ไม่สามารถเล่นวิดีโอนี้
ขออภัยในความไม่สะดวก
Computers Barely Work - Interview with Linux Legend Greg Kroah-Hartman
ฝัง
- เผยแพร่เมื่อ 3 มิ.ย. 2020
- + Build Video: • Building a Whisper-Qui...
+ Gskill 256GB RAM Vid: • Checking Out GSkill’s ...
+ Forum Thread: forum.level1te...
+ Greg's Reddit AMA: / im_greg_kroahhartman_l...
**********************************
Thanks for watching our videos! If you want more, check us out online at the following places:
+ Website: level1techs.com/
+ Forums: forum.level1tec...
+ Store: store.level1tec...
+ Patreon: / level1
+ L1 Twitter: / level1techs
+ L1 Facebook: / level1techs
+ Wendell Twitter: / tekwendell
+ Ryan Twitter: / pgpryan
+ Krista Twitter: / kreestuh
+ Business Inquiries/Brand Integrations: Queries@level1techs.com
IMPORTANT Any email lacking “level1techs.com” should be ignored and immediately reported to Queries@level1techs.com.
-----------------------------------------------------------------------------------------------------------
Intro and Outro Music By: Kevin MacLeod (incompetech.com)
Licensed under Creative Commons: By Attribution 3.0 License
creativecommons...
"There's nothing scarier than a software engineer with a soldering iron"
This man *knows* what's up. Mad respect Gergkh, mad respect.
Those folks are called firmware engineers. ;)
Never seen Wendell so comfortable in his own space... this is my people vibe.
Great interview, thoroughly enjoyed it.
The intro did not have its normal music. I stopped and restarted several times, tweaking with sound settings trying to figure out why it wasn't working. Turns out there was no sound because there was no sound.
@@TomBorowski real talk, pulse audio has become amazingly stable
@@miawgogo If your drivers do. Somehow every single PC/laptop I own either has sound problems, stops working after waking up, or doesn't work at all.
But maybe in a few days I could fix it and get this into kernel. One guy (Conmanx360) brought support for last 15 years of Creative cards, with no prior knowledge
@@TomBorowski I wanted to make a pulseaudio -k joke, haha
I have waited for this interview. This is my favorite kind of content.
this 'content' is straight magic. what is happening!? lol
When I gave an Chromebook 5 years ago to my wife, I told her too "when it breaks, it's cheap".
The "damn thing" has been working fine until today. Acer Chromebook 13 with a cheap ARM chip... :-)
What brought me here was the title. As someone who was a systems engineer back in the 90s I can say, the title is true. Most people don't understand some of the interesting physics going on in computer. There was a time when balancing a 66MHZ PCI Bus (balancing means make it stable) was considered impossible, beyond the human capacity to solve for. The same thing goes for memory buses and clock timing. Now, we have progressed beyond that point but much of what is keeping your system working is on the level of "beyond human capacity to solve for". We use software to solve the engineering issues. It's far from trivial. When you are talking about the clock speeds we are at today trace capacitance and length of trace are required to be.. perfect. This bleeds into everything, the NVME issues as you say, the internal engineering required to make the hardware function. I mean, most people alive today don't recall the time when peripheral devices in a PCI bus you had to swap them around to find out what worked in what slot due to the timing/balance issues. It's really no different internally to your NVME device, your graphics cards, or even your USB devices. Also, the standards are.. ugh. Some things never change. It was hell in the 90s too.
As for solving calculus or linear algebra, yeah fun stuff when you realize those variances can actually cause different end results.
Half an hour in and this might be one of my fave linux interviews - great job Wendall really appreciate it
This is the content I need in my life
This is why open source hardware is important
🙌 🙌 Good software support for Threadripper incoming !
Nice to see Wendel "fan-girling" hard 😁 Great interview.
I read on Twitter that there was an interview - have been checking and waiting for this every few hours, on both channels :)
This is so cool. Way above my head but interesting is still interesting.
