Slashdot Mirror
X11 Chrome Reportedly Outperforms Windows and Mac Versions
An anonymous reader writes "In a curious contrast to conventional wisdom, there are reports of X11 Chromium being faster than Windows or Mac versions. In the thread titled 'Why is Linux Chrome so fast?,' a developer speculates that it is due to the use of X11 capabilities: 'On X-windows [sic], the renderer backingstores are managed by the X server, and the transport DIBs are also managed by the X server. So, we avoid a lot of memcpy costs incurred on Windows due to keeping the backingstores in main memory there.' Has the design of X11 withstood the test of time better than people tend to give it credit for?"
From the later discussion on that topic, it seems the conclusion was that windows had a large history in the profile and may be bitblt'ing the first draw operation from main memory. Both of which have an impact on how slow it feels to the user.
09F91102 no, 455FE104 nope, F190A1E8 uh-uh, 7A5F8A09 that's not it, C87294CE no. Ah! 452F6E403CDF10714E41DFAA257D313F.
Framebuffer is an unaccelerated bitmap display, X11 is an accelerated graphics layer (that can use a framebuffer)
something that writes directly to a framebuffer is going to need a lot of additional programming in order to be as fast as X11 is.
Paying taxes to buy civilization is like paying a hooker to buy love.
This doesn't say it's optimized for Linux - it says that it runs well on *X11*. X11 is used on almost all Unix derivatives, not just Linux. Most importantly though, all indications are that while Google intends to use Linux (the kernel) for ChromeOS, they have made some statements that would indicate that they likely will not use X11.
"People who think they know everything are very annoying to those of us who do."-Mark Twain
If you can take risk of re-compiling every X related app/library in case you give up in future, try the semi official/unofficial at http://xquartz.macosforge.org/ , it is newer than the Apple bundles. Install anything with the help of Fink/Macports like Konqueror from KDE 3 and see the amazing GUI speed, scroll speed, widget drawing speed.
I don't understand, as an OS X user, why a modern x.org on a good, supported hardware should be surprising to give better results. Also remember the insane things x.org has to do on OS X like using Aqua layer.
I think he means applications where Linux or a BSD is the primary development environment and those APIs were the target. Agreed it is a sloppy use of "native". In the case of Firefox, XP is a the primary dev environment and it also benefits from Profile Guided Optimizations when compiled. The Linux port could in principle but the work hasn't been done (and probably won't be).
To be fair I think the gp meant Linux exclusive, not native. But even then Firefox is a pretty bad choice, since its development has always had Linux in mind as well as other platforms.
If you benchmark some random 3D games between Linux and Windows there is no Linux slow-down. If you benchmark the responsiveness of a well written GUI environment on Linux vs Windows, there's no slow-down. In fact, I've only rarely run into a situation where Linux is slower than Windows in a GUI or otherwise. The primary reason I've come to realize is lazy programmers writing slow client software, and in some cases, horribly inefficient GUI toolkits (Gtk, I'm looking at you).
X11 isn't the bottleneck, and ever since a few tweaks were done for desktop users, the Linux kernel isn't either.
- Michael T. Babcock (Yes, I blog)
I had the same problem with Google Desktop. It was a great tool and worked well for a while, but eventually its little database file was immense and was dragging my system down with it.
- Michael T. Babcock (Yes, I blog)
> It's been called X-Windows for a long time. ... incorrectly. It's actually the X Window System.
http://en.wikipedia.org/wiki/X11
25% Funny, 25% Insightful, 25% Informative, 25% Troll
MS Windows = Explorer.exe Linux = X11 MacOS = Quartz
A loop, by its nature, continues. If that didn't make sense, start reading this sentence again.
memcpy of 1000s of bytes is slower than sending a message. Many of these systems that provide direct access to RAM require lots of copying too. (OSX one example I'm most familiar with)
X11 also supports direct access to memory, but it is only used in very specific circumstances because it's extra work to set up.
“Common sense is not so common.” — Voltaire
Funny, but the real answer is GDI.
http://en.wikipedia.org/wiki/Graphics_Device_Interface
Buckle your ROFL belt, we're in for some LOLs.
