True, thank you for pointing out the mistake! Quick summary because i'll pin this comment: EEVDF replaced CFS in 6.6 BUT was incomplete, as of 6.12 it is now complete.
@@FoobarDesign I think the posters complaining are talking about it's not the British/International pronunciation: "shed-u-al". I don't know, if they even know, that it's a "k" sound in American English. If that's not the complaint, I'm just going to pretend it is.
RT is all about _deterministic behavior_ not hard time scales. The point is knowing _exactly_ how a complex system will behave. For example, knowing how much cpu time a processes will have _and when._ It does not mean your "script" will always run in exactly the same time, just in a predictable time. In the CNC example, inputs (interrupts) will be processed in a known time range.
Idk, It's in the end often about hard time scales. If you know clock rates and how many cycles something deterministically takes you can also deterministically make something take exactly 109ns. And that's what you usually need RT for.
@@zapl80 Not exactly. RT is all about giving assurances things will happen within specific time frames. Linux (without RT voodoo patches) has never been able to offer any such guarantees. For example, a driver can turn off interrupts for as long as it likes, some kernel system calls cannot be interrupted (i.e. prevents the scheduler from switching tasks), etc. The _need_ for RT is a very rare thing, but when you do, it's an absolute. (Even with the RTLinux stuff, I've never used linux when RT is required.)
So making the video worse just because one guy said he didn't like the pronounciation? That would be childish. Actually the pronounciation irritates me too because all other words are pronounced well so it stands out and takes my focus away from what he's actually saying. I would love to not be de-focused from such a small thing but i can't unhear it. So, for me, this "critique" has nothing to do with trying to offend the voiceover and eveything to do with the desire to watch the video :)
36:15 musicians are still using firewire. audio interfaces are expensive, and music ppl get weird about the specific analog characteristics of stuff, so there's a strong incentive to never replace a known-good interface. i imagine a lot of scientific and industrial fields also make use of older firewire gear that's still decades away from obsolescence
I use FW, with my old XPS over the USB C cables (the thunderbolt 4). The legacy hardware is the 1st gen MOTU 896HD, and like everyone else who has something like this, it's for the analog front end. The microphone preamps are beautiful (subjectively, but they do have nice specs, too), and care nothing about the bus. It's robbery that Apple sells one of the only ways to connect FW devices to the TB bus, two dongles that cost as much as 100 USD. :(
@@Maple-Circuit plus you're getting money for that engagement. You're literally taking it to the bank. I think that's Joe Rogan's entire strategy for becoming a millionaire.
@@squirlmy For my channel standards? yes i'm taking it to the bank, all the engagement boosted this vid and money wise, 280USD which is the best for my channel. Not craizy but i have to pay for that davinci and obsidian licence (; I would get shit on any day of the week for that exposure XD
The PCIe Multipath (around 30:00) is basically a feature to connect one PCIe card using multiple slots. The reason is, that fast NICs and IB adaptors are faster than a single x16 PCIe port. Physically the solution is still a 1 slot card, but has a extra cable that goes to a "passive" board to cable adaptor in the next slot. There are then some benefits if you have a dual-socket system - these connections can go to each CPU, and freeing up precious bandwidth in QPI or infinity-fabric between the sockets.
I do appreciate the fire wire support because you have alot of people who have old tape recorders that will one day want to transfer their old memories from the tapes to the PC and without that stupid apple fire wire support they would be screwed. I being one of those people. So happy to have the support. Thank you Takashi Sakamoto.
Thank you for this insightful video. I never knew there was so much involved behind the scenes of the Kernel. I knew it was involved, but this was all next level. Exciting stuff.
When I clicked the video I thought it was going to be a brief update lol Didn't expect an hour long video. Glad I stuck with it though for the qr code feature at least lol that's pretty sick
Daily reminder que c'est "as a frenchman" ou "as a French person", "as a french" ça veut rien dire. M'enfin ça colle avec l'esprit du commentaire remarque.
36:05 - Firewire is still widely used in AV production, I don't think it'll die (to be fair I am absolutely furious with Microsoft killing and leaving out drivers for it, makes old and expensive hardware unusable unless you stick to Windows 7)
I always have a hard time describing what RT is really used for in real life. I usually use network traffic as an example, but I end up explaining the entire network stack hahaha. Thanks for those CNC examples.
It's also crucial for SDR and radar systems in real-time TDoA (Time Difference of Arrival) applications, as deterministic processing times are required, with precision in the nanosecond range.