Great interview Wandell! It's always a pleasure to watch your interviews. It's lovely when passionate people are coming together.
Just want to correct something regarding the Afterburner card for Mac Pro 2019. That card is for decoding ProRes/ProRes Raw streams not for helping with visual effects (unless you count video as visual effects :] ). Btw, it's an Intel/Altera FPGA. (I did a bit of FPGA developing while studying computer science and I really wish that they will be adopted in more mainstream hardware, they are really cool.)
"Nothing scarier than a software eng with a soldering iron" , how about a hardware eng with a C compiler ! What could possibly go wrong ?!?
*What could possibly go right. XD
Looking at you Toyota...
Actually not really It depends on the developer self. Hardware developer work with C a lot. Maybe other languages might not be a good thing.
I always learn something new when I listen to Greg KH. Thanks for this interview!
i love the conversation/interviews that you do wendell. you have managed to get some pillars of the community, awesome! Hello to GKH, and thank you so much for your time.
Very insightful. Was always wondering what to work on in the kernel since it is soooo big. I appreciated that conversation a lot. Thanks!
Best part of this whole episode is when Wendell build|gifted GPU with couple of bugs & GKH generously accepted it with open heart. What else a kernel developer needs is couple of bugs and Wendell gifted the same to add up some, in his ToDos .
Most detailed and most personal chat with Greg I've seen-such an engaging and fun listen and watch. Thanks very much!
This was awesome. More like this pls. I love this guy.
Thank you for this interview, such a entertaining and interesting insight to that area. Now we all want more!! 🙂
Wendel, please keep making more videos like this! You are awesome.
Awesome interview, would definitely watch/listen to a podcast by these two
Well it's very touching to see you excited and nervous as you interview somebody we all admire so much.
Kernel developers are unsung heroes. They work on software that if it works you don't even know it's there. They need to know how hardware works in order to make sure their software works but do not design the flashy hardware everyone sees.
I am really curious now about Greg's favorite keyboard shop in Hong Kong and Tokyo. I would pay them a visit, well sometime in 2034 anyway, when they lift travel bans.
If it's Hong Kong, by that time they might be part of China and who knows what it will be like... ;-)
Just go to Akihabara, there are lots of shops with hundreds of keyboards which you can try in person before buying. Mouses too. If you're in Osaka go to Denden Town.
Yeah, that's the answer to why kernel has 30 million lines. You hear some ignorant people just complaining about how kernel is bloated and how OS-es in general are not required at all and let's just run programs on bare metal like in the good ol' days. Well, here's a news flash for you: hardware has become much more complicated and much more varied since the f-in 80's. Whatever worked back then would not scale to the ecosystem we have today. A user buys hardware in a store, they expect it to work. They don't care how. It's so easy to be ignorant of all these problems when someone else is working tirelessly so that you don't have to know.
This is a wonderfully geeky interview. Thank you!
Excellent to not only to build the mini super computer, but to get a really in depth interview. Loved watching this episode.
Huh, I finally got around to watching the patreon version of this last night. Awesome stuff!
What a wholesome, well done interview. many thanks wendell !
I learned to dread 80% of IT management. the more you learn about them, the more you find they don't understand. the part I really did well, was interfacing everything and getting it into production, without it being a nightmare to support. the management would almost NEVER support putting in production monitoring /support software and a competent admin support staff. you deploy it, you support all calls on it. to many companies are at this level of support.
Its greg the legend
15:35 "We just know, what would work really well and so we ... Ryzen."
This was what I wanted today, I needed intelligent content. Wendell, great interview and awesome content, it was over my head and very interesting. Thank you, stay safe and keep well...
A desktop application is an abstraction which is an abstraction of a JIT language, which is an abstraction of the OS, which is an abstraction of good and bad hardware alike.
Break through these abstractions and appreciate the hardware for it's quarks and the hard work that the Linux kernel devs put into giving you the experience you have today!