I guess maybe I'm making bad assumptions, as I don't really know what Chrome's intent with multiple processes is... but I think the answer to your question is no, it wouldn't necessarily impact security and certainly wouldn't fully negate the multi-process approach's advantages.
The major advantage to keeping separate tasks in separate processes, it seems to me, is that they have separate memory spaces. I can't sneakily inject code into another task's buffers if I'm not in the same process. In particular, if the browser spawns a process to execute some plug-in or whatever, there's less risk that the plug-in or whatever can trick the browser into executing malicious code. (Or cause it to crash, but that's more "stability" than "security".)
In other words, the biggest security risk comes from two processes sharing the same process at the same time. I don't think re-use is as big a problem. Sure, if you did a bad job of wiping buffers, then in theory one process could see its predicessor's data; and I guess there are scenarios where that could be an issue, though I'm a little skeptical that a malicious process would go rooting around its uninitialized space "just in case" it was handed a process with something it would recognize as sensitive data from a previous task...
The problem is that when monitoring processes, people can "see" X11 using CPU cycles, whereas in Windows they only "see" the application doing. It's six of one, half dozen of the other -- but it makes it look like X11 is CPU-resource intensive. In reality, the same cycles are used for windows based apps (perhaps more? I certainly don't know...), but they look like the app using the CPU which is somehow more expected.
No.
Explorer is a cross between your shell and your window manager, its like the Gnome or KDE window managers except most of the window manager functions are the responsibility of process itself in Windows, although they are provided by the standard libraries, which us things like the uxtheme.dll and company to provide a consistent interface until the app goes well out of its way to do otherwise. You can use cmd.exe in place of explorer and apps won't notice the difference unless they interact with explorer, such as things that put items in the notification area (systray to most). Apps do not talk to explorer to display windows any more than apps on Linux talk to Gnome or KDE, or Finder on OS X. None of them have to be installed or work in order for Windows to be displayed. You can kill all explorer.exe processes in windows and you just won't have a start menu or clickable desktop until it restarts. You can not kill the X11 client and do the same, all processes using it will be disconnected and exit or crash.
Quartz is a toolkit/API used within the 'window_server', like DirectX to some extent on Windows. It is not the generic low level API like GDI is on Windows.
'window_server' would be almost the direct equivalent of X11 on OS X in native applications, if you exclude X11 for OS X, which acts as a translator basically between X11 servers (applications) and your X11 client (the gui you see) and passes that along to the window_server process to display.
In reality, all three of these systems use a different mix of the way these components interact and at which layer things are done due to their different designs. There isn't a 1 to 1 relationship between any of the components.
I do not recall which process on Windows handles the GUI, but it is more or less untouchable, unlike in OS X and traditional UNIX where you can easily kill the gui portion, doing so in Windows traditionally would result in a blue screen, this is no longer strictly true in the Windows 6.x versions (Win2k8/Vista/Win7), but I don't recall what process owns that part of the system off the top of my head.
Persistent Volume manager for Kubernetes - https://github.com/dwimsey/openshift-pvmanager
The thing is, X11 gets network-transparency essentially for free. The system requires some sort of inter-process communication, even when running on a local machine, and Unix-domain sockets are one of the fastest IPC mechanisms available short of shared memory (which is orders of magnitude harder to get right). X11 supports shared memory for local processes, to speed up communication of large objects like pixmaps, but the core protocol is socket-based and there is absolutely no reason to change that. So long as this is the case, why not support network transparency? BSD network sockets are interchangeable with Unix-domain local sockets once the connection has been established, so it's not like there's much more effort involved.
"The state is that great fiction by which everyone tries to live at the expense of everyone else." - Bastiat
The Windows and OSX abstractions for the display don't provide an API that allows these sorts of optimizations to be done behind the scenes.
That is not true. Windows GDI was designed for hardware acceleration. As an example: Circa 1998 I got an ISA Diamond Viper video card which performed orders of magnitude faster than comparable VESA cards because the drivers took advantage of the hardware rasterization. For example, dragging a window didn't redraw the window, it moved the bitmap from one place in video memory to another. Drawing/filling lines and shapes was absurdly fast because GDI offers primitives for those operations and the driver mapped those to the hardware functionality.