TH-cam recommended me a gem, I had heard about "real time" thing but never bothered looking for it and then was like lets see what's new in 6.12 and didn't thought I'd be watching an hour long video about a kernel update... anyways the explanations are really good, I'm not into kernels neither do i do anything LL, but i still understood majority of it. Also chad Takashi carrying the old devices' lifeline, one of my friend still used a firewire cable with an external hdd till recently cuz the case and hdd itself was so old, and some people still use old camcorders, hd video camera and other devices from old tech. I've seen Japanese tech culture (in videos) and it's a blend of older and newer tech (which is kinda cool to me and one of the reasons I'd go to japan, to to use their tech), which is the reason Takashi is maintaining its support i think, or maybe just because he have some cool old tech and wanna maintain its support for as long as possible.
1:04:15 I would guess this change was noticed by mutation testing which intentionally breaks code logic (here probably adding `&& false` to the whole if test to disable the whole branch) and instead of kernel crash, they saw a huge performance boost. After that it was time to figure out why it was okay to ignore that whole branch.
dynamic scheduler is amazing, needing some high frequency kernel setting for one audio app, but do not want to generally overload the system with overhead .... awesome
I wish the PC community would make more noise about 10-bit color support. Color has to be the single most neglected specification within all of display technology. We're stuck at 8-bit on consumer hardware, which is pitiful in the grand scheme of things. There's 10-bit, 12-bit, 14-bit, and 16-bits per channel, and we've been at 8 bit ever since we left CRT's behind, which were a superior display tech.
He's technically correct in assuming the accent is on the second syllable of 'SCHEduler'. But in English we have exceptions for every rule, and this is a first syllable accent word, something you would not suspect unless you heard a native speaker. Another word like this is DEVelopment.
I had no idea there was this much of latency difference. Funny you mention time critical applications as it reminds me of a link system on "slave" trains that used Linux. When the tech was troubleshooting a connection issue he rebooted the system, I smiled a bit when I seen the boot up messages.
Thanks for the cool review. The real time feature and scheduler will be really helpful. Just a note, it is pronounced with the emphasis on the "SChed" part...not the dule, so it really sounds odd when you say scheDULer
Pretty certain maintaining any older technology, like Firewire, is for existing deployments of production machines. (or things like cars or uses in buildings or airplanes, etc.) Anything which has a long life as a device.
1:28 This is a freaking feature if it's working perfectly, I'm not kernel guy or even good C/C++ programmer, but I use a lot of software and I have ( _software review website_ ) not in CNC only this feature is vital or anything related to controlling a machine, but all other software that not rely heavily on human interaction, synchronized, automated tasks
There are many quality firewire devices that people have no need to replace, can't replace, don't want to replace, even if they could. I wished my Alesis io device didn't bite the dust, cause I would still use it today.
Great video. I am always amazed when people try to correct others speech, especially since they obviously understood what you said. Not to even point out how you are most likely bilingual and they seem like they can barely speak 1 language.
I have a kit with what I believe is the first gen raspberry pi. I'm finally ready to start messing around with it. Wondering if this new kernel would work on it.
just found your channel, I must say the info and review of the changes is pretty good, just one note about the AI voice used, it seems a little broken with some words, besides that great video
KVM always brings to my mind KVM switch ("keyboard, video, and mouse"). Maybe we should ease up on using TLA for so many things without saying what they mean...
Can you imagine the combination of the new EDAC feature with real time Linux ? If we can guarantee how long a single fault takes and the rest of the system continues in the right time, would be amazing. Maybe we now can.
I think the Rust progress should be followed by monitoring the amount of crashes per 100k lines. If the Rust code actually gets less (critical? memory corruption?) bugs than C code, then the transition to more Rust makes sense even if it's painful. If I've understood correctly, there's still too little Rust code to have statistical trust in the results right now so kernel community must keep adding Rust until the results are statistically meaningful. If Rust code is statistically less broken then it means that Rust usage should increase in the kernel. If Rust code is equally broken to C then I'm not sure if adding any new Rust code is worth the effort. This far the Apple AGX GPU driver has been the most complex Rust part and that has shown a lot of promise already.
It's almost a certainty that Rust code will display fewer memory corruption errors and vulnerabilities than C code. IMO, Rust is simply a much friendlier and less error prone language than C. Programs coded in Rust are normally rock solid, whereas C programs ... not so much. TBH, I'm still a bit surprised that a language with a relatively high level "feeling" to it is really suitable for doing kernel work, but perhaps that's just my lack of imagination. If it works out, that would be fantastic.
