Please create an account to participate in the Slashdot moderation system

 


Forgot your password?
Close
typodupeerror
×
AMD Intel Hardware

Linus Torvalds Discusses Intel and AMD's New Proposals for Interrupt/Exception Handling (linuxreviews.org) 149

Posted by EditorDavid from the making-an-exception dept.
"AMD and Intel have both proposed better ways of doing interrupt and exception handling the last few months," reports LinuxReviews.org.

Then they share this analysis Linus Torvalds posted on the Real World Technologies forum: "The AMD version is essentially "Fix known bugs in the exception handling definition".

The Intel version is basically "Yeah, the protected mode 80286 exception handling was bad, then 386 made it odder with the 32-bit extensions, and then syscall/sysenter made everything worse, and then the x86-64 extensions introduced even more problems. So let's add a mode bit where all the crap goes away".

In contrast, the AMD one is basically a minimal effort to fix actual fundamental problems with all that legacy-induced crap that are nasty to work around and that have caused issues...

Both are valid on their own, and they are actually fairly independent. Honestly, the AMD paper looks like a quick "we haven't even finished thinking all the details through, but we know these parts were broken, so we might as well release this".

I don't know how long it has been brewing, but judging by the "TBD" things in that paper, I think it's a "early rough draft"."
In the article (shared by long-time Slashdot reader xiando), LinuxReviews.org summarizes the state of the conversation today: Torvalds went on to say that while AMD's proposed "quick fix" would be easier to implement for him and others operating system vendors, it's not ideal in the long run. Intel's proposal throws the entire existing interrupt descriptor table (IDT) delivery system under the bus so it can be replaced with what they call a new "FRED event delivery" system. Torvalds believes this is a better long-term solution...

While the pros and cons of Intel and AMD's respective proposals for interrupt and event handling in future processors are worthy of discussion, it's in reality mostly up to Intel. They are the bigger and more powerful corporation. It is more likely than not that future processors from Intel will use their proposed Flexible Return and Event Delivery system. Their next generation processors won't, it will take years not months before consumer CPUs have the FRED technology. Remember, the above-mentioned technical document was published earlier this month [in March]. Things do not magically go from the drawing-board to store-shelves overnight.

Intel isn't going to just hand the FRED technology over to AMD and help them implement it. We will likely see both move forward with their own proposals. Intel will have FRED and AMD will have Supervisor Entry Extensions until AMD, inevitably, adopts FRED or some form of it years down the line.
They also note that Torvalds took issue with a poster arguing that microkernels are more secure than monolithic kernels like Linux. "Bah, you're just parroting the usual party line that had absolutely no basis in reality and when you look into the details, doesn't actually hold up.

It's all theory and handwaving and just repeating the same old FUD that was never actually really relevant."
This discussion has been archived. No new comments can be posted.

Linus Torvalds Discusses Intel and AMD's New Proposals for Interrupt/Exception Handling More

Linus Torvalds Discusses Intel and AMD's New Proposals for Interrupt/Exception Handling

Comments Filter:

  • FRED (Score:3, Funny)

    by nemyax ( 1376001 ) on Sunday April 04, 2021 @07:13AM (#61235146)
    > FRED event delivery Right, said Fred, gotta take the wall down, that there wall is gonna have to go!

    • Cuppa?

    • Bah, you're just parroting the usual party line that had absolutely no basis in reality and when you look into the details, doesn't actually hold up.

      In other words, your whole argument hinges on unreality. In a world that doesn't exist, and no reasonable person expects to exist or even has a realistic plan for, in any kind of reasonable timeframe, your argument simply makes absolutely zero sense at all.

      The name of the forum is "real world tech". Not "my private fantasy tech"."

      Ah, Happy Fun Linus is back :-).

  • Haven't they read the news these past years?

    That's like saying that huge dinosaur that got its head bitten off by the smaller dinosaur is bigger and more powerful because it's still in motion and warm. ;)

    Sure, it's Intel. They might grow a new head over time. But things still have changed quite a bit.

    Also, Intel is mostly bigger because they still make (most) of their own chips, and because of monpolism. The former doesn't count, as by that logic, you could add GlobalFoundries and part of TSMC to AMD, and t

    • Re: (Score:3)

      by Osgeld ( 1900440 )

      I am huge amd fanboi, but you do know intel a gazillion chips in markets where amd is still hoping to get higher single digit marketshare

  • FRED is overloaded (Score:3)

    by drinkypoo ( 153816 ) <drink@hyperlogos.org> on Sunday April 04, 2021 @08:16AM (#61235254) Homepage Journal

    FRED already refers to a whole bunch of stuff, least relevantly the freespace mission editor and most relevantly the Fucking Rear End Device that replaced cabooses.

