 --> 00:38:10.614 CodeCrafters offers a curated list of exercises for learning programming 00:38:10.614 --> 00:38:13.624 languages like Rust or learning skills like building an interpreter. 00:38:14.099 --> 00:38:17.639 Instead of just following a tutorial, you can instead clone a repo that contains 00:38:17.649 --> 00:38:21.379 all of the boilerplate already, and make progress by running tests and pushing 00:38:21.379 --> 00:38:24.719 commits that are checked by the server, allowing you to move on to the next step. 00:38:25.359 --> 00:38:28.329 If you enjoy learning by doing, sign up today using the link at 00:38:28.329 --> 00:38:32.449 sdr-podcast.com/codecrafters, or use the link in the show 00:38:32.449 --> 00:38:33.679 notes to start your free trial. 00:38:34.099 --> 00:38:36.619 If you decide to upgrade, you'll get a discount and a portion of 00:38:36.619 --> 00:38:38.159 the sale will support this podcast. 00:38:38.604 --> 00:38:42.294 That's sdr-podcast.com/codecrafters. 00:38:42.994 --> 00:38:45.374 Thanks to CodeCrafters for sponsoring this episode.