Those same things happen today under Windows Vista, Windows 7, and OS X. That is a big part of why the driver model changed for Windows Vista/7: Microsoft wanted to expose even more layers through the video driver to permit those kinds of optimizations. X11 would have a hard time trying to do the Windows 7 alt-tab or OS X expose features where Windows move around in 3-dimensions on the screen. X11 doesn't expose that kind of stuff.
VNC and RDP give you the ability to interact in an explicitly remote sense, the windows in question operate on the remote server instance and will remain in existence regardless of what happens to the local system.
That's one reason I stopped using X11 forwarding even though I could: If I lost connection on my laptop for any reason, every application I had open was dead. With VNC (or RDP), they were always running remotely.
Also, if I have an application open on display :0 I have no way (that I know) of moving it from :0 to :10 and having it appear uninterrupted on my local display.
I'd say that they're extremely useful hacks that solve issues that are, at least for me, unresolved in X11.
It certainly works much faster than Firefox on Linux in most circumstances. One thing I have noticed, though, is that Flash applications seem to run much more slowly in Chrome than in Firefox (to the point where Chrome becomes unusable on my computer for some sites that run acceptably in Firefox.) That seems a little weird to me, since it's the same plugin. Hopefully it's a problem that will be corrected before the official Linux release.
That and the lack of a decent ad blocker are the main things that keep me from switching from Firefox. I really like the speed and the interface is generally a big step forward, but I don't really like having to switch between two browsers.
How can we continue to believe in a just universe and freedom to eat crackers if we have no ale?
I don't care what Wikipedia labels it. I've been around since back in the day, and nobody went around saying "The X Windows System". We called it "X Windows". Maybe we weren't being 100% correct, but I'm pretty sure it was acceptable enough that nobody would have quoted us with a snide little "[sic]".
Why are you referencing the average PC of 20 years ago, which ran Windows 3.1, not X11, to the average workstation from back then, which the GP referred to as 'woefully inadequate' today? Did you know that there were actual framebuffer cards back then in typical use that cost more than your '386-DX 25 with 16MB of RAM' that all the secretaries were impressed by?
The point is that while X is "an afterthought" and "badly designed" it has been fixed so well and so often by so many people (simply because there was nothing better) that it has already had all the major bottlenecks taken out of it ...
It is an afterthought in that it is stacked on top of the rest of the system rather than integrated, and it has a client server model because originally the display was normally on a different machine than the server (this is often raised as a bad design), but this can actually help to isolate it from the rest of the system so it is not dependent on it, and because it was always (on very old hardware) considered very slow, it was optimised to be as fast as possible within the design limitations it had .... this means that it is now very fast on fast hardware ....
Puteulanus fenestra mortis
Windows does not have a bevy of toolkits to choose from out of the box.
Yet Windows developers often find the core Win32 API doesn't fit their needs, and waffle between toolkits like GTK, wxWidgets, Qt, and the usual cross-platform suspects. They also turn to .NET, Java (technically a whole language), MFC and others.
I constantly run into issues where I need to download a runtime to run a Windows app, because many developers don't simply develop for what is out of the box with a standard Windows install.
I've seen native X11 apps offer up checkboxes, form elements, and the like. So it is possible. It just doesn't look nice.
Developing audio on Windows is not necessarily standard, not easy. That is why apps might ship with their own separate audio library like OpenAL, tie into QuickTime, etc.
I have some non-trivial Windows apps that were built for Win95 that still run in Windows7.
Sadly, there were many Windows 95 era apps that were still 16 bit, which don't work in x64 versions of Windows 7. Half of my games (mostly from the XP era, not Win95 era) wouldn't work properly in Windows 7. Windows breaks compatibility all the time, even with service packs. Saying you have a particular app that still works doesn't mean the Win32 API has stayed the same, because it hasn't.
The GTK API has stayed mostly the same over the 2.x life-cycle. That is likely to change with GTK 3.
Now, I'm not a fan of GTK by any means. However, the GTK 2.x API has been around since 2002. Anything written for that API should still work fine today.
If they wanted the easy way out, they would have used wxWidgets, which at least feels native (by using the native toolkits) across Windows and Mac, and GTK on other Unixes.
That is basically what I said, except you apparently didn't get that. I suggested they use Qt (like wxWidgets is a cross-platform toolkit). Qt makes more sense for a variety of reasons, such as native bindings to Webkit (the heart of Chrome), and very easy audio development where you write code once, and then the audio works on Windows, Mac, Linux, etc.
http://blindscribblings.com - Tasty pop-culture in conceptual fashion.
