Announcement

Collapse
No announcement yet.

AMD Zen 2 Improvements For LLVM Have Been Held Up For Months By Code Review

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • #21
    Originally posted by agd5f View Post
    Why are you getting mad at AMD? Michael is the one complaining. As was stated in the article, AMD provides branches and packages for just about everything so you don't have to wait for everything to get upstream. Upstream is unpredictable. Realistically, it's almost impossible to get everything upstream ahead of launch. In most cases hardware, distros, and components (like llvm, mesa, gcc, etc.) are on different schedules and none of them align. You could try and have everything ready by 6-9 months before launch, but in most cases that's not feasible. Hardware development schedules are not designed that way. You generally don't write support for something and then sit on it for 6-9 months wanting for the part to launch. Generally development goes right up until launch, so even putting out early development code much earlier is not likely to work 100% once production hw launches. You have a bit more time on the CPU side because the schedules are longer, but even those are getting compressed. You work to get a solution out to your customers for launch and you try and align with upstream as much as possible to make sure you get the changes in for the long term.
    I realize it's not easy, but your competition (Intel) doesn't seem to have much difficulty doing it. If your answer is that Intel has more resources, well... That's true. But it's kind of pointing out a major problem if you intend to compete with them.

    We work to ultimately get everything upstream, but the fact is very few if any customers actually want upstream.
    And that's the actual issue. AMD doesn't care about this, because they don't think their customers care. I wish they viewed actual consumers as customers, but it's obvious they don't.

    Comment


    • #22
      Originally posted by Space Heater View Post

      Yep, it's pretty apparent that AMD as a company doesn't understand how important it is to work upstream. Just look at how the community has needed to commit temperature reading support for every iteration of AMD's CPUs, and how perpetually under-staffed their Linux GPU team is. They should feel embarrassed.
      Man, what's the issue here? You have purchased a system with an AMD CPU because when you purchased it it was the better option for you. If you do have a Zen 2 CPU, what I read here is that there is something that AMD could have done better, but they have done it in a suboptimal fashion, yet you have purchased the AMD system because it was the better choice.

      Now you read an article showing that your AMD system has a chance of becoming better still, short to mid term, when those changes are finally upstreamed and distributed to a distribution near you. Yup, AMD could have done better. And what about the myriads of things any company of any size could do better? This one got publicized, that's all. Yet you make choices and buy products from one company over another's because product A fits your needs and is better than product B.

      You have chosen to complain and be mad about... well, nothing since the system you have, is working for you NOW. And it is going to get better in a rather close future if you keep your Linux distrib up-to-date. Else... it will still be as good as it is now.

      You're criticizing the whole company because one man... well upstreamed one thing recently and you think it could have been upstreamed before. But the whole company should now be embarassed. Man, I hope you NEVER EVER do a single thing wrong, or in this case, just a bit late, else your WHOLE family, colleagues, town, community etc should feel embarassed.

      Oh, in the off-chance you haven't actually purchased an AMD Zen2 system and are still complaining, then... your posts would just amount to trolling. But if you have... well enjoy your system and just know that the world is full of things that aren't perfect and we simply navigate our way between proposals, chosing the one that seems to fit best.
      Last edited by cyclistefou; 23 October 2019, 08:15 AM.

      Comment


      • #23
        Originally posted by Space Heater View Post

        That's a false choice. You don't need to delay a product, you start the software development earlier. Other companies manage to do this, and this hurts AMD's adoption in the server space.
        The only thing that happens when the compiler is not optimized for the CPU at launch is that... it's not optimized for the CPU at launch.

        If it works and pulls out very competitive scores with whatever compiler support there is at launch, why would customers chose to opt for another CPU which is already fully optimized? I mean, I'm not saying one company should not optimize compiler before launch... if it can, then by any means do it - what I mean is what matters 1st is the actual fit that one architecture has to your needs and budget at the time of purchase, plus the perspective of evolution / maintenance according to your needs (consumer / enterprise / datacenter etc).

