The Challenge In Delivering Open-Source GPU Drivers

[mugginz]

Surely, it's possible to update Mesa, X, and the kernel in a non-rolling distro, as long as it provides the relevant packages.

I guess if updating your kernel so that you can upgrade a graphics driver never say broke your wifi then I guess that'd be quite good.

[pingufunkybeat]

If it does then it's not a non-rolling distro, no?

No, not really.

You still need to do a complete re-install from time to time, because the undelying libraries tend to be stable and changing them requires a recompilation of everything that links against them.

Upgrading one or two packages is easy. Upgrading everything, starting with libc, is a lot more difficult, which is why distros prefer to package everything from scratch twice a year. The rolling distros a la Gentoo get to upgrade everything all the time and occasionally deal with the related fallout.

[jari2]

GNU/Linux on the Desktop: It is happening

I wanted to point out the success of GNU/Linux on the desktop despite the critics. In Oregon a number of companies and organizations have emerged to support desktop GNU/Linux. ThinkPenguin is selling fully support desktop GNU/Linux systems and accessories. You are going to see more people buying them who are average users.

People don't seem to realize that the average user has 10GB of data, checks his or her email and posts a few pictures on the web.

Most wouldn't even notice if 3D accelerated graphics were missing. About the only real benefit that 3d accelerated graphics give you is better battery life on laptops and the ability to watch video where cheap proccessors are used.

Desktop GNU/Linux systems are here. The Penguin Wee sold by ThinkPenguin is doing very well. It sells cheap and runs better than any comparable Microsoft or Mac system for a fraction of the cost. The cheapest decent GNU/Linux desktop system (the Penguin Wee) is about $250 with an Intel Atom Processor and 3d accelerated graphics vs $500 for the cheapest Microsoft Windows desktop computer at a typical store like Staples or a $1000 for a Mac. Used computers might sell for less, but not new ones that are comparable in terms of speed performance, and meet customers needs.

Assuming that the drivers are free. Binary blobs cause problems and it isn't just in GNU/Linux. Microsoft's ecosystem is full of dead printers, graphics cards, and other devices that no longer work. In fact many of those devices do work in GNU/Linux still because unlike in Microsoft Windows the drivers are free. Users aren't demanding the latest and greatest hardware. They are demanding systems that work.

Apple has marketed that image. Sadly they don't actually produce much of anything that works and users end up in the same cycle that is Microsoft Windows. Purchasing new hardware every other week shouldn't be a requirement. In Apple and Microsoft land it is.

These are the reasons GNU/Linux is succeeding on the desktop in the places where it is being heavily marketed.

The year of the GNU/Linux desktop is really just a perspectives game. No major company has succeeded yet small players are succeeding and making allot of $$ off it. Corporations like HP and Dell may eventually become disadvantaged by smaller players who are currently succeeding in the market with Desktop GNU/Linux.

Don't forget that there is also System76, Open-PC, and others who while may not have been successful at penetrating the masses have been selling systems and doing well with a technical audience.

The Open-PC project did an excellent job designing even if it can't fulfill the masses needs. If nobody is buying though I question the success of the project. I think that may have partly to do with a failure to market it though. No press release was ever done. Or it was done many many months before the actual release.

[deanjo]

It sells cheap and runs better than any comparable Microsoft or Mac system for a fraction of the cost. The cheapest decent GNU/Linux desktop system (the Penguin Wee) is about $250 with an Intel Atom Processor and 3d accelerated graphics vs $500 for the cheapest Microsoft Windows desktop computer at a typical store like Staples or a $1000 for a Mac. Used computers might sell for less, but not new ones that are comparable in terms of speed performance, and meet customers needs.

Wow are you ever reaching there. The penguin wee is nothing but a base atom system. It doesn't give superior performance in linux, quite the opposite. Video drivers alone hamper the system where intel's windows blobs blow by the FOSS intel drivers. Also Apple has never made a system with that low of system specs (intel era). The very first mac mini with intel smokes any atom based system to date in performance, (even the original ones with the intel graphics IGP). For the same price of the Penguin Wee you can go out and many of the dirt cheap atom systems out there with windows included (which are usually dual core atoms to boot).

[allquixotic]

People don't seem to realize that the average user has 10GB of data, checks his or her email and posts a few pictures on the web.