@@dschledermann C works with enough eyeballs doing code review with decades of C programming experience. The only question is if even the LKML has enough eyeballs to make C safe enough compared to Rust?
I thought there was a clock like a timer chip in the processor or on the motherboard that did the timing? I don't see how you can do timing in code alone. Linux never used a clock before? That should have been priority number one a long time ago.
@@tom-hy1kn if you're talking about the realtime part, you're throwing unrelated stuff together. Obviously Linux uses the same hardware as everyone else. And yes that includes a clock. Realtime is about predictability and determinism. You don't need a realtime os for games, e.g. windows isn't really realtime, you'd have to look at windows ce or iot versions.
@@SimonVaIe I don't know anything about it. Why would controlling timing of a CNC machine be any different than controlling the timing in a video game?
41:00 If this does what it think it does, it's a good thing (haven't taken a look yet). But, we already had something like this for a few vendors like Lenovo (AMT, DYTC), and while it's using ACPI you'd still need a driver per vendor. Also, /sys/firmware/acpi/platform_profile is a horrible, windows like, interface.
I only use Windows for Cubase, that's the only program I have installed on my workstation for music making. If I could run it on Linux with better performance and support for my interface, it would be Christmas.
Ubuntu is pushing the gas on their kernel, if you follow the non LTS release of Ubuntu, you get the latest kernel at the time of release... even if that kernel is still in the works! (ex: kernel 6.x.0-RC3)
Absolutely lovely video, very well put and thought. Just a little observation, please look up the pronunciations of Management Scheduler Register/Registers Your english is already very good, no harm in honing it up a bit.
I was going to install 6.12 today, only because I need the Intel BE200 320MHz WiFi 7 network card to work correctly, it will rarely and only briefly connect to it, it usually connects to the 160MHz AX band with poor results. On this same machine on a different SSD is Windows 11 Pro, and the card screams, to the router the connection is 5.7 Gbps and a speedtest maxes out my ISP Gigabit bandwidth at 1.2 Gbps. Intel says it was in kernel 6.5, no it wasn't , I'm running 6.8 right now and it doesn't work right. I have another WiFi 7 card that is a MediaTek, and it's not kernel supported at all right now. I'm waiting for the kernel to catch up to my hardware, which is unusual for me, I usually have older equipment. Maybe 6.12 will solve it for me, maybe not, we'll see.
Yea, you know you did a good job when no one complains about the factuality of your content. The amount of comments would have made it worthwhile to say something wrong on purpose XD
0/10 The opening promised a golden Kermit. There were no frogs in this video. Only nerds talking about commits. Edit: Really good explanation of the realtime stuff. I am amazed bcachefs is still being included after the most recent conflict with the developer. Shows that Linus has a lot more patience than I ever would.
Ya would think so right? turn out app dev really want that feature. Isolation has always been part of Android but the only way to make 100% sure you are isolated from everything else would be to VM your app. Now for a random game, that makes no sense. For a banking app... We also tend to overestimate the impact of VMs on performance. When implemented correctly, they can be minimal and this "Perfect" integration is probably what Google is aiming for. PS. If you have experience with VirtualBox.... i can understand why the idea of a VM on a phone might be crazy, just keep in mind that X86_64 emulation is really heavy for some reason...
@@Maple-Circuit You seem to have it backwards. A VM does not protect the software inside it from the host system, rather it protects the host system from the software inside the VM.
@@phillipsusi1791 was true in the past, now tech like AMD's SEV-SNP or even Intel's TDX are there to solve the protection of the VM from the host. Now to be 100% honest, i have no clue if ARM has an equivalent, but i would be very suprise if that wasn't the case (;
@@Maple-Circuit Yea, I don't really believe that, but even if it were so, it would be far less efficient to use that to isolate individual applications on your cell phone rather than using a one kernel to isolate applications from one another ( which is already its job ), than to introduce multiple kernels that are isolated from one other by a hypervisor. That is something you want when you are a cloud provider and are selling virtual machine instances to different clients that can run their own kernel and you want to make sure that none of them can interfere with each other on your cloud provider hardware. That is of no concern when you are the provider of the firmware image on a cell phone that only needs to run multiple applications in isolation from one another, not multiple kernels, which applications don't get to run their own.
Linus being the way he is, is eventually going to cause the kernel to get forked out from under him. Regardless of what you think of Linus, having one unelected guy being able to dictate rules to everyone else in regards to the kernel goes against everything FOSS is *SUPPOSED* to be about, and eventually this will catch up to him and have disastrous consequences for reasons that should be completely obvious.
