xorg-server 1.