Most wouldn't even notice if 3D accelerated graphics were missing. About the only real benefit that 3d accelerated graphics give you is better battery life on laptops and the ability to watch video where cheap proccessors are used.

This may be true for a certain narrowly-defined "average user", matching a specific demographic. I could take a wild guess at what that demographic might be, in the context of the rest of your post. Just for the sake of argument, let's say that you draw your conclusions about what the "average user" wants from personal experience, surveys, and indeed, in-depth statistics about what suburban, middle-aged, working-class users in Oregon expect out of a home personal computer.

Given these demographics, it is perfectly reasonable to expect that these users would not get very excited about their new computer, if it is functionally equivalent to their old one, but has new "whizzy" desktop effects, flashy new hardware-accelerated 3d games, and counterintuitive new desktop features like displaying all your windows on a cube, or as a mosaic. In fact, they might even have a negative reaction to these new features complicating a system that they already understand, and they would resist the change just for the sake of remaining in the familiar. I'd even go so far as to say that they probably want to continue using whichever computer operating system they are most accustomed to today, and will only change when they are absolutely forced to by hardware failure with a lack of replacement parts.

Now, let me paint a completely different picture of a different group of citizens, with completely different requirements. Take for example, young adults in their mid to late 20s, 30s, and early 40s. Middle-class working folks (some unemployed due to the economy) living all over the country. If they used computers when they were very young, they probably grew up on DOS or Windows 3.1. If they didn't start until they were in their teens, they might only know Windows XP, or Mac OS X.

But they have social reasons to be open-minded about change. Word of mouth is an extremely powerful marketing force, and everyone in this demographic cares about what's "cool" and new. So when something visually attractive like Android or iPhone or Windows Aero hits the market, this demographic is enthused, and the topic spreads like wildfire through social networks.

Maybe in 2001, this demographic would have matched the expectations of the one you have defined. But this is 2011. Everyone and their grandmother has a smartphone with "whizzy" hardware-accelerated window transitions, accelerated video and Flash, and games whose graphics detail is quite frankly remarkable considering the size of the device.

And then there's the desktop competition. Windows 7 and OS X are downright beautiful in their presentation and in the desktop accessories (like the Aero feature that lets you see all your windows zoomed out in a panel-like layout). Nearly every user interaction except typing text sets off an elaborate, detailed animation or transition to show the user that an operation is in progress, and to do so with fashion. Aesthetics matter.

In the 2010s, I'd argue that more and more users are starting to value what we deemed "eye candy" in the 2000s and earlier. Animations, transitions, dynamic / haptic feedback, advanced scene graphs.... none of this is really possible without 3d acceleration, unless you have a fairly powerful CPU -- and then the CPU is inefficient for this kind of graphics-intensive workload.

But it depends upon who your target market is in particular. There is no rigorous definition for "average user", only perspectives. Demographics, generalizations, surveys, statistics, personal experiences... it all comes into play when you are trying to capture and define what the average user wants.

You might think that it doesn't even matter what the average user wants, and that as long as you can market your offering to some substantial sub-market of the world, then you are successful as a business. That's fine, but it assumes the microcosm perspective. When analyzing a global movement like FOSS, it is more sensible to take the macrocosm perspective. The little successes found here and there tend not to affect the status quo of the companies and users who operate in the "mass market" segment, where dollar values are discussed in billions and user statistics are aggregated in tens of millions.

But when I say "average user", what I refer to would more precisely be defined as "The overwhelming majority of non-technical business and home computer users living in the industrialized nations of the world". Gathering conclusive factual data about the values of this large and diverse group of people is difficult if not impossible, so I won't claim I have the facts. I can just tell you that I think I have a fairly good pulse on what those within my own demographic care about, and from that I tell you that computers bereft of 3d acceleration would simply not be acceptable to those users. In fact, many of these users are picky about the extent of the 3d acceleration: gamers, a moderately small sub-class of the demographic defined above, care about whether the 3d acceleration is fast enough for animations and transitions to be "seamless" versus "jittery". They also care if the graphics stack is so poorly maintained as to fail to work properly with the applications they want to use (mostly games). In the case of most open source 3d drivers, the performance is still very, very poor for even the most basic games, and they still fail to run the vast majority of both native and emulated hardware-accelerated games.