I've never had performance issues running X11 over a LAN.
VNC and RDP are useful hacks for systems that weren't designed for remote access, but they're no replacement for real network transparency.
Oh no you don't!
Try using X11 over something slightly slower as LAN. Just try it, over ADSL, whatever
I tried. And X11 is totally and utterly USELESS. A well configured VNC (and you have to really play with the knobs) is usable. RDP is the best (of course, it wasn't developed by Microsoft...)
how long until
Let's just say you're wrong and I've seen flickering on plenty of Mac OS desktops.
And with X11, the flickering you get is more likely due to the program ignoring X backing store and "doing their own thang". Well guess: their failure isn't the fault of X11, is it.
"To be fair to X, managing compositing et al isn't it's job---but it should be!"
Compiz: It IS!!!
Sheesh.
And Enlightenment had compositing freaking YEARS ago.
" Between X's by-design paucity of features"
You mean like C's "paucity of features" that means libraries that do whatever you damn well want?
There is no "by-design paucity": by design X11 is extensible. See X extensions.
"Had X been designed a little more smartly (eg, for actual people and not for computer scientists) this probably wouldn't be such a problem."
Uh, what design WOULD have been for "actual people"? This statement, bald as it is, makes no sense.
X11 is designed for the task it has to solve: drawing a GUI.
One program: one purpose. Expose capability and don't impose process: someone may think of a use you never considered when writing it, so don' t write a program that will hate them for it.
The UNIX way.
Which, oddly enough, Apple have embraced to a large extent since bringing out Ten.
"By comparison, again, we have MacOS X's system, which again just works, even if in theoretical terms it's a little slower."
Two problems: the dissing of X is how slow it is. So Ten's system being slower should be more dissed, yes?
Secondly, ten's system doesn't just works else there would be no problem with "But Mac can't support clones, they have to have a limited selected hardware to deliver the eXPerience!". Ignoring that this just works meme is wrong. I've seen it often just stop a lot.
"The UNIX Hater's Handbook, which is a little bit out of date now, goes into the design errors of X "
And here we see where you've been misled.
The UNIX paradigm is extensibility. Policy is set by the use of the program. not by its programming. And the UNIX haters hate UNIX so hate the UNIX paradigm. Ergo they hate X too.
Maybe they're just a little bit predisposed to a priori conclusions...
And it's not the "why does X drive people nuts" it's why do people get a stiffy when the opportunity comes to diss X?
(oh, and a quick look at that, hmm, *discourse* seems to be a person who gets a real big boner over getting to rant and rave about how X is teh devil. Could he be any less coherent?)
No X11 can do Windows 7 and Vista and OSX expose features ... and does so .....
The whole point is that X11 does not draw Windows it draws tiles ... Window Managers draw windows ... and they can draw 3d glass dancing Windows on X11 without X11 caring about it ...
On Windows the layers are Driver - GDI - Application ...)
On X11 the layers are Driver - Kernel - X11 - Window Manager - Application (there may be more
The point is that you do not need to Expose the low level stuff to the application. .. just to the window manager, the application should not have to worry about redrawing itself, or resizing the window etc... it should let the window manager worry about that
Puteulanus fenestra mortis
Uhhhh... have you seen Compiz, Beryl, Metisse, Gnome-Shell or any of the other whiz-bang screen-flipping and warping and cubing desktops? They do run X11 apps... through an X11 extension, be it AIGLX or XGL or something similar. X11 exposes whatever you want through the use of extensions, including the stuff needed to do 3D window manipulation. It did this when Vista was still Longhorn, let alone Windows come lately annex 7...
--frank[at]unternet.org
(and "client" and "server" roles were reversed from convention, for some strange reason)
What? Since when? I always thought that the X server was the behemoth application that ran, waiting for connections from other apps (the clients), consolidated the requests, acted on them, and responded back to those clients. You're telling me that X itself is termed the client, and those little apps that all connect to it are called servers? Yeah, that IS backwards!
