Want to read Slashdot from your mobile device? Point it at m.slashdot.org and keep reading!

 



Forgot your password?
typodupeerror
Operating Systems Linux

Linux 4.17 Kernel Offers Better Intel Power-Savings While Dropping Old CPUs (phoronix.com) 136

An anonymous reader writes: Linus Torvalds has released Linux 4.17-rc1. This kernel comes with a significant amount of new capabilities as outlined by the Linux 4.17 feature overview. Among the new features are AMDGPU WattMan support, Intel HDCP support, Vega 12 GPU enablement, NVIDIA Xavier SoC support, removal of obsolete CPU architectures, and even better support for the original Macintosh PowerBook 100 series. Phoronix testing has also revealed measurable power savings improvements and better power efficiency on Intel hardware. The kernel is expected to be stabilized by June.

Linux 4.17 Kernel Offers Better Intel Power-Savings While Dropping Old CPUs

Comments Filter:
  • But, but (Score:1, Offtopic)

    Only Apple drops old hardware.
    • by Anonymous Coward

      These aren't exactly mainline architectures they are dropping and they are being dropped due to the current lack of active use.

      • Just like Apple what drops.

      • The architectures that were dropped are Blackfin, CRIS, FRV, M32R, Metag, MN10300, Score, Tile, and POWER4.

        Blackfin is a DSP-oriented processor made by Analog Devices. It is not normally used for general purpose computing, so its applications rarely call for a full OS running on it. It's used in embedded applications, not as a user-visible CPU.

        POWER4 was used in IBM mainframes circa 2004, and was never found in any Apple computers. The PowerPC 970 that was used in some Macs, which was derived from POWER4,

    • Re:But, but (Score:4, Informative)

      by Dorianny ( 1847922 ) on Monday April 16, 2018 @01:31AM (#56444271) Journal
      They were dropped because there are no developers to maintain the code, If you want to pick up the slack, I'm sure the community will be more then willing to rummage through their junk-piles to send you testing hardware
      • by ReneR ( 1057034 )
        junk pile of hardware? that would be so nice, I currently mostly need an AAUI adapter (Apple AUI, Ethernet) https://www.youtube.com/watch?... [youtube.com] If you find other goodies, let me know. Other most wanted is some Sgi Octane HD tray, or this Mac 8100/80 HD and CD-ROM tray: https://rene.rebe.de/wishlist/ [rene.rebe.de]
      • For most purposes, it's not necessary to test it on hardware. We know the underlying functions work and that side-effects are minimal. The requirements are that the API presented matches the API the kernel uses, the semantics are as expected and that the code compiles. That requires a continuous integration tool and a way to describe the semantics.

        You can get the software for free. The effort is a bit more, depending on how much it is worth.

        It won't prove things work, but it will prove things will probably

    • So they're dropping support for "obsolete" CPUs but somehow the first-generation Motorola 68000 in the PowerBook 100 doesn't qualify as obsolete? I mean, I was a Mac user in the '90s, and I still have some fondness for the some of the old hardware, but this is ridiculous. What kind of mental gymnastics do you have to go through to keep 68000 but drop the much newer Blackfin DSP, and the M32R that still seems to be manufactured for embedded applications?

      • Re:The irony... (Score:4, Informative)

        by religionofpeas ( 4511805 ) on Monday April 16, 2018 @02:47AM (#56444417)

        What kind of mental gymnastics do you have to go through to keep 68000 but drop the much newer Blackfin DSP

        No mental gymnastics required. You just need a developer willing to support the 68k, and not have a developer for the Blackfin.

        • by _merlin ( 160982 )

          If that's the case, just say you're dropping unmaintained architectures due to lack of interest or active maintainers. Talking about dropping "obsolete" architectures is somewhat misleading when the architectures being dropped are still in being manufactured or developed, but undeniably obsolete architectures are still being actively supported.

          • I suppose if they'd said "unmaintained" instead of "obsolete" it would have been a tautology. Perhaps "unpopular"?

            • I read the release notes, most of it was already broken on those dropped architectures.

              I'd say "nonworking and unmaintained"

        • Typically they'll leave old hardware in of anyone says they are USING it. It won't necessarily see new development. I've seen several lkml posts asking "is anyone using _____? If not, we'll remove it."

          • I've seen several lkml posts asking "is anyone using _____? If not, we'll remove it."

            lkml is mostly for developers, though.

      • Re: The irony... (Score:4, Informative)

        by Z00L00K ( 682162 ) on Monday April 16, 2018 @02:56AM (#56444437) Homepage

        The 68k is kind of baseline for that tree, and relatively simple even for people without specialized knowledge, so it may stick around for a while. It's harder to maintain more complex architectures where you need instruction order and cache management to think about as well.