        Comment


        • #24
          Originally posted by cyclistefou View Post

          Man, what's the issue here? You have purchased a system with an AMD CPU because when you purchased it it was the better option for you. If you do have a Zen 2 CPU, what I read here is that there is something that AMD could have done better, but they have done it in a suboptimal fashion, yet you have purchased the AMD system because it was the better choice.
          You clearly didn't read my post if you think this is just a one-off event or that I'm demanding perfection.
          Originally posted by cyclistefou View Post
          You have chosen to complain and be mad about... well, nothing since the system you have, is working for you NOW. And it is going to get better in a rather close future if you keep your Linux distrib up-to-date. Else... it will still be as good as it is now.
          Except it doesn't always work, GPU crashes that hang your entire system aren't what I'd call "working". Having no temperature reading support within 4 months of launch isn't what I'd call an advanced feature request. But you're not hear to actually have a discussion, you're just upset someone isn't singing AMD's praises 24/7.
          Originally posted by cyclistefou View Post
          Oh, in the off-chance you haven't actually purchased an AMD Zen2 system and are still complaining, then... your posts would just amount to trolling. But if you have... well enjoy your system and just know that the world is full of things that aren't perfect and we simply navigate our way between proposals, chosing the one that seems to fit best.
          It's a bit pathetic seeing someone who is so blindly brand-loyal lose their mind at any criticism of their favorite company. Seek help.

          Comment


          • #25
            Originally posted by agd5f View Post
            AMD provides branches and packages for just about everything so you don't have to wait for everything to get upstream.
            If only they were in some central place... instead you have to get one part from github, another part from people.fdo and yet other things from fdo git. Then there was packages.amd.com but that appears defunct now. And finally (for ROCm on APUs) you have to get 3rd party patched stuff because AMD can't be bothered to support their very own products.

            Sorry AMD, with that mess you are making both distro packagers' and users' life unnecessarily hard.
            Originally posted by agd5f View Post
            but the fact is very few if any customers actually want upstream. Most customers want support for a 2 year old enterprise distro or their own custom distro or some other slow moving embedded distro depending on the part, at launch.
            AMD is under no obligation to make it work under any possible situation. However the most basic requirement is that you take an AMD product on release day, boot the latest Ubuntu release, and can expect it to work, even in a sub-optimal fashion, until you install the latest driver code. But AMD even fails this test three months after launch. With a distro that is among the most willing to cooperate with hardware vendors in order to get OOTB user experience in order, and that is on a highly predictable release schedule. Heck did nobody at AMD stick the first Ubuntu 19.10 beta into an RX 5700 machine and notice it not boot to a desktop? Canonical will even do that for you if you send them a unit.

            Comment


            • #26
              Originally posted by Space Heater View Post
              You clearly didn't read my post if you think this is just a one-off event or that I'm demanding perfection.
              I quoted your entire post my love. You didn't provide any fact to back your claims up apart from the temperature stuff which I agree is an issue on AMD side.
              You based your rant on an article about lack of optimization of the compiler and made it look like it equates with other issues. Temperature reading stuff _IS_ an issue, lack of optimization of the compiler at launch just isn't - it's just a delayed opportunity. Then you blamed the entire company for that.
              It is for this reason I have stepped in, trying to deal with the fact that lack of optimization of the compiler should not be treated as an actual issue, and trying to put some rationalization in the discourse.

              Originally posted by Space Heater View Post
              Except it doesn't always work, GPU crashes that hang your entire system aren't what I'd call "working". Having no temperature reading support within 4 months of launch isn't what I'd call an advanced feature request. But you're not hear to actually have a discussion, you're just upset someone isn't singing AMD's praises 24/7.
              I haven't experienced the GPU crashes as I don't use any AMD GPU, but they are certainly a PITA. I hear everyday about GPU crashes and I see so many of them, most of them on Intel or nVidia GPUs because I don't use AMD GPUs (so much for your AMD fanboy gratuitous accusation) so the only thing I can say is "crashes happen". Each time they do, I am tempted to curse the company... but I try to avoid doing it on the net... or do it in a constructive manner.
              For the temperature stuff, yes as said it's a pity AMD doesn't do it well and you know what? If it's too important to you then by all means delay your AMD purchase, or purchase Intel stuff as it has better temperature reading, hence better product for you. That's what competition and comparison gives you man: choice.
              Oh and I don't have AMD CPUs either. I have only Intel stuff because when I was shopping for computer stuff, Intel ruled and I chose what was best then. If I was shopping today, I would probably take AMD, but it's just about evaluating the available products when the need to purchase something comes. Who knows if Intel or ARM will beat the crap out of AMD tomorrow, I will take the product that fits my needs best at my price.