  • IMHO they are beating a dead horse, x86 carries so much cruft from the last 50 years. It is bloated like Windows. Those propositions exist just to milk the cow as long as possible.

    • Re:x86 is a dead horse (Score:4, Informative)

      by Z80a ( 971949 ) on Sunday April 04, 2021 @09:52AM (#61235458)

      Sadly it's the best horse we have.
      The tie with the IBM PC compatible platform guarantees that the bootloaders, standard devices etc.. are all in the same place, instead of the DRM hell that is the ARM devives, where they can make every system boot and behave a different way, making efforts like linux impossible on em.

      • Re: (Score:2)

        by Mojo66 ( 1131579 )
        ARM is DRM hell? You obviously haven't heard about the M1 Macs...

      • Uhhh, as a firmware engineer I can't comprehend WTF you're even talking about.

        I assume you're talking about cell phones or something? Rather than processors? You know that Android is Linux, right?

        • Re: (Score:3)

          by Z80a ( 971949 )

          I'm talking about how x86 is specifically tied to the IBM PC compatible platform that is quite known and standardized, while ARM is not tied to any platform in special as Android, MacOS etc can be rebuilt by the manufacturers.

        • Re: (Score:2)

          by amorsen ( 7485 )

          If you buy a random x86 system you can be almost guaranteed that there is an UEFI firmware with reasonably well supported ACPI. This is true even if you buy extremely low end implementations (Raspberry Pi-style and below) or extremely expensive ones (Juniper MX routing engine). There is a good chance that an unmodified Linux distribution will boot on them, and if it doesn't, the changes required are usually tiny. Drivers can be hit and miss, but SSD and screen and keyboard and so on will work, perhaps with

      • > ARM devives, where they can make every system boot and behave a different way, making efforts like linux impossible on em.

        The majority of the CPUs in existence are ARM chips running Linux.

        • Re: (Score:3)

          by Z80a ( 971949 )

          A proprietary binary blob created by the corporation that manufactured the phone. Running your own build on it is a chore if not nearly impossible.

      • instead of the DRM hell that is the ARM devives, where they can make every system boot and behave a different way, making efforts like linux impossible on em.

        What are you talking about? Raspberry Pi, Android are just two examples or ARM on Linux. Just because an implementation on ARM can have DRM does not mean it must have DRM.

        • Re: (Score:3)

          by Z80a ( 971949 )

          You can run linux on ARM of course, it is even what android basically is, the difference is that you're vendor locked unlike with the PC platform.
          You can't make a linux iso that works with all the android devices for example because there's no standard platform tied to the CPU architecture.

          • You can run linux on ARM of course, it is even what android basically is, the difference is that you're vendor locked unlike with the PC platform.

            Again, what the hell are you talking about? Please tell me how I am vendor locked on a Raspberry Pi. Or on my ARM based home router. On Android you can install Linux on Android [tuxphones.com] with Ubuntu being the most popular.

            You can't make a linux iso that works with all the android devices for example because there's no standard platform tied to the CPU architecture.

            Errr what? I have no idea what you are saying. No version of Linux ISO works on every x86 device. No version of Linux ISO works on [insert platform here] including MIPS, Power, ARM whatever. Your complaint belies a fundamental misunderstanding of Linux.

            • Re: (Score:3)

              by Junta ( 36770 )

              Well, your Raspbian image isn't going to install very well on a Mediatek based platform. Going a bit upscale to UEFI equipped aarch64 servers, and now you are talking about an area where distributions do tend to be pretty viable across multiple vendors. I do think ARM is well on its way for the norm to be an x86-like ecosystem of well defined standards being the norm to enable cross-vendor compatibility, but there's a lot of lingering of the bespoke incompatible solutions. One brand new ARM device I have w

              • Well, your Raspbian image isn't going to install very well on a Mediatek based platform.

                And I never said that it would. I am merely contesting his point that somehow my Raspberry Pi is vendor locked.