      • by raymorris ( 2726007 ) on Monday April 16, 2018 @03:09AM (#56444469) Journal

        Typically old hardware isn't dropped if we know someone is using it. Of course there are exceptions when a stack gets a major rewrite. If there is discussion about dropping support for something you use, just let us know on lkml.

      • Blackfin was discussed. It's not being removed at this time.
        A manufacturer using Blackfin, or someone who wants to keep support in newer kernels, will need to step up, though, or it will be removed soon.

      • Dragonball is still in use and is 68K based. It is also more likely to run a Linux OS than the Blackfin or M32R which usually run an RTOS.

        Also, a lot of enthusiast systems run 68K CPU's, but I know of no enthusiast systems at all based around Blackfin or M32R. And since enthusiasts are likely to be contributing to Linux, support works itself out. Therefore support for some obsolete CPU's is dropped, while support for other obsolete CPU's remain.

        No mental gymnastics required.

      • Blackfin is still being made. But hardly anybody ever ran Linux on it. A common usage scenario for Blackfin is as a specialized DSP processor for a system that also contains another CPU (these days most likely ARM) that runs a full OS.
    • Only Apple drops old hardware.

      I was coming here to see how many people who bashed Apple on here last week for WARNING about STARTING to Deprecate 32-bit Support in macOS were going to completely DEFEND this move by "Linux" (whoever that is!).

  • by SeaFox ( 739806 ) on Monday April 16, 2018 @01:32AM (#56444275)

    I'm curious what's considered an "obsolete CPU architecture" if a Powerbook 100 is still supported.

    • things like Power4, Blackfin, M32R etc.
    • by ShanghaiBill ( 739463 ) on Monday April 16, 2018 @01:39AM (#56444297)

      I'm curious what's considered an "obsolete CPU architecture" if a Powerbook 100 is still supported.

      Obsolete = Nobody is stepping up to maintain the code.

      Obviously someone still cares about the Powerbook 100 enough to do the maintenance work.

      Proprietary software drops support when they no longer care. Open source drops support when you no longer care.

      • by houghi ( 78078 )

        That is not what obsolete means. That is what un-maintained means. If I am the only person using and maintaining the source for hardware I developed my self, then it would perhaps still be maintained, even if it is obsolete.
        If nobody maintains the source even when there are millions of users, then it would be dropped. Hardly obsolete.
        Yes, very often obsolete hardware will become un-maintained, as there are not any developers that have the hardware or the will to maintain it.
        Obsolete =/= un-maintained.

        • If I am the only person using and maintaining the source for hardware I developed my self, then it would perhaps still be maintained, even if it is obsolete.

          You're describing a vendor actively maintaining a product in active use. That by its nature means it's not obsolete.

    • Anything before Skylake. Oh wait nevermind I thought we were talking about Windows 10.

    • Things like Unicore, Hexagon, S+core, OpenRISC, M32R, Cris i.e. stuff most people didn't even heard about.
      The long version at (as always) excellent LWN:
      https://lwn.net/Articles/74807... [lwn.net] and
      https://lwn.net/Articles/74929... [lwn.net]

      • by mysidia ( 191772 )

        The article summary didn't do it justice and implied they might be dropping out some form of support for older chips regardless of architecture such as older X86 generations ..... That's not the case.

        plan that emerged from this discussion is to remove score, unicore, metag, frv, and m32r in the 4.17 development cycle, while hexagon and openrisc will be retained. There will be a brief reprieve for blackfin and tile, which will be removed "later this year" unless a maintainer comes forward. And mn10300 wil

        • by Anonymous Coward

          Were likely feature-complete anyways.

          None of them supported USB or later versions of USB, PCIe, etc. Meaning none of them have been gaining anything other than bloat from newer iterations of the kernel. Furthermore some of them may have been broken thanks to the minimum kernel size exceeding the maximum kernel image the bootloader could load (sparc ipx was limited to 1 meg maximum kernel size, which was impossible with later 2.4/2.6 kernels and the minimum features for bringup, notably ipv4 support.)

          While I

    • by nateman1352 ( 971364 ) on Monday April 16, 2018 @03:38AM (#56444523)

      I'm curious what's considered an "obsolete CPU architecture" if a Powerbook 100 is still supported.