Oh, hold it, that's not what you're saying at all. You think that "server" is a designation of the size of the machine it runs on, not a designation of the model of communication the application itself uses. You do realise, though, that even a "web server" (which could just be a wall wart acts as a client for DNS querying, right? That "client" and "server" are fluid terms based on what the app is doing, and not where it is?
A server responds to incoming request(s), usually from multiple sources. A client initiates those request(s), usually to a single target. That is all. X uses these terms perfectly. The application sitting on my desktop machine is the server, and the xterm I'm running on the Linux/zSeries box is the client. For this particular purpose. Of course, that Linux/zSeries box is also the ssh server that I use to connect to it in the first place, over which a tunnel is created in the reverse direction to allow that xterm to come up at all. It's not the convention that is being ignored. It's just that you're using the wrong definition.
Konqueror doesn't even run on Webkit.
On my machine here, in Konqueror, you can go to the View Menu, select "View Mode", then click on "WebKit", to change Konqueror's renderer. So yes, it does. It just doesn't do so by default.
I'm dragging a Firefox window over a dead RDP session right now. It's erasing the contents of the window. I've never, ever managed to do that on MacOS X and not often on Windows Vista.
If I use a compositing window manager, that problem goes away, but in it's place comes a whole new set of issues. Ever notice how common "turn off Compiz/KWin compositing" comes up when people try to troubleshoot X issues? At least Vista is smart enough to turn it off for you when the situation merits it.
On no other platform do you ever have to think about syncing refresh rates. On X, you're at the mercy of the driver and window manager. On an nVidia card you're probably ok. On a recent Intel card you might be. On ATI it's a complete crapshoot. Throw in compositing and it gets even more complex.
Meanwhile, on much of the same hardware, you could run a hacked version of Mac OS X and not see a lick of tearing or artifacting.
I'm not saying that IRIX didn't do have that problem, but that, at the time, you could buy a commercial UNIX workstation and get a decently-integrated X server. The problem then was that you had to pay an astronomical sum to get the same window management performance that you took for granted on a Mac, and that a heck of a lot of tuning, testing and integration had to happen to get your (very expensive) video or 3D application to work, again, as well as a Mac.
Nowadays, you don't have to spend a fortune tog get decent X. What you have instead is that you're stuck with one card family (nVidia) or checking experimental code out of various git respositories, and even then you're not guaranteed the level of seamless video behaviour that you'd see on a Intel-based machine running Windows Vista.
I'm glad some people get decent X performance, but spending more than a little time on, say, Phoronix or Ubuntu's forums should disabuse you of the notion that everything works.
--srj/mmv
- As to overwriting. This occurs because the update events follow behind the UI. The problem is resolved by the composite extension, or by enabling backing store -OR- by increasing network bandwidth. Some old X Servers didn't have backing store (and certainly no compositing), AND ran over constrained pipes. It hasn't been a problem with desktop X for years.
- X is extensible by design. Multiple display support, accelerated 3D, video playback and compositing do work. For $DEITY sake, I use these features on my stinky little Acer Aspire One using Linpus! No particular problems -- "it just works" (tm). I don't like transparent windows, so I just don't bother, but it does work. Why the hell would a user want to know about the alphabet soup? Just use a packaged OS. The alphabet soup comes about because the development of X is an open process.
- And, in comparison with the Mac, you do notice that Apple packages an X Server with OS X? When running in a heterogeneous environment, it's necessary.
- Finally, you bring up the Unix Hater's Handbook. Ok, let's break it down:
1 - xload, xterm and xclock are possibly among the LEAST used programs under modern X based systems. They weren't
even installed on my Acer when I got it.
2 - Motif isn't used anymore.
3 - Cut and Paste really isn't an issue anymore, either.
4 - ssh -Y is usually used to remote X servers - authentication isn't an issue anymore either.
5 - Gnome and KDE provide the "customization methods"; since xterm isn't used anymore (or xcalc, or xedit),
the xresources issues are also gone.
6 - imake has been deprecated for YEARS.
7 - Pretty much nobody uses raw X protocol or XLib anymore either.
8 - NeWS was "killed" because IBM and DEC didn't want a repeat of NFS - they didn't want to send SUN any more money. So, they marketroids forced the issue. I agree the superior technology didn't win, but X is still around. Sucks to be the customer when they get what they have been told to ask for.