1:02:50 -> poor little babys being removed doing his job only because his russian nationality... it is wild.. but understandable, I realy want to see this guys agin in the project in the the future, poor babys..
49:20 I thought EEVDF already replaced CFS in Linux 6.6?
It was noted to be "incomplete" at the time iirc.
True, thank you for pointing out the mistake!
Quick summary because i'll pin this comment:
EEVDF replaced CFS in 6.6 BUT was incomplete, as of 6.12 it is now complete.
You somehow managed to explain a kernel change log in a way that non-kernel devs can understand while still being entertaining. Bravo!
Thanks!
Very good video indeed 👍
Skijeweler is my new favorite word
Funniest thing is he nails it at 3:23 so he CAN say it correctly.
That one was great, but lo-CAH-lio @ 11:24 is my favorite
It's great for engagement :P
@@FoobarDesign I think the posters complaining are talking about it's not the British/International pronunciation: "shed-u-al". I don't know, if they even know, that it's a "k" sound in American English. If that's not the complaint, I'm just going to pretend it is.
You can thank his thick Quebec accent.
RT is all about _deterministic behavior_ not hard time scales. The point is knowing _exactly_ how a complex system will behave. For example, knowing how much cpu time a processes will have _and when._ It does not mean your "script" will always run in exactly the same time, just in a predictable time. In the CNC example, inputs (interrupts) will be processed in a known time range.
Idk, It's in the end often about hard time scales. If you know clock rates and how many cycles something deterministically takes you can also deterministically make something take exactly 109ns. And that's what you usually need RT for.
@@zapl80 Not exactly. RT is all about giving assurances things will happen within specific time frames. Linux (without RT voodoo patches) has never been able to offer any such guarantees. For example, a driver can turn off interrupts for as long as it likes, some kernel system calls cannot be interrupted (i.e. prevents the scheduler from switching tasks), etc. The _need_ for RT is a very rare thing, but when you do, it's an absolute. (Even with the RTLinux stuff, I've never used linux when RT is required.)
This is so true. Hard realtime OSes limit how long latencies are, not how short they can be.
Please don't stop saying scheduler that way just to piss off the guy who said "please stop saying scheduler that way".
Lol, spite based spelling XD
It is "spite based pronunciation".
So making the video worse just because one guy said he didn't like the pronounciation? That would be childish. Actually the pronounciation irritates me too because all other words are pronounced well so it stands out and takes my focus away from what he's actually saying. I would love to not be de-focused from such a small thing but i can't unhear it. So, for me, this "critique" has nothing to do with trying to offend the voiceover and eveything to do with the desire to watch the video :)
Don't worry I'm working on it (;
@Cranked1 Shut up nerd
36:15 musicians are still using firewire. audio interfaces are expensive, and music ppl get weird about the specific analog characteristics of stuff, so there's a strong incentive to never replace a known-good interface. i imagine a lot of scientific and industrial fields also make use of older firewire gear that's still decades away from obsolescence
I don't know what OS is running on fighter jets, but firewire is also F-22s' system bus.
I use FW, with my old XPS over the USB C cables (the thunderbolt 4). The legacy hardware is the 1st gen MOTU 896HD, and like everyone else who has something like this, it's for the analog front end. The microphone preamps are beautiful (subjectively, but they do have nice specs, too), and care nothing about the bus.
It's robbery that Apple sells one of the only ways to connect FW devices to the TB bus, two dongles that cost as much as 100 USD. :(
My Sony DV Cam uses FireWire for video transfer
36:30 Some Airplane black boxes use FireWire (IEEE 1394), so... you know, it might be important sometimes
Wow! I didn't know that!
CLAP!
I use sche'duler btw
Shed jeweler
Bro is getting roasted for a good video , the internet is ruthless 💀
It doesn't bother me, i'm infused by internet cancer.
@@Maple-Circuit plus you're getting money for that engagement. You're literally taking it to the bank. I think that's Joe Rogan's entire strategy for becoming a millionaire.
@@squirlmy For my channel standards? yes i'm taking it to the bank, all the engagement boosted this vid and money wise, 280USD which is the best for my channel. Not craizy but i have to pay for that davinci and obsidian licence (;
I would get shit on any day of the week for that exposure XD
This is the best kernel series ever, I've seen no other place talk about the linux kernel with this much detail and hype.
Thanks!!