      The full list of dropped CPUs is Blackfin, CRIS, FRV, M32R, Metag, MN10300, Score, and Tile. Also under consideration are Unicore32 and Hexagon, but they are not officially gone yet. Apparently this change removes about half a million lines of code, a substantial reduction in complexity. I had never heard of any of these before and I suspect most other people haven't either, so I don't think they will be missed.

      • I had never heard of any of these before and I suspect most other people haven't either, so I don't think they will be missed.

        Blackfin I've heard of, but don't know what it is. Tile is the instruction set used by Tilera's parallel CPU, which had 64 cores on a chip about 10 years ago. Looks like they switched the thing over to use ARM cores after the first couple of generations.

      • Re: (Score:2, Informative)

        by Anonymous Coward

        If you're an audio engineer, the Blackfin is used by the legendary Bricasti Model 7 [bricasti.com] (actually, the M7 uses six Blackfin processors). It's also used in devices like the DR-70D [tascam.com]. I'm actually surprised they are dropping support for it.

        • by arth1 ( 260657 )

          It's also used in devices like the DR-70D [tascam.com].

          Is that running Linux, though? I was under the impression that Tascam used a "true" embedded system without kernel/userspace separation.

      • Just FYI, Mikrotik's latest Cloud Core Routers use the Tile CPUs. I sure hope this mean's Tile isn't the end of the line for Mikrotik!
        • I sure hope this mean's Tile isn't the end of the line for Mikrotik!

          I would never have even considered using a new Mikrotik product after how unreliable the last generation was. I got two different routerboard devices and both were flaky AF. Putting something into customer hands that has to be power cycled frequently to function correctly is unacceptable.

    • by tlhIngan ( 30335 )

      if a Powerbook 100 is still supported.

      Actually, how does Linux support the 68K CPU? It doesn't have an MMU that's required. I know there was a version of Linux called uCLinux that worked on MMU-less (and MMU-lite) processors like the 68K, but I've found it quite unstable. No hardware support for protection means a bad pointer literally will take down the system.

      So it was neat, but completely pointless in the end. All it took was one bus error and you'd be rebooting.

      Now, there were later 68K CPUs with MMUs (

      • Heck, there were a couple of bad instructions in the 68k that made a "modern" OS impossible (i.e., you can call them from user mode) - that's why it was revised into the 68010 to allow proper user-supervisor mode separation.

        I think this is what you are referring to: https://en.wikipedia.org/wiki/... [wikipedia.org]

      • by Jody Bruchon ( 3404363 ) on Monday April 16, 2018 @08:45AM (#56445219)
        You're right about the stability; that's just the nature of the beast. When they refer to supporting the 68000, they're probably using a NOMMU system [nommu.org] which is a quirky thing but within its several limits it DOES work. As you noted, some features of modern CPUs and MMUs such as memory protection are not available. uClinux's kernel work was folded into the Kernel during the 2.5.x series. As of 2016 they were still alive and kicking.
      • by Megol ( 3135005 )

        The 68000 have no MMU internally and no support for virtual memory even with an external MMU, the "bad" instructions just exposed system state making pure virtualization impossible. But virtualization isn't a requirement for a modern OS and x86 fixed their abstraction leakages much later (20+ years) and in a much more complex way (but with advantages).

        The 68010 was hardware compatible with the 68000 but fixed both the virtualization problem and the MMU problem (the MMU itself had to be added externally).

    • Same as with any other kernel function - no verifiable use.

      The problem is that this buggers up my plan to travel back in time and upload the latest Linux source into Linus Torvalds' head when he was a student. The architectures removed were the only ones around then. How's that going to convince him, it he can't run it?

    • The powerbook 100 isn't and couldn't be supported--it's a 68000, with no potential for an emu (other than that bizarre hack that used two, and stopped one mid-instruction on a fault, using the other to fix things before resuming).

      Oddas it sounds, the Powerbook 100 was not part of the Powerbook 100 series.

      Apple basically handed the Macintosh Portable to Sony and said "shrink it." The result got called the 100.

      There was one insignificant difference, other than size, iirc, but I forget what it was.

      hawk, who

    • I'm curious what's considered an "obsolete CPU architecture" if a Powerbook 100 is still supported.

      The question is "obsolete" for who? Obsolete depends on who defines it.

      Obsolete due to lack of users and lack of working hardware?
      Obsolete due to lack of vendor support?
      Obsolete due to lack of maintainers?

      Ultimately in every traditional version of obsolescence one group decides quite arbitrarily when something becomes obsolete and it has nothing to do with the orderly discontinuation of products in a linear timeline.

  • Kernel 2.6 works just fine thank you very much sincerely my bosses sticking with Redhat 5 and 6.