"desktop application is an abstraction which is an abstraction of a JIT language" This is a modern trend that I hate. I miss the days of native desktop applications. I feel like people need to learn to code in a "real" language and leave the Javascript toys at home. Still...I have it better than most. I'm using Gentoo so just about everything that I run is a native application. When I reboot into Windows to play video games, I'm appalled at the resources and lag of what should be simple programs.
Windows didn't used to be like this. 10 years ago if you opened a windows application, it was probably written in C or C++, and running natively on the actual hardware.
You need to have this guy back on
great interview! I hope there's going to be a follow-up
Totally engaging - thanks to both of you!
I loved this video, amazing to get a view of how things work.
woooo
Such a great video Wendell! Keep these coming!
I can't think of a better interviewer for GKH.
Very awesome, indeed! Thanks for posting!
THIS IS ONE OF THE TOP VIDEOS I HAVE WATCHED FROM YOU! THANK YOU!
I still have a working Abit BH-6.
i wish wendell would do more of this type of videos... absolutely enjoyed it!
Linus just lifted the curtain on certain components he chose for his system, but he did get worse RAM and slower storage than GKH. :)
super interesting interviews. Looking forward to see more :)
You've gotta have him back for another one. Great interview!
Great interview!
I know i'm late to comment here, but i just wanted to chime in with a big +1 recommendation for considering studying Embedded Systems, or doing a project or course in it. Learning about the hardware the software runs on gives you a much deeper appreciation and understanding of what is actually going on, and gives important context for many kinds of problems and systems. Just having a sense of scale of the different resources and performance domains lets you avoid making some classes of mistakes in code, or be able to skip the first 2-3 rounds of benchmarking and performance tuning (or depending on use-case not even have to) because you inherently understand the characteristics of your code and hardware well enough that you can write semi-high-performance code from the get-go. And on the other hand, also give you some understanding of where and when you need to care about performance and not. 1-10 ms doesn't matter in most user-interaction scenarios or in a top level control flow procedure, but in a nested loop scenario that can execute thousands of times it can be the difference between a usable and unusable program/solution.
Ok, this was just awesome! Thanks Wendell and Greg!
"My wife was not happy about that one."
I know exactly what that had to felt like. Got into similar situation too :-)
Thank you, Wendell and Greg.
Best technical chat ever!!! Thank You
I know some of these words. I love this content. It expands my horizons and gets me excited to keep up my nerdy hobby.
this was really interesting :D more of these kinds of videos :D
Very interesting video, would like to see more of this!
Great interview, think I'll compile the kernel for the first time this weekend!
Never!
Man I love this interview
I wouldn’t mind a 102-button mouse :)
Imagine dual wielding two of them. That would be 204 buttons!
I'm curious if Greg has any thoughts on the Mill CPU architecture?
Is it still being researched?
@@lex4o It is not researched or a research project really. It is being designed and slowly put into actual silicon design, not a specific node yet, but later the better, because they still want to be on a leading edge manufacturing node, but that is changing every year, but slowly flattening. So they are postponing doing that part when they have all the rest of the design finalized mostly (including having full compiler, Linux kernel, microkernel hypervisor and userland working and all with intended performance).
I am really interested in it, as it has really good single threaded performance, very high code density, and syscalls, context switches, interrupts and functions calls are so freaking fast (1 cycle, ignoring any cache misses). Not freaking 500-2000 cycles on x86.
movax20h it really has all these advantages? I thought the only up on x86 it had was safety features
@movax20h thanks for the awesome reply. It's a radical design that could really turn things on its head. The belt and maybe portal calls are two stand outs for me.
Reading through the Mill wiki, I get a sense of "this is how computers are meant to work". Per-operand metadata, temporal register addressing, split protection and address translation, compactified instructions, decentralized compute slots. It's all amazing stuff that I really hope catches on.
What a great interview. Loved it.
Great interview. Really enjoyed it.
Thanks for that, Brilliant.
I like these videos. If it would be possible I would like to see more of them!
He's so fun to listen to!
Fantastic interview.
Great interview, Thankyou!