The PCIe Multipath (around 30:00) is basically a feature to connect one PCIe card using multiple slots. The reason is, that fast NICs and IB adaptors are faster than a single x16 PCIe port. Physically the solution is still a 1 slot card, but has a extra cable that goes to a "passive" board to cable adaptor in the next slot. There are then some benefits if you have a dual-socket system - these connections can go to each CPU, and freeing up precious bandwidth in QPI or infinity-fabric between the sockets.
We are reaching levels of insanity that are unheard of
Oh my god, amazing video! Not only reading changes, but giving practical examples. Dude, you are awesome, instant subscribe!
Glad you liked it!
I do appreciate the fire wire support because you have alot of people who have old tape recorders that will one day want to transfer their old memories from the tapes to the PC and without that stupid apple fire wire support they would be screwed. I being one of those people. So happy to have the support. Thank you Takashi Sakamoto.
Apple Fah-wah?
CLAP!
TH-cam algorithm, just recommended this video, and channel that explains in depth an Linux Kernel release 👍
Savage.
Thank you so much for explaining all the detailed changes coming to the next Linux release!
Wow i thought, that i would just watch for 10 minutes or so but you posted an entire hour of material and its great!
Thanks!
Thank you for this insightful video. I never knew there was so much involved behind the scenes of the Kernel. I knew it was involved, but this was all next level. Exciting stuff.
Glad you enjoyed it!
I watch your video at normal speed and I enjoy it
This made me subscribe 3 seconds into the video.
Nothing like a good testimonial to know you’re in the right place ❤
he has incredibly calm voice. i agree
I watch it at 0.75x 🥴
When I clicked the video I thought it was going to be a brief update lol Didn't expect an hour long video. Glad I stuck with it though for the qr code feature at least lol that's pretty sick
As a french, i really couldn't care less how you say scheduler
lol
As someone who's not French 🥖 I don't care how many times you surrender or make horrible food.
@@xephael3485 i'm french canadian, i never surrender and i make poutine XD
Daily reminder que c'est "as a frenchman" ou "as a French person", "as a french" ça veut rien dire. M'enfin ça colle avec l'esprit du commentaire remarque.
@@GeneralKenobi69420 je me disais que ça sonnait bizarre aussi. Merci
Thanks again, your kernel updates videos are always very informative
Excited for ext3, also eBPF if you remember ;)
Of course!
Sometimes I genuinely forget my phone runs on the Linux kernel
I mean... unless you actually have a Linux phone, your phone runs on a Android Common Kernel. Which is based on Linux-LTS, but is not Linux-LTS.
holy shit , GREAT INFORMATION . subscribed
36:05 - Firewire is still widely used in AV production, I don't think it'll die (to be fair I am absolutely furious with Microsoft killing and leaving out drivers for it, makes old and expensive hardware unusable unless you stick to Windows 7)
Good vid. Very cool that we can just listen with no need to watch the screen.
I always have a hard time describing what RT is really used for in real life. I usually use network traffic as an example, but I end up explaining the entire network stack hahaha. Thanks for those CNC examples.
compiling linux-6.12-rc7 right now
make LLVM=1 -j4 pacman-pkg
It's also nice for audio
@@spicybaguette7706 that what i heard too, but i not familiar with DSP.
It stands for Real Turd.
It's also crucial for SDR and radar systems in real-time TDoA (Time Difference of Arrival) applications, as deterministic processing times are required, with precision in the nanosecond range.
TH-cam recommended me a gem, I had heard about "real time" thing but never bothered looking for it and then was like lets see what's new in 6.12 and didn't thought I'd be watching an hour long video about a kernel update... anyways the explanations are really good, I'm not into kernels neither do i do anything LL, but i still understood majority of it. Also chad Takashi carrying the old devices' lifeline, one of my friend still used a firewire cable with an external hdd till recently cuz the case and hdd itself was so old, and some people still use old camcorders, hd video camera and other devices from old tech. I've seen Japanese tech culture (in videos) and it's a blend of older and newer tech (which is kinda cool to me and one of the reasons I'd go to japan, to to use their tech), which is the reason Takashi is maintaining its support i think, or maybe just because he have some cool old tech and wanna maintain its support for as long as possible.
1:04:15 I would guess this change was noticed by mutation testing which intentionally breaks code logic (here probably adding `&& false` to the whole if test to disable the whole branch) and instead of kernel crash, they saw a huge performance boost. After that it was time to figure out why it was okay to ignore that whole branch.
Your channel is really worth supporting. Nice video.
Thank you!
Dev: Accidentally introduced 2ms delay.
CEO: Great work! Here's your golden trophy.
OMFG... appreciate this presentation. Scheduler is awesome 😎 much respect from Thailand 👏❤💯 SUBSCRIBED!!