                . I do think ARM is well on its way for the norm to be an x86-like ecosystem of well defined standards being the norm to enable cross-vendor compatibility

                Considering the wide range of capabilities of stock ARM cores, that is a hard ask as ARM chips run ultra lower power computing devices like watches to advanced smart phones. Some vendors still use 32 bit processors because it suits their needs.

                ISO images from all major distributions are able to pretty much work on any x86 based system commonly sold in the last 15 years

                ANY? I would bet you a x86 distro does not work on a x86-64 machine and vice versa. 15 years ago, Intel just released their Core 2 microarchitecture which wa

                • Re: (Score:2)

                  by Junta ( 36770 )

                  I said 15 years ago to specifically catch the era where the vast majority of chips would be x86_64 capable, and the modern distros frequently are x86_64 focused for Intel. Yes, you can find pretty tortured examples to defy it (a brand new 32-bit pentium clone that is embedded and usually runs freedos), but the general purpose x86 platforms don't require the user to give a second thought as to whether the OS installation image they have will work or not.

                  Sure, ARM is seen in microcontrollers, but even ignori

                • An x86 distro will work fine on an x86_64, they are all back compatible and can run x86 code.

                  An x86_64 distro won't work on an x86.

                  Of course this is ignoring specific driver issues. Newer equipment may require a newer kernel for driver support.

            • Comment removed based on user account deletion

              • He's talking about the fact that while you can make a boot image for a Samsung G-500 Supergood Phone Plus G5, you can't make a boot image that'll work on that, and all the other Samsung phones, and the Raspberry Pi.

                Considering that the two devices use vastly different CPU and hardware capabilities, why would anyone assume you could us the same boot image. Even if we talk about just the Raspberry Pi, early generations used 32-bit ARM processors.

                Whereas you can make an ix86/amd64 image that'll boot on virtually all modern PCs, whether made by Dell or HP, whether they contain AMD or Intel chips, whether they're in a mini PC style form factor or a full tower, or a laptop, and so on, as long as that image includes the relevant drivers. This is achieved via standardized firmware (including a standardized booting mechanism) and standardized hardware allowing standardized methods for probing the available devices.

                But you are missing that the x86's last major advance was x86-64 and that required maintaining 2 versions for years because consumers could have 32-bit or 64-bit processors. Today vendors can still use 32-bit ARM processors in their devices whereas AMD and Intel do not make 32 b

        • What are you talking about? Raspberry Pi, Android are just two examples or ARM on Linux

          Actually, they're examples of Linux on ARM.

          Raspi still has binary blobs and Android devices pretty much all have closed drivers, however, so they are still examples of platforms which are not fully open.

      • Re: (Score:2)

        by tlhIngan ( 30335 )

        The tie with the IBM PC compatible platform guarantees that the bootloaders, standard devices etc.. are all in the same place, instead of the DRM hell that is the ARM devives, where they can make every system boot and behave a different way, making efforts like linux impossible on em.

        But there are significant differences still. Enabling A20, an operation dating back to the 80s, can be done by one of 4 different incompatible methods. And you don't know which one would work. So you have to try them all.

        Then t

    • So, is it a dead horse or a milking cow?

    • Re:x86 is a dead horse (Score:4)

      by serviscope_minor ( 664417 ) on Sunday April 04, 2021 @11:12AM (#61235688) Journal

      The death of x86 has been greatly exaggerated.

      Basically people have got awfully excited that the M1 is competitive with x86 in one niche (mid range laptops, essentially). If you want something faster for general computing, x86 is still your best bet.

      • Basically people have got awfully excited that the M1 is competitive with x86 in one niche (mid range laptops, essentially). If you want something faster for general computing, x86 is still your best bet.

        Many, many reviewers of the M1 Macs have said that the M1 is very competitive for general everyday computing. It is not competitive for niche uses like video productivity in some cases. In some cases, like h264/h265 encoding it actually beat older Intel Macs.

        • Well, a mid range laptop is ok for everyday computing, so I guess we agree?

          • You said: " one niche (mid range laptops, essentially)". What about lower end laptops? What about lower end desktops (Mac Mini)? That is three categories.

            • Well it's use in low end laptops is completely unproven. Apple don't sell low end laptops, and the chip uses HBM which is pretty pricey stuff, but offers higher bandwidth and much lower power draw then DDR4.

              And yes I forgot about low end desktops. Good point.

              • Anything with 16GB is a low end laptop, sorry.
                It's a Gucci chromebook.

                • Anything with 16GB is a low end laptop, sorry.