              Originally posted by Space Heater View Post
              It's a bit pathetic seeing someone who is so blindly brand-loyal lose their mind at any criticism of their favorite company. Seek help.
              It is now established you're trolling. I am no fanboy of any company. Just want to provide reasonable arguments to help avoiding needless brain heating for details, and focus on informed discussion with apples-to-apples comparisons. But your nick "Space heater" suits you well. Cool of a bit man. May I return you your last advice?

              Comment


              • #27
                Originally posted by Space Heater View Post
                You clearly didn't read my post if you think this is just a one-off event or that I'm demanding perfection.
                I quoted your entire post my dear. You didn't provide any fact to back your claims up apart from the temperature stuff which I agree is an issue on AMD side.
                Your rant was based on an article about lack of optimization of the compiler and made it look like it equates with other, actual issues. Temperature reading stuff _IS_ an issue, lack of optimization of the compiler at launch just isn't - it's only a delayed opportunity. You put the blame on the entire company for that.
                It is for this reason I have stepped in, trying to deal with the fact that lack of optimization of the compiler should not be treated as an actual issue, and trying to put some rationalization in the discourse.

                Originally posted by Space Heater View Post
                Except it doesn't always work, GPU crashes that hang your entire system aren't what I'd call "working". Having no temperature reading support within 4 months of launch isn't what I'd call an advanced feature request. But you're not hear to actually have a discussion, you're just upset someone isn't singing AMD's praises 24/7.
                I haven't experienced the GPU crashes as I don't use any AMD GPU, but they are certainly a PITA. I hear everyday about GPU crashes and I see so many of them, most of them on Intel or nVidia GPUs because I don't use AMD GPUs (so much for your AMD fanboy gratuitous accusation) so the only thing I can say is "crashes happen". Each time they do, I am tempted to curse the company... but I try to avoid doing it on the net... or do it in a constructive manner.
                For the temperature stuff, yes as said it's a pity AMD doesn't do it well and you know what? If it's too important to you then by all means delay your AMD purchase, or purchase Intel stuff as it has better temperature reading, hence better product for you. That's what competition and comparison gives you man: choice.
                Oh and I don't have AMD CPUs either. I have only Intel stuff because when I was shopping for computer stuff, Intel ruled and I chose what was best then. If I was shopping today, I would probably take AMD, but it's just about evaluating the available products when the need to purchase something comes. If Intel or another competitor (maybe even a solution based on ARM) beats the crap out of AMD tomorrow, then I would take the product that fits my needs best at my price.
                I don't praise or defend AMD. I rather try to put things in proportion, that's all.

                Originally posted by Space Heater View Post
                It's a bit pathetic seeing someone who is so blindly brand-loyal lose their mind at any criticism of their favorite company. Seek help.
                It is now established you're trolling - I am no fanboy of any company and don't even have any AMD product currently - and that's by choice! Please could you try to listen to reasonable arguments, and avoid needless brain heating for things that aren't remotely important? I try to focus on informed discussion with apples-to-apples comparisons.
                Your nick "Space heater" suits you well, but try cooling of a bit man, it will make you a more agreeable person. Are you like this IRL?

