Linked by Hadrien Grasland on Thu 13th Jan 2011 17:13 UTC, submitted by Michael
Linux "Now that the kernel mode-setting page-flipping for the ATI Radeon DRM kernel module has been merged into the Linux 2.6.38 kernel and the respective bits have been set in the xf86-video-ati DDX, we're in the process of running new open-source ATI graphics benchmarks under Linux. Our initial results (included in this article) show these latest improvements to cause some major performance boosts for the open-source ATI driver as it nears the level of performance of the proprietary Catalyst driver."
Thread beginning with comment 457898
To view parent comment, click here.
To read all comments associated with this story, please click here.
RE[7]: great
by lucas_maximus on Fri 14th Jan 2011 13:41 UTC in reply to "RE[6]: great"
lucas_maximus
Member since:
2009-08-18

Stable in-kernel AP/BI isn't gonna happen in the mainline Linux kernel, ever. It's logistically infeasible and most kernel developers would swear by the ability to change APIs as development progresses.


Constantly having to change an interface reeks of poor software design.

The whole point of an interface is that it stays the same and the implementation behind it changes.

What's possible and actually being done is maintaining stable ABI in the same distro release and for enterprise products this often can span many years. In that respect, it isn't too different from commercial operating systems.


Enterprise != My Desktop/Laptop.

If for whatever reason stable ABI is a must for you (it escapes me why it would even matter for usual desktop users tho), go ahead and use something else.


So according to your logic it is acceptable for a desktop machines for hardware to stop working after an update (which does happen).

A driver released 10 years ago for Windows XP will still work with Windows XP Service Pack 3 with all the latest updated ... In fact I am using Windows 2000 drivers on my old laptop because there are no Windows XP drivers for it ... The Interface has stayed the same therefore the older code still works.

Quit whining about something which isn't feasible.


It is feasible. The Kernel maintainers have the power to do this anytime they want.

Anyway none of this changed my original point. If there was a Stable ABI, less effort would have to spent on hardware support whether that code was open or closed since code doesn't have to be constantly changed.

Reply Parent Score: 2

RE[8]: great
by asdf on Fri 14th Jan 2011 14:07 in reply to "RE[7]: great"
asdf Member since:
2009-09-23

Constantly having to change an interface reeks of poor software design.


It's something called evolution and an extremely fast paced one at that. In the end what really matters is scaling the scalability of development and that's the biggest reason why Linux got where it is today.

A driver released 10 years ago for Windows XP will still work with Windows XP Service Pack 3 with all the latest updated


The same goes for RHEL and SLE SPs (for the most part). If you talk to MS kernel engineers that they have to maintain internal ABI identical across different kernel generations, they'll probably give you a pretty silly look too. And, you know what? They don't either.

The difference doesn't primarily come from development itself. It comes from how they're packaged and released. Linux distros are much faster paced, which has its benefits and drawbacks. A lot of that is by choice but at the same time with the current money flow (at least for desktop), it's quite difficult for distros to sustain such long maintenance cycles. It takes a lot of money to do that and that's why you see much longer cycles with enterprise distros.

Really, it's not about kernel devs trying to screw everyone else. If it were that simple, don't you think someone would already have come up with a branch or something which maintains the supposedly superiorly designed stable ABI? It's about how the whole thing is structured and the economy around it is built and I personally think that it may be different from other but nevertheless a model which has potential for sustainable improvement over long period of time.

So, think a bit more about it. It's okay to complain but don't draw conclusions when your understanding is very shallow. Just say hardwares working in one release and not in the next is very annoying or unacceptable. Don't jump to the unwarranted conclusion that that's a result of kernel developers' whims.

Reply Parent Score: 1

RE[8]: great
by sj87 on Fri 14th Jan 2011 14:28 in reply to "RE[7]: great"
sj87 Member since:
2007-12-16

So according to your logic it is acceptable for a desktop machines for hardware to stop working after an update (which does happen).

Your own fault, if you upgrade packages that break compatibility.

A driver released 10 years ago for Windows XP will still work with Windows XP Service Pack 3 with all the latest updated

And Linux 2.6.35.10 is 100-% compatible with the ABI/API of Linux 2.6.35.1.

If you're gonna compare Windows XP updates with each other, then you also must compare the same kernel revision updates with each other. Otherwise it's meaningless.

It's not like "Linux 2" or "Linux 2.6" were major versions that have to keep compatibility throughout their life spans. "Linux 2.6.36" and "Linux 2.6.37" are major versions, really.

If you have hardware that's considered 'unstable', then you shouldn't try to act like it was something better. It's your own fault in the end.

In fact I am using Windows 2000 drivers on my old laptop because there are no Windows XP drivers for it ... The Interface has stayed the same therefore the older code still works.

Only the legacy interface has stayed the same. Microsoft is constantly upgrading Windows' interfaces, too.

Microsoft supports the old/legacy interfaces only because they have no choice. They aren't supporting them because the interfaces are so god damn great and never become technically out-of-date.

MS also hides the incompatibility by delaying public releases of Service Packs and such. It gives time for the manufacturers to check their stuff against the new version and fix regressions and other breakage in time for the masses.

Linux kernels are there for everyone to test any time, it's just not many third parties give a crap whether their products work all the time or not. Probably partly due to the fact that a Linux release happens four times a year whereas Microsoft rarely releases any real updates aside bug fixes.

Edited 2011-01-14 14:37 UTC

Reply Parent Score: 2