                  No. This is (a) snobbery and (b) unrealistic. You can still buy laptops with 4 and 8 G, and 64 is about the max you get in anything less than a mobile workstation (I hesitate to call those beasts laptops unless you own asbestos underpants). If you care about mobility, then 16G is the max you can get in a lightweight laptop, even with a good CPU.

                  A 16G laptop is fine for even mid sized dataset analysis, hence mid range.

                  • Of course it's snobbery.
                    The "ranges" are entirely subjective.

                    You can still buy laptops with 4 and 8 G, and 64 is about the max you get in anything less than a mobile workstation

                    I hesitate to call something with 4 and 8GB a laptop. That's an iPad with a keyboard.
                    As for mobility... 64GB (and 128GB) laptops are quite mobile. Sure, they're heavy. But tools generally are, no?

                    (I hesitate to call those beasts laptops unless you own asbestos underpants)

                    Quite right. I wouldn't put any of my beefier laptops on my lap for any kind of real use. They're too heavy.
                    Heat actually isn't that big of a problem. They generally have very adequate airflow, so the chassis itself doesn't get that hot. The person or

                    • Of course it's snobbery.
                      The "ranges" are entirely subjective.

                      Snobbery is stupid. It sounds like you're defining midrange as capable of "real work" where you alone define what "real work" is and anything less memory intensive isn't "real work". This is also completely blinding you to the idea that people have other usecases to you.

                      I hesitate to call something with 4 and 8GB a laptop. That's an iPad with a keyboard.

                      A low spec laptop is still a laptop. Can still run arbitrary code, run whatever OS you can inst

                    • Snobbery is stupid. It sounds like you're defining midrange as capable of "real work" where you alone define what "real work" is and anything less memory intensive isn't "real work". This is also completely blinding you to the idea that people have other usecases to you.

                      While it sounds to me like you're defining "midrange" as capable of your "real work" (what was that about analyzing "mid sized" data sets?)
                      It's not blinding, it was me pointing out that what you define as mid range is entirely subject to you.
                      It looks like we agree here ;)

                      A low spec laptop is still a laptop. Can still run arbitrary code, run whatever OS you can install etc. You can easily compile code on an 8G laptop, etc. You benefit from lots of RAM if you're working with large codebases, but not everyone is.

                      It is still a laptop. And so is a chromebook.

                      Portable. n. has wheels and not technically bolted to the floor. Also, see crowbar.

                      Unsure what you're trying to say. That a laptop that may weigh 2-3kg more than a lightweight laptop is only mobile out of technicality?

                      lolwtf no. Tools have a purpose, and bigger and heavier is not necessarily better.

                      No one said better. You inserted that word entirely of you

                    • No one said better.

                      You are literally saying that below:

                      Sure, but would you be annoyed if your 2kg hickory handled club hammer was 5kg, and capable of much more?

                      This definitely qualifies as "not even wrong". The 5kg sledge isn't capable of more, it's capable of different. There is very little overlap between the tasks I would accomplish with a 5kg long handled sledge and a 2kg short handled club hammer. No one even makes such a thing; I've never seen a club hammer over 2 kilos.

                      So no: I would be very annoyed

                    • You are literally saying that below:

                      Only if we're using the millennial definition of literally.
                      I literally said heavier was more capable. Is that not the case?
                      Better is putting words in my mouth.
                      More capable isn't remotely synonymous with better. I'll trust I don't need to provide examples to demonstrate that logic.

                      This definitely qualifies as "not even wrong". The 5kg sledge isn't capable of more, it's capable of different. There is very little overlap between the tasks I would accomplish with a 5kg long handled sledge and a 2kg short handled club hammer. No one even makes such a thing; I've never seen a club hammer over 2 kilos.

                      You missed the point of the hypothetical.
                      It wasn't a matter of what actually exists.
                      Ultimately, a hammer is a hammer, which makes it a terrible analogue for a computer- and that's on you. Perhaps it's partially my fault for in

                    • I literally said heavier was more capable. Is that not the case?

                      I don't know what misunderstanding about tools would lead you to believe that heavier hammers are "more capable".

                      Tell you what, since you don't believe me try the following:

                      1. Buy a 5 kilo sledge
                      2. Attempt to put in a panel pin

                      A heavier sledge is not "more capable" because there are jobs that are essentially impossible due to its size and weight. Not to mention the steel head is completely unsuitable for many tasks and would in fact completely

                    • I don't know what misunderstanding about tools would lead you to believe that heavier hammers are "more capable".