So, conclusion? Go ahead and celebrate the victory of GNU/Linux systems in some limited demographic in Oregon, but keep in mind that on the larger scale, when factoring in especially the younger generation, the continued success of GNU/Linux is directly related to its ability to produce awe-inspiring, slick, hardware-accelerated graphics. It is becoming almost as essential as speedy internet access.

[popper]

No, this is our job, and we blew it for Sandy Bridge. We're supposed to do development well ahead of product release, and make sure distros include the necessary code to get things working (we have separate teams to do development and aid distros in backporting, though most of them can handle it by themselves these days).

I could give you all sorts of explanations as to why this is (Sandy Bridge is a big architectural change, we made some mistakes in defining our development milestones, and we didn't work hard enough to get our changes backported), but really there's no excuse. Fortunately we've learned from this and are giving ourselves more time and planning better for Sandy Bridge's successor, Ivy Bridge.

As for a stable ABI, yes it would definitely help situations like this and make lives easier for distros, device manufacturers, and probably users. But it would make life harder for developers, and as we all know, open source development is driven by developers, and we're a whiny and lazy bunch. (Yes, this is a flippant remark, don't take it too seriously since I omit much of the complexity behind the "life harder for developers" part that has big implications for users, distros, and device manufacturers; it's a complicated issue.)

never mind Jesse, these things happen i guess.

if you/other intel dev's ever get bored one day or need to clear your Linux head for a while , you might like to take a look at the AROS OSS code http://aros.sourceforge.net/ and port some Intel related gfx to their hidd drivers etc, you Did own and have some FUN writing code for the amiga OS once right

[popper]

"Mesa updates come every quarter too and are generally released by Intel's own Ian Romanick."

BTW Michael, you have probably already seen it, but Ian said he was readying a release of 7.9.1 and 7.10 for friday

"I plan to release 7.9.1 and 7.10 on Friday
afternoon (pacific time). I'll ping folks on IRC first, just to be sure."

[Lennie]

Browsers

So, conclusion? Go ahead and celebrate the victory of GNU/Linux systems in some limited demographic in Oregon, but keep in mind that on the larger scale, when factoring in especially the younger generation, the continued success of GNU/Linux is directly related to its ability to produce awe-inspiring, slick, hardware-accelerated graphics. It is becoming almost as essential as speedy internet access.

I would just like to add that, I don't know if you folks noticed but the browser vendors are adding hardware acceleration to their tricks. No plugins, but real hardware acceleration. Have a look at the efforts by Mozilla Firefox 4, Google Chrome and IE9. I'm not just talking about OpenGL-like WebGL, I'm especially talking about faster rendering of normal pages.

And example article:

http://www.webmonkey.com/2010/09/a-guide-to-hardware-acceleration-in-modern-browsers/

[BlackStar]

No, it's not a sad excuse. It's an excellent excuse. Using a computer should be easy and intuitive. Like Mac OS and Windows.

Expecting users to deal with Git and make is ridiculous.

I was responding to the comment about "distro dependency". You are off-topic.

[stoatwblr]

Not just Sandy Bridge

No, this is our job, and we blew it for Sandy Bridge.

Not just Sandy Bridge. This just brought ongoing issues to a head.

I've been frustrated enough to recommend to our procurement teams that Intel-based hardware be avoided entirely until things were sorted out.

Your own Intel Desktop Boards have been mildly unstable for years under linux. It's relatively easy to provoke them to Segfault X sessions and in some cases completely lock up the system (especially the older 8xx and 9xx series boards such as the DQ965GM but I can enumerate several other examples)

I work at a major UK university and we can't afford this kind of problem. Linux gets the blame when it's clearly a hardware/firmware fault and that plays right into the hands of MS's FUD generators.

Having your devs state in irc sessions that they don't care about a X stability problem triggered by Firefox because the product is unsupported is not encouraging for a board less than a year old, etc - in all liklihood if you go back and address the issues with your older hardware it's likely you'll cover a lot of problems with the newer ones too.

What's very clear is that a culture change is needed at Intel. It's intolerable that products are being certified as working with linux (and specific distros such as redhat) when they clearly don't work properly in the real world. Perhaps less reliance on software Development Vehicles and some more hands on experienxce with real world hardware is a good starting point.

I'm trying hard not to flame, but Intel really needs a wake up call.