The UGH was relevant in the early '90s. No longer.
The "MAC UI Experience" could be planted on top of X. I am disappointed that Apple isn't driving that. It would involve developing several extensions that would be useful to X users. But, if Apple doesn't want to do it, others will:
http://www.freedesktop.org/wiki/Software/CompositeExt
http://en.wikipedia.org/wiki/XRender
http://keithp.com/~keithp/talks/randr/randr/
http://en.wikipedia.org/wiki/X_video_extension
http://docs.sun.com/app/docs/doc/801-6662/6i1196cd6?l=ja&a=view
The first four are generally implemented. The last is not (X/DPS). But, MAC OS X only implements a subset of X/DPS anyway (and, of course, it isn't compatible).
Just another "Cubible(sic) Joe" 2 17 3061
Don't use X11 raw over remote links. Use NX. It meets your criteria from start to finish and actually brings to X11 the things you talk to with actual performance, when compared to the others.
Comment removed based on user account deletion
Us losers? Are you implying that somehow you are exempt from being a loser while still posting to slashdot just because you are anonymous?
An anonymous coward who knows what he or she is talking about is far more valuable than a poster with an account who doesn't.
The closest thing to screen for X11 that I know about: xpra. A bit rough around the edges, but usable.
On MacOS X, it's just about impossible to get into a situation where a) video tears or flickers, or b) menus and windows can "rub out" other menus or windows (eg, you can't drag a window around like a giant eraser on Mac OS). On X+whatever, it's pathetically easy to do either. Windows is somewhere in-between the two.
a) That is largely an issue of the video driver in use. Given the poor state of documentation on most video hardware these days it isn't difficult to understand why this might be happening. Back in the SVGA era you didn't see this happening because there was enough documentation of a standard interface to do things right. Some of this is server related as well. X was never designed to support opaque window moves. That has been grafted on in time.
b) That is something that afflicts the X.org server. You don't see these sort of compositing flaws in commercial X servers. This is not a fundamental problem with X. Just a common implementation of the server.
I am becoming gerund, destroyer of verbs.
If VNC is usable, you'll love NX. It is *far* more responsive for a given bandwidth/latency and it is persistent too (the session keeps running if your client disconnects.).
You can even run VNC to other machines through NX and it feels faster on limited bandwidth (NX creates a session on the Linux client that runs a fullscreen vncviewer to another system.)
It's my standard way of working remotely. My default work desktop lives on a Linux machine at the office and it resizes automatically depending on what screen size the client uses (as long as your Gnome or KDE version is recent.) Even at the office I run NX to my work session - over a LAN I can't tell the difference between local and NX.
thegodmovie.com - watch it
back when that actually was an intended purpose
Lots of people still use X in its intended purpose. Many people in a corporate or academic environment who need to run X apps nowadays will run them on a server or compute cluster and have the display shown on their PC (Typically running Windows with Exceed or other server). Don't poo-poo that design feature just because you haven't made use of it. Everybody's hot to jump on the Citrix and VNC bandwagon these days to do what could be done with X 20 years ago.
The performance issues you see from running remote apps largely stem from the abundance of eyecandy in modern window managers and applications that have to throw a lot of pixmaps across the network to get the job done. Before this was commonplace, remote apps worked as smoothly as a local one. Actually remote apps would be even faster in the days when typical X usage involved a low power, low memory workstation as the server. That model is what X was designed for. Most workstations would be brought to their knees with swapping if you tried to run a non-trivial application locally. It isn't fair to criticize an architecture developed in the 80's that was flexible enough to still be in use today even if some design decisions would be different in a clean architecture started now.
I am becoming gerund, destroyer of verbs.
I'm dragging a Firefox window over a dead RDP session right now. It's erasing the contents of the window. I've never, ever managed to do that on MacOS X and not often on Windows Vista.
For RDP that makes sense- compositing doesn't happen over Terminal Services. What normally happens on Windows, is that when the Window Manager (not Aero) detects that an application has stopped responding to WM_PAINT messages, it actually swaps out the window for a special one that looks just like the last known good copy of the windows display surface. If it can't for some reason, it just dumps a plain white window with loading cursor instead.
That's actually intentional.
For a site about things like basic rights, Slashdot users sure do like to censor "dissent".