Google's Crostini Project brought LXD containers running Linux to ChromeOS. It's a very useful feature, although still has some limitations (no access to video or audio hardware). I bought my latest Chromebook mostly for the 10+ hour battery life, so I can take it with me for an entire day without worrying about a recharge. The small size of the persistent storage available (mine has 32 GB) continues to be its biggest drawback. With its painfully slow access times, the removable SD Card slot doesn't really provide any relief.
Great interview.
Really hope he likes the machine too.
Great talk, thank you both! :)
2:52 Our evil-genius Bond villain stereotype listens to the magic words....
I greatly appreciated this and learned more than I wanted to know.
fantastic, thank you Wendell
For anyone looking to build a PC form factor Linux workstation on RISC-V, I'm joining your company.
Call up ASRock or Gigabyte or whoever, figure out how to design the socket to make it easy for them to make an ATX board around it with standard peripherals. Make an SoC with the same facilities as a standard PC SoC like Ryzen. Write the board firmware for them, make the interop documents public if it doesn't sacrifice your board partner relationship. See about employing somebody to write a lightweight DBT runtime for i686 applications now that the patents for SSE2 are on the death bed.
Whoever builds this platform could win really big, and if they don't, the human resources left over will be a massive value for other ventures.
The trick for any RISC-V or ARM to be usable, it must: 0) Be affordable, even if it means shipping a silicon with some cores disabled, or lower clocks. 1) Have PCIe slots. 2) Have DIMM / SO-DIMM slots for memory, 3) Support UEFI, 4) Allow booting from USB FAT32 and from network (DHCP+TFTP) with no extra tricks, 5) Make it a small ATX board accepting 12V from normal ATX PSU. IBM POWER finally made it all possible and does work and using it as a workstation is now possible. You can't do the same with any ARM or RISC-V systems. Some ARM systems almost approach it (Rasberry Pi 4 finally solved point 4 above, and has 8GB version now, but it fails on other fronts). ARM server stuff fails on points 0 and 5 above. RISC-V fails on many other fronts.
=\ The left audio out on my Creator TRX40 shorted out and is now much quieter than the right channel, I'm going to have to rip apart my entire build and RMA the thing. =(((
Weird enough, 1752 was also a leap year beginning on a Wednesday and has the same calendar as 2020 until September 14th.
No mention of Mill computing processors. Weird...
Thanks Wendell! What a great interview:D
Great Interview
This was great, Greg seems like a cool guy.
Watched it a second time just to see you at the brink of fangirling for 50 minutes ;) Thanks again for sharing
Nothing scarier than software engineer with soldering iron? What about electronics technician with IDE and compiler?
Wait, that might actually be a good thing.
You need two things to create a universe. First, a turtle. Second, recursion.
Stay awesome.
1 hour went fast without noticing , great interview
My first into to the name Greg in Linux was when i got in to Android kernel development when i was 12 (& did till 14). I was trying to port Doble Tap To Wake and other stuff and no matter where i looked, be it google or HTC or other kernels for stuff, the version inclements were always done by GHK and that's when i found him as another core member other than Linus who was involved not in Android, but Linux kernel as a whole (Well, Some CI bots were there at the top too :P ). Mad respect for him.
Thats a nice background you got there.
btw, Turemetal makes fanless cases that can take these high-end CPUs and GPUs.
Hey Wendell, great video, but I am really interested in the gadget you're having in your hands. I am smoking way too many cigarettes on a normal workday and I've found out it's mostly because when I'm not typing I want some distraction. Any Info on that :)?
Thanks
Took over "everything" but the desktop, aka the most used performance oriented piece of hardware.
All software is inherently bloated in some way or another, but the lack of acknowledgement of the sheer complexity of even the most simple programs like "hello world" is appalling and the drive to reduce that complexity is nonexistent. Something must be done about it before the damage is irreversible.
Good comment
I'm in the same boat. I'm not sure where to start. The suckless project seems to be stalling and not fixing the fundamental issues, and the kernel itself has so much bloat too...