                Comment


                • #28
                  Originally posted by cyclistefou View Post
                  I quoted your entire post my dear. You didn't provide any fact to back your claims up apart from the temperature stuff which I agree is an issue on AMD side.
                  You quoted my post but ignored the contents of it. The facts are evident from the git history and the corresponding phoronix articles talking about temperature support.
                  Originally posted by cyclistefou View Post
                  You put the blame on the entire company for that.
                  It is for this reason I have stepped in, trying to deal with the fact that lack of optimization of the compiler should not be treated as an actual issue, and trying to put some rationalization in the discourse.
                  You claim rationality but all you haven't contributed anything to the discussion. AMD has left temperature reading support up to the community since forever, this is not their first zen-based product launch (Ryzen 1k and 2k, Threadripper 1k and 2k, Epyc, RavenRidge, Ryzen 3k). We are supposed to believe that they can't provide simple temperature offsets for any of these within 1 month of launch? They have taken months to release their fusion hub controller driver, critical for hybrid AMD laptops, and it's now debatable if they will ever release it.

                  Your go-to response is essentially "oh well it will maybe be fixed later!" or "nothing's perfect!", but the fact is that other companies do manage to have better support, and it's not just a matter of money. AMD should be embarrassed, this isn't their first product launch, and this isn't their first time working with Linux.
                  Originally posted by cyclistefou View Post
                  Your rant was based on an article about lack of optimization of the compiler and made it look like it equates with other, actual issues.
                  If you had read my posts (and not just quoted them) you'd clearly see they never were solely talking about the lack of compiler optimizations, that's just a convenient fiction you invented so you could justify yelling at me. You also just sidestep any content in my posts and just call them rants.
                  Originally posted by cyclistefou View Post
                  I try to focus on informed discussion with apples-to-apples comparisons.
                  Your nick "Space heater" suits you well, but try cooling of a bit man, it will make you a more agreeable person. Are you like this IRL?
                  Are you like this in real life? Unable to read comments without adding your own delusions? Double-posting because you're so upset after reading my posts? You're just a troll.

                  Comment


                  • #29
                    After reading through this entire thread I found it difficult to sort out what the real issue is that people are having with this article. Seriously.

                    If a compiler can create code that will run on a target CPU, then that target CPU can be used for that code.

                    In this case I got the impression that code can be created for these processors, so those processors can be used by the public. What seems to be the issue is this: "Is the resulting code optimal for that processor?"

                    Anyone that has observed Linux kernel development over the past few decades, since 1995 in my own case, probably realizes that code optimization in the kernel takes place over a period of time. Rarely is code ever "optimal" on Day 0. It works, but ongoing research usually find ways to make it better. I think the same logic applies to compilers. Unless a target CPU is completely unsupported by a compiler, the resulting code for the target CPU should run, it just won't run "optimally". Distributions like Gentoo permit such tweaking of x86 CPU features for the compiler, but Gentoo code compiled with more general X86 CPU features still runs, just not "optimally".

                    The AMD code will find it's way upstream, eventually. As AMD's income streams grow let's hope they return to something like their OS Research Center.

                    In closing, all of the ranting in this thread about code not being upstream, but nobody actually complaining that they cannot create compiled code for the target CPU, reminds me of nothing more than:

                    "bike shedding"

                    HTML Code:
                    https://en.wikipedia.org/wiki/Law_of_triviality

                    Comment


                    • #30
                      Originally posted by Space Heater View Post
                      _some actual facts (yes really)_
                      _too much crap unfortunately_
                      I read your last message. You have some points which I won't debate because I agree and have always agreed on several of them (not all) - indeed AMD is far from a perfect OSS citizen (though better than many). My intention in this thread was simply to demote the topic of compiler optimization so that it is not equated to actual issues, some of which you have mentioned.

                      However your aggressivity and tendency to gratuitously insult people makes you the actual troll (you are definitely not the worst of them, but still you behave like one) and frankly you should be the one embarassed, but you need to listen to your conscience first. Here's hoping that you are only like this on the Internet forums - due to the apparent freedom to launch ad hominem made-up attacks on random people - and that you don't jump randomly at flesh and bones people to spit nuisance.

                      Well anyway, it's your problem. I'm not offended, just... well a bit sad to see that it is pretty hard to discuss like civilized humans on a forum without one or several participants jumping at the other's throat with... baseless insults.

                      Not worth discussing this anymore, won't respond to future offenses if any - I don't care. Maybe we will have other more civilized exchanges in the future if both of us want to react to an article.
                      Last edited by cyclistefou; 24 October 2019, 04:55 AM.

                      Comment

                      Working...
                      X