A little tip: LOCALIO just means two words: local I/O (Input/Output)
Kent circle of conflict LMAO 😂
A great example of a narcissist making a fool of himself, it's as funny as it is sad.
dynamic scheduler is amazing, needing some high frequency kernel setting for one audio app, but do not want to generally overload the system with overhead .... awesome
love your videos like this
I wish the PC community would make more noise about 10-bit color support. Color has to be the single most neglected specification within all of display technology. We're stuck at 8-bit on consumer hardware, which is pitiful in the grand scheme of things. There's 10-bit, 12-bit, 14-bit, and 16-bits per channel, and we've been at 8 bit ever since we left CRT's behind, which were a superior display tech.
He's technically correct in assuming the accent is on the second syllable of 'SCHEduler'. But in English we have exceptions for every rule, and this is a first syllable accent word, something you would not suspect unless you heard a native speaker.
Another word like this is DEVelopment.
3:30 replay button 🥵
Awesome overview! Thank you!
So epic most wont notice.....
A shot for every time you say sche JULER :D haahahah, love the video, no disrespect :D
Very interesting information. Thank you.
thank you very much for these videos that are actually very educational. I love how you explain everything "in english please"
I had no idea there was this much of latency difference. Funny you mention time critical applications as it reminds me of a link system on "slave" trains that used Linux. When the tech was troubleshooting a connection issue he rebooted the system, I smiled a bit when I seen the boot up messages.
seeing linux in unexpected places is always waaaayyyy too smile inducing (;
Glad I found this channel.
Thanks for the cool review. The real time feature and scheduler will be really helpful.
Just a note, it is pronounced with the emphasis on the "SChed" part...not the dule, so it really sounds odd when you say scheDULer
Pretty certain maintaining any older technology, like Firewire, is for existing deployments of production machines. (or things like cars or uses in buildings or airplanes, etc.) Anything which has a long life as a device.
1:28 This is a freaking feature if it's working perfectly, I'm not kernel guy or even good C/C++ programmer, but I use a lot of software and I have ( _software review website_ ) not in CNC only this feature is vital or anything related to controlling a machine, but all other software that not rely heavily on human interaction, synchronized, automated tasks
There are many quality firewire devices that people have no need to replace, can't replace, don't want to replace, even if they could. I wished my Alesis io device didn't bite the dust, cause I would still use it today.
BTRFS actually is garbage, but Kent Overspeed does not increase the trustworthiness of his file system by his behavior.
The name.... is sche'duler agen
jokes apart, really good video!!
actually, debugging a windows operating system kernel over firewire is still something that there is at least one single person in the world doing :)
I'm subscribing, hot, shmexy accent, I'm melting...
He doesnt pronounciate the H...Like for "happy" , he says "appy".
@ami443, I don't care 😆
@@denisvolin7489 yes but I was just noticing, and it is very strange to me.
XD
So is there anything special we need to do to take advantage of the RT features? If I have 6.12 is it just on or is there a flag etc?
You just need to compile your kernel with the option
12:07 I assume it should be pronounced as local I/O?
yes (;
It’s not 67 minute historic. Self editing is a virtue
Scheduler is commonly *SKED-yule-er* but a single schedule is often pronounced softer as in *SHED-yule* yes, this language is a mishmash.
Que buen video muchas gracias
Great video. I am always amazed when people try to correct others speech, especially since they obviously understood what you said. Not to even point out how you are most likely bilingual and they seem like they can barely speak 1 language.
Lol true (;
I have a kit with what I believe is the first gen raspberry pi. I'm finally ready to start messing around with it. Wondering if this new kernel would work on it.
Yep it would
@@Maple-Circuit oh that is beautiful, thank you
LinuxCNC is a absolutely still a ting :)
I saw that, I n s a n e
This update is a Schejeweller in the Crown on Tux's head
just found your channel, I must say the info and review of the changes is pretty good, just one note about the AI voice used, it seems a little broken with some words, besides that great video
I'm not an ai XD
@@Maple-Circuit J/K mate hahaha
Skedjewler?
I dropped dead out of laughter XD
KVM always brings to my mind KVM switch ("keyboard, video, and mouse").
Maybe we should ease up on using TLA for so many things without saying what they mean...
You are right! I try to say the full name of everything as much as possible... will think about KVM next time (;
Can you imagine the combination of the new EDAC feature with real time Linux ? If we can guarantee how long a single fault takes and the rest of the system continues in the right time, would be amazing. Maybe we now can.