@@timh.6872 I like a lot of what suckless does, except they have nonsensical ideas of what is “simple”. In general among many C devs, there's a cargo-cult-like aversion to C++.
I'm thinking of trying to transition to Gentoo and rolling my own stuff.
I don't know, you seem to be operating from the wrong premise. Yes there is a lot of underlying complexity that is executed alongside Hello World that Hello World doesn't per se need; but it also doesn't actually harm Hello World very much, you can still run it as quickly and as often as if it was much lighter weight. This underlying complexity exists in order to support COMPLEX software which has significantly tougher requirements, it needs a big world to operate in, and if you don't provide it this world, it would have to invent its own, and another complex piece of software would invent its own too, and when two are run together, even if they were individually lighter weight, their impact would be compounded. Furthermore it exists in order to provide many way compatibility, so the complex software doesn't have to be tied to a specific hardware, doesn't hardcode too many assumptions about hardware, which you would agree is a good thing, since all this effort creating this software is not invalid when you buy new computer components every couple of years - even if it's technically bloat and results in layers of complexity underneath. But as it happens, when the complexity is overburdening and leads to its own issues, like performance issues, resource consumption issues, compatibility issues, you're by then necessarily already dealing with software that is fundamentally more complex, with software which is generally not Hello World.
So maybe reducing complexity is a good idea, but you can't start from the bottom, you can't start from Hello World. You have to start from the top, from the most capable and complex packages possible - which also makes it a borderline insurmountable task to do systematically. Or maybe ones you use a lot, a lot of instances etc, so the impact adds up a lot - but if they're simultaneously infrastructure, like a library or a standard, you don't want this infrastructure to become practically less capable, you don't want 10 other packages to reinvent things just because you had to simplify your end. So what i can say - there is no fundamental solution, there is no silver bullet. Sure you should chip away at extraneous complexity and try to maximise software quality in any way possible, but when you do, you probably want to focus the effort to do it where it counts, rather than on Hello World, because the grand total sum of human effort is limited.
@@SianaGearz In education, yes you MUST start at the bottom to properly set expectations for future devs to be like "here's how we can write a simple program that can print a string on the screen and can also be easily loaded onto the boot sector" and not have it be 150kb in size for the executable. It's all about setting the expectation for absolutely maximum resource efficiency no matter the cost because it just works given the tradeoffs because that's just what engineering is all about. There will never be "the perfect solution for everything", most likely it'll be "solutions that best fit the problem given all these conditions" with the added benefit of better security because less code that can be exploited by 1337 h@x0r5.
In the entrepreneurial world, you would have to advertise some weird Rube Goldberg machine that is likely to simplify most of the complexity that is running the world when it is just a simple MxN interlocked ring buffer machine under the hood (as an over simplified example). I'm not saying people should temper their expectations for "muh features/ hardware drivers", but they're not paying for software/SoC by the line of code either. Gameifying software engineering to be less of a waste of time for the user, the developer, and the CPU would help us all. I'm just not optimistic that people will ever learn their lesson until bloated software kills them (sometimes literally in a RTS device) or grossly invades their privacy to the point if never feeling safe in their own house without being a slave to corporations mentally screwing them for life and exploiting them for supet invasive targeted ads. There are times where I wonder if forcing people into using worse hardware like those UNIX/SGI workstations from the 1990s will make them learn once and for all that you can unleash your creative potential with weaker hardware that feels better than computers that were just manufactured last week.
Like Intel has said in the past, they're capable of doing R&D for the best possible minimal SoC that harkens back to the old days of computers where everyone wrote their own OS, but normies don't demand for that sort of thing, which tells me this is an epic cultural failure we'll never recover from for centuries unless we speak up and say "ENOUGH IS ENOUGH."
This was awesome
I think work from home work great, but you have to meet people and you get an understanding with the people.
But it's a different way of work.
So many bad USB to SATA bridges, all with the same USB ID. Is it a bad chip or a missing capacitor?
I didn’t know GKH lives in the Netherlands as do I .... cool
I love this.