Announcement

Collapse
No announcement yet.

A Quick Look At The Firefox 66.0 vs. Chrome 73.0 Performance Benchmarks

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

  • #61
    Originally posted by treba View Post

    Thanks for the benchmark! Although the motionmark benchmark puzzles me...it's one of the things where webrender should shine. For comparison, see https://www.phoronix.com/scan.php?pa...der-beta&num=2
    My guess is it isn't turning on for some reason. The chance is very slim two different renderers should perform basically the same in several scenarios. Or maybe there's a bottleneck elsewhere holding both renderers back.

    Comment


    • #62
      Originally posted by starshipeleven View Post
      Really, the main reason he does not want C++ in the kernel right now is that he does not see benefit in using most C++ features at all (which is 100% understandable as they were made for userspace application development, aka to abstract complexity so that the developer can deal with it faster)
      That's bollocks. C++ has several very low overhead features that enforce better quality code no matter if it's kernel or user space. Userspace programs often use more abstractions, but some of the higher level features already become useful in projects in the range of 10k-100k LOC. For instance namespaces come to mind. Many C programs emulate namespaces using naming conventions. This is a tradeoff and requires a bit more developer attention in C. Generics and generative programming are also highly useful. You can emulate generics with untyped code and/or wrappers, with lower performance and less type safety. Or you could repeat yourself, which leads to more expensive maintenance and larger codebases. C++'s templates can replace external code generators in some cases. C++'s RAII also extends C's auto memory management. C++'s better types make things like type safe tagged unions possible. Sure these are rather tiny features compared to the whole Linux codebase. C++ might also introduce overheads that are not so fancy. Still even if C++ had some interesting features, I'd rather consider a kernel written in Rust than C++.

      Comment


      • #63
        Originally posted by kyrios View Post
        In real world situations in which specific cases is there a noticeable performances difference?
        Basically any more advanced "web application" that works like an actual application running a JS client-side. I use some "websites" that are more like actual programs. They pull data off websockets and process it and present it in various graphical ways. The difference between Chromium and Firefox on such sites - even with websocket rendering and layers.acceleration.force-enabled and gfx.canvas.azure.accelerated - isn't just huge, it's works and is usable vs really laggy and basically unusable. If you're browsing some random old-school static webpage with some HTML formatted text and some pictures then it really does not matter which browser you use, there's no way you'll notice any difference.

        The web in general is going in the direction of more and more JS and more advanced "web applications" so Firefox is basically out of the game unless it by a miracle manages to become a whole lot faster. I would ideally like to be able to use Firefox for everything, I liked how it used to respect my privacy and used to be highly configurable with plugins. It's sad that for an increasing number of sites it's just not a choice. In some specific cases using Firefox on my Ryzen 1600X desktop with a RX570 GPU feels worse and slower than Chromium on my Penium N4200 laptop with a iGPU.

        Comment


        • #64
          Originally posted by starshipeleven View Post
          I was referring specifically to that quote already, and since I'm not a little shit like you I'm not going to post only a snippet of it.
          Doesn't fucking matter, "subpar programmer" is a very clear and concise definition, no amount of your fantasy dreamland will change that fact.

          I know it hurts but it's just facts bud.

          Originally posted by starshipeleven View Post
          As you can clearly see above, more than 70% of the actual message is spent whining about C++ features or libraries. I'm personally in agreement about his "object model is crap" general theme.

          Of course he is also calling idiots all the people that think the features he hates are good, but that's just a logical consequence of hating C++ language features.
          Actually no, what else would he fill his message with? Just insults? Of course he had to detail things he disliked for the rest of it. He's not a little shit like you.

          Furthermore most of his whining is not about the language itself but what it "leads to", aka the people using it and the community. It's not the fault of the language because it doesn't force you to use them. It's the community, again, the trash programmers. And same with Rust.

          Sure you can code with unsafe blocks everywhere in Rust, but the language was not designed for that (unlike C++ which works ok as a C replacement with the same features!) and is ugly as hell to do it. Therefore Rust (the language) creates an even shittier community which attracts even more idiots.
          Last edited by Weasel; 25 March 2019, 12:40 PM.

          Comment


          • #65
            Originally posted by hreindl View Post
            amazing how you prove your stupidity whenever there is a opportunity to do so

            first there is a difference between a kernel and userland, second I have other than you no problem with any programming language no matter if I use it personally or not

            finally given the age of rust the chances are good that Linux if it would start from scratch now would use it

            your head is so full of shit that you don't see why a still compiled language with more security against programming errors is
            a way better decision than pure c/c++ which was chosen for many projects because at that time where nothing better available
            Spoken like a true script kiddie.

            Comment


            • #66
              Originally posted by hreindl View Post
              says the number one idiot of phoronix - why do you fucker use C/C++ at all instead ASM? :-)
              At the risk of repeating myself: because asm is not portable across uarch in terms of performance (which is its goal after all). I do use asm though, just not all the time.

              tl;dr: There's no "-mtune=bla" in asm to tune the same code for two different uarch.

              Comment


              • #67
                Originally posted by hreindl View Post
                i thought that's all from the devil because you handrcat optimize your holy code - dumb liar :-)
                Am I supposed to know what "handrcat" means?

                Comment


                • #68
                  Originally posted by hreindl View Post
                  handcraft - idiot
                  So what does that make you writing unreadable typos?

                  And I don't know what "handcraft" is supposed to mean. If you mean writing it instruction by instruction, sure, but as I've already said the benefit/efficiency varies with each uarch. Sometimes do "handcraft" though, whether by inline asm or just straight up asm functions.

                  But stuff like organizing your data structures or code flow to be more efficient is a design decision (especially when vectorized), doesn't matter the language. Thing is, only C (and C-like C++, aka proper C++) and asm programmers tend to think that way. Others think about "mah beautiful codez omg" and other abstract subjective emotional hysteria. Rust fits there.

                  Comment


                  • #69
                    Originally posted by hreindl View Post
                    either you are capable to do it without any "but" or shut up!
                    You're a lost cause. It's not about writing it once, it's about having to write it multiple times if you want absolute top performance. Which is applicable only to some hot functions in general.

                    I'm not saying writing in assembler once is "harder" (it is just a bit, but not significant). I'm saying having to tune it to another uarch is too much waste of time if you want to write the entire code that way.

                    Comment


                    • #70
                      Originally posted by hreindl View Post
                      what you little idiot refuse to realize is that there is little no no performance win C++ versus Rust and the same time you even admit "applicable only to some hot functions in general" what let you look like a moron, but well, you where also the shithead shouting against compiler hardening flags because of your holy perormance while you as always never provided any numbers backing your bullshit
                      No, your time is applicable only to a few hot functions. You can rewrite hot functions for more uarch and update them constantly. But not the entire code base. Unless you can stop time, but most people can't. So try harder.

                      I also don't give a shit about your numbers, if we are to go by "applicable" performance. For example, something that takes 2 clock cycles versus 1 is twice as fast, a 100% performance boost, but obviously you wouldn't care about it unless it was in a tight loop. That doesn't mean the optimization doesn't apply, because it does, you just don't give a shit about it. And I don't care about your opinion so...

                      Comment

                      Working...
                      X