  • Soo it is yet not bug fixed. it become stable till june so is it under developer option.
    • RC1 stands for Release Candidate 1. Meaning you're free to try it but is is not a mainline release so don't complain about bugs, help them get fixed.

  • ... these days. Even before flakey audio and bluetooth management.

    Seriously. Every other OS get's an easy 8 hours out of todays regular portable hardware, only Linux barely scrapes 4 hours. I'm a big Linux and FOSS fan but this is a problem that is really annoying and needs fixing ASAP. Windows, macOS and even Chrome have been on top of this for the better part of a decade and the Linux kernel still wastes gobs of energy. Unacceptable. This update is a step in the right direction. I hope it continues that way. Thumbs up for the kernel crew.

    • by sad_ ( 7868 )

      Seriously. Every other OS get's an easy 8 hours out of todays regular portable hardware, only Linux barely scrapes 4 hours... Windows, macOS and even Chrome have been on top of this for the better part of a decade and the Linux kernel still wastes gobs of energy.

      you do know that chromeOS uses the linux kernel?

    • Albeit no doubt modified. However those modifications should have been pushed upstream to the main line.

      The power wastage in desktop systems probably comes from the bloated GUIs that tend to come as the default on linux these days (hello Gnome and KDE) which seem to have their processes permantly at the top of the "top" cpu usage list if my system is anything to go by.

    • fwiw a lot of that can be fixed by turning down the brightness on your laptop screen. The backlight draws a lot of power, and Apple aggressively dims the screen.
    • Re: (Score:2, Interesting)

      by Dr. Evil ( 3501 )

      To much abuse, ridicule and disbelief, I’ve posted here many times about my problems with Linux on the desktop...

      This hasn’t been one of them. I have an Acer Cloudbook which gets 12-17h per charge. It can’t hibernate properly, crashes on resume from suspend, bit the fact that it gets such wicked battery life while ON, means I just leave it running in my bag all day.

      Although my audio stopped working, no fricking clue why. Who has time to deal with this stuff?

      • What OS are you running? I used to have issues with older Ubuntu and Debian and basically anything that uses old PulseAudio. Most have switched to ALSA which isn't too much better but works. I used to manually have to strip pulse and force the OS to use ALSA. But those days are fortunately over. And for ease of use Ubuntu 16.04 and Debian 9 from experience.

    • Which is quite interesting, given my experience with a Thinkpad Carbon X1 4th gen. I get around 10 hours on that running Windows, and can squeeze that to almost an extra hour if I really push hard on running nothing but Scrivener - but that requires Windows Update to behave, which it usually doesn't, pushing battery life below 9 hours.

      Running Linux, with Scrivener in Wine, I get 12-14 hours on it without even trying.

      In both cases WiFi is on, screen brightness at near minimum, and I'm using it pretty much co

    • Checklist:

      1. Are you running KDE, Gnome or Unity? If so, bury them with dignity at the bottom of the garden and install Enlightenment.
      2. Are you running SystemD? Ditto and install OpenBSD's init or Gentoo's OpenRC.
      3. Is the kernel tuned? It should not contain things you don't use. If power is a problem, that is the governor you should select. Run with the worst latency you can afford because you need more power to get equal performance with low latency.
      4. You might want to use the hoard memory allocator ove

    • by Hydrian ( 183536 )

      The biggest issues with power use is the hardware drivers. Many times open source drivers are written without any involvement of the hardware manufacture's company. The manufacturers often don't publish the open hardware/firmware specifications. If they do, they can be incomplete. To get the full specs, you often have to be apart of some big company and sign an NDA agreement. This is very incompatible with open source code.

      This leaves devices driver writers often black boxing the driver writing process so i

  • Ok, maybe from the mainstream kernel, but space missions use old hardware and you don't encourage recycling eWaste this way.

    Maybe have a tree of obsolete architectures, where those who want old stuff can go.

    Better yet, since architectures are split off anyway, have a subtree for each architecture (inline or otherwise). Let arch-specific distros merge the architectures they want and no others, let developers cleanly see what does what. It should make adding new architectures to (and removing old ones from) t

    • I don't get the complaint. Chances are the stuff in the new kernels won't even work on legacy hardware. Put your big boy pants on, download an older flavor and get some back ported security patches and you're in the game.. Problem solved.. Linux on old hardware is easy, sometimes. The hard part is finding a distro that works with little modification. If you're trying to run Linux on old hardware you're already doing it for fun anyways because if it was for production you would already.

What is algebra, exactly? Is it one of those three-cornered things? -- J.M. Barrie

Working...