49:19 🤣🤣 Oh boy, the way you say “scheduler” 😁
I thought it’s a US vs British pronunciation thing but this is something completely wild 😁
I'm pretty sure the brits say shedule or something idk i dont talk with beans in my tea
I don't understand anything but I hope one day i'll understand
Lol we all get there (;
@@Maple-Circuit lmao I just switched to linux on my main pc like a month ago
I think the Rust progress should be followed by monitoring the amount of crashes per 100k lines. If the Rust code actually gets less (critical? memory corruption?) bugs than C code, then the transition to more Rust makes sense even if it's painful. If I've understood correctly, there's still too little Rust code to have statistical trust in the results right now so kernel community must keep adding Rust until the results are statistically meaningful.
If Rust code is statistically less broken then it means that Rust usage should increase in the kernel. If Rust code is equally broken to C then I'm not sure if adding any new Rust code is worth the effort. This far the Apple AGX GPU driver has been the most complex Rust part and that has shown a lot of promise already.
It's almost a certainty that Rust code will display fewer memory corruption errors and vulnerabilities than C code. IMO, Rust is simply a much friendlier and less error prone language than C. Programs coded in Rust are normally rock solid, whereas C programs ... not so much. TBH, I'm still a bit surprised that a language with a relatively high level "feeling" to it is really suitable for doing kernel work, but perhaps that's just my lack of imagination. If it works out, that would be fantastic.
@@dschledermann C works with enough eyeballs doing code review with decades of C programming experience.
The only question is if even the LKML has enough eyeballs to make C safe enough compared to Rust?
PURE MAPLE SYRUP: Fedora
I just love generated voices.
You think my voice is generated? Lol
@ might be biologically generated, but still, you’re close to a voice that’s been trained on French / Belgian young male with heavy accent 🤪.
XD
Take a shot for every skeyuler.
23:32 this also sounds like it has the potential to be used for storage.
I thought there was a clock like a timer chip in the processor or on the motherboard that did the timing? I don't see how you can do timing in code alone. Linux never used a clock before? That should have been priority number one a long time ago.
Modern CPU can modulate their frequency (clock) to give them more or less processing power depending on needs.
@@Maple-Circuit Linux took their time in dealing with it. Video games on Linux were changing speed too?
@@tom-hy1kn if you're talking about the realtime part, you're throwing unrelated stuff together.
Obviously Linux uses the same hardware as everyone else. And yes that includes a clock.
Realtime is about predictability and determinism.
You don't need a realtime os for games, e.g. windows isn't really realtime, you'd have to look at windows ce or iot versions.
@@SimonVaIe I don't know anything about it. Why would controlling timing of a CNC machine be any different than controlling the timing in a video game?
what's a good mini pc that will run linux and have WIFI 7?
41:00 If this does what it think it does, it's a good thing (haven't taken a look yet). But, we already had something like this for a few vendors like Lenovo (AMT, DYTC), and while it's using ACPI you'd still need a driver per vendor. Also, /sys/firmware/acpi/platform_profile is a horrible, windows like, interface.
Nvm, still kinda meh. This change is for things in the FFH, platform_profile is not in FFH. So still, a vendor's clown fiesta.
I only use Windows for Cubase, that's the only program I have installed on my workstation for music making. If I could run it on Linux with better performance and support for my interface, it would be Christmas.
Reaper?
36:29 Takashi Sakamoto is a real trooper for keeping Firewire alive on GNU+Linux. Make sure you don't "CLAP" him too hard...
Lol
The virtual machine instruction bug is also in 5000... VMware workstation on Windows 100% shutdown host with motherboard showing an unknown error code
Anyone familiar with kerigrid?
Can something like that work for this kernel with an asymmetric multiprocess ‽‽‽
Not familiar
01:03:57
you re welcome
I would assume you were already able to set scheduling priority on a process om linux. Doesn't windows allow you to do that?
Yep It does, if you are talking about SCHED_DEADLINE the problem happened when all processes (deadline and fifo) would have the same priority level.
Now to see if AMD GPU performance issues were fixed. Been experiencing issues since 6.9.
cant wait for Ubuntu/Mint to get to that release as stable .. just updated to 6.8.0 on all my systems so probably going to be a while.
Ubuntu is pushing the gas on their kernel, if you follow the non LTS release of Ubuntu, you get the latest kernel at the time of release... even if that kernel is still in the works! (ex: kernel 6.x.0-RC3)
Absolutely lovely video, very well put and thought. Just a little observation, please look up the pronunciations of
Management
Scheduler
Register/Registers
Your english is already very good, no harm in honing it up a bit.
lol, just saw I had made the same observations xD
Will work on it! Thanks!
one more: uring in io_uring is you-ring. I think the u is for usespace and ring is for ring buffer.