                      You conveniently skipped over:

                      You missed the point of the hypothetical.
                      It wasn't a matter of what actually exists.
                      Ultimately, a hammer is a hammer, which makes it a terrible analogue for a computer- and that's on you. Perhaps it's partially my fault for indulging your bad analogy.

                      I suspect it was no accident. You love beating on dead horses.

                      No, it is not the case.

                      Except it is, in the context with which we were speaking.

                      Right: me providing examples of how a 5kg sledge is not the most capable hammer in the world is not a straw man. No one even makes 5kg single hand hammers because such a device is completely useless, not "more capable" as you keep insisting.

                      Sure it is. It's an attempt at building up an argument you can win that isn't remotely relevant to the logic train we were actually arguing.
                      So desperate are you to continue trying to lift this poor man of straw, you won't even acknowledge that I conceded that your analogy was ridiculous, and I was silly for indulging it.

                      ORLY? Allow me to quote the original thing I was replying to:

                      Anything with 16GB is a low end laptop, sorry.

                      That's "anything" not "any mac".

                      OK, that's fair. Rather, I should have s

                    • So yes, your $1500 16GB "ultrapremium" laptop is a Gucci Chromebook.

                      In addition to tools (specifically hammers) it appears you also have no grasp of fashion or chromebooks.

                      One wears Gucci because (charitably) you want to look nice and (less charitably) because you want others to know you wear Gucci. Reality for any one person will lie between those extremes. Lenovo isn't a fashion brand: you don't buy a Lenovo because you want to show off. I mean you might, but it'd be a pretty odd choice unless you wanted

        • What you are describing is EXACTLY what he said. the M1 is good enough as a md range processor (that is everyday computing), but still is greatly outpaced at the top end. mid range is actually all most people need though.
          • His said it was good for one niche which is mid range laptops while ignoring lower end to mid desktops and lower laptops.
  • What are the chances new bugs will just be introduced and require another set of workarounds to maintain?

    • This is what undermines the "Intel is big, they can do whatever they want." AMD is expected to go from 20% server market share by the end of the this year. By the time this is in products, Intel is going to have a hard time forcing anybody to do anything. They'll have to convince people to support it, and if they try to make it proprietary, that's going to be harder to do.

    • Re: (Score:3)

      by ghoul ( 157158 )
      Thats how engineers keep their jobs. We are not doctors. We dont get an endless supply of broken things to fix. If we want things to fix we have to break it first.
  • A new cpu that eliminated all the backward comparability may have been a better choice. Never actually used one, but who knows.

    • You may want to give the ATtiny84 a try.

    • Itanium did have a backwards compatibility subsystem, but in the first 733MHz 64 bit chips was equivalent to 100 MHz Pentium. One of many things on the long list of failures and poor performance that doomed Itanium to be killed by the x86-64 chips of Intel and AMD

    • I always wondered who bought Itanium based systems. The only reason seems like you needed OpenVMS on modern hardware. I occasionally check eBay because Itanium seems interesting from a collectors standpoint but the prices are insane. Wonder how many Itanium boxes are still out there running.

      • > the prices are insane. Wonder how many Itanium boxes are still out there running.

        Legit, some people threw their Itantic boxes off the roofs of their office building - the supply must have rapidly diminished.. There must be an old story on here about it.

        They say it was epic.

    • A new cpu that eliminated all the backward comparability may have been a better choice. Never actually used one, but who knows.

      The Itanium VLIW architecture was dependent upon a magic compiler to produce good performance, and Intel was never able to produce a compiler which produced performance in line with their promises. There was a lot of hope that they would manage it since their x86 compiler was so very good at optimization, but it never happened and Itanic died.

  • Code segments should be immutable, write-once. This requires ending INT3 injection as the break-point insertion approach. Instead there should be a sparse-per 8 page debugging bitmap. So many security problems would be erased if CPUs enforced immutable code segments.

  • Why should we believe Intel's "solution" will be any better?

    Besides, 10 years from now - I'm betting/hoping Intel is (at most) just a fab provider, and no one will be running Intel's architecture anymore.

  • We're going to have Supervisor Entry eXtensions along with Flexible Return and Event Delivery. SEX with FRED!

Slashdot Top Deals

Some people manage by the book, even though they don't know who wrote the book or even what book.

Close