I was going to install 6.12 today, only because I need the Intel BE200 320MHz WiFi 7 network card to work correctly, it will rarely and only briefly connect to it, it usually connects to the 160MHz AX band with poor results. On this same machine on a different SSD is Windows 11 Pro, and the card screams, to the router the connection is 5.7 Gbps and a speedtest maxes out my ISP Gigabit bandwidth at 1.2 Gbps. Intel says it was in kernel 6.5, no it wasn't , I'm running 6.8 right now and it doesn't work right. I have another WiFi 7 card that is a MediaTek, and it's not kernel supported at all right now. I'm waiting for the kernel to catch up to my hardware, which is unusual for me, I usually have older equipment. Maybe 6.12 will solve it for me, maybe not, we'll see.
Stop saying scheduler that way, wtf.
ikr!!!
Scheduuuuuling
@@Lesfehlerable sounds incorrect.
Drawing incorrect implications is a sign of stupidity however
@@Lesfehlerable random word generator
No.
What will you do?
😂😂😂
Kernel 6.12 shows as invalid in Canonical Mainline Kernels
Jesus you have the craziest comment section I've ever seen.
Yea, you know you did a good job when no one complains about the factuality of your content.
The amount of comments would have made it worthwhile to say something wrong on purpose XD
no firewire, no lighting, no water cooler, no grounding. i guess my rig is element-o-phobic
Not even thunderbolts and lightning?
Very very frightening....
a variant of SCHED_EXT was used by Po in Kung fu panda against tiger villain
Hell yeah, the kernel explorer for ya.
SKED-u-ler, not sked-DULER. How have you never seen or used this word pronounced correctly? Thank you for this video.
0/10 The opening promised a golden Kermit. There were no frogs in this video. Only nerds talking about commits.
Edit: Really good explanation of the realtime stuff. I am amazed bcachefs is still being included after the most recent conflict with the developer. Shows that Linus has a lot more patience than I ever would.
>Shows that Linus has a lot more patience than I ever would.
Insane right?
Huh? What android devices run in a virtual machine? Why would you want the overhead of a hypervisor on your phone?
Ya would think so right? turn out app dev really want that feature.
Isolation has always been part of Android but the only way to make 100% sure you are isolated from everything else would be to VM your app.
Now for a random game, that makes no sense.
For a banking app...
We also tend to overestimate the impact of VMs on performance.
When implemented correctly, they can be minimal and this "Perfect" integration is probably what Google is aiming for.
PS. If you have experience with VirtualBox.... i can understand why the idea of a VM on a phone might be crazy, just keep in mind that X86_64 emulation is really heavy for some reason...
@@Maple-Circuit You seem to have it backwards. A VM does not protect the software inside it from the host system, rather it protects the host system from the software inside the VM.
@@phillipsusi1791 was true in the past, now tech like AMD's SEV-SNP or even Intel's TDX are there to solve the protection of the VM from the host.
Now to be 100% honest, i have no clue if ARM has an equivalent, but i would be very suprise if that wasn't the case (;
@@Maple-Circuit Yea, I don't really believe that, but even if it were so, it would be far less efficient to use that to isolate individual applications on your cell phone rather than using a one kernel to isolate applications from one another ( which is already its job ), than to introduce multiple kernels that are isolated from one other by a hypervisor. That is something you want when you are a cloud provider and are selling virtual machine instances to different clients that can run their own kernel and you want to make sure that none of them can interfere with each other on your cloud provider hardware. That is of no concern when you are the provider of the firmware image on a cell phone that only needs to run multiple applications in isolation from one another, not multiple kernels, which applications don't get to run their own.
I feel like the "performance improvements" are to be incorporated with state sponsored vulnerabilities
Linus being the way he is, is eventually going to cause the kernel to get forked out from under him. Regardless of what you think of Linus, having one unelected guy being able to dictate rules to everyone else in regards to the kernel goes against everything FOSS is *SUPPOSED* to be about, and eventually this will catch up to him and have disastrous consequences for reasons that should be completely obvious.
Nvidia display drivers are not even installable on some linux kernels. Or if you do install them they don't work correctly. Let's see what happens.
1:02:50 -> poor little babys being removed doing his job only because his russian nationality... it is wild.. but understandable, I realy want to see this guys agin in the project in the the future, poor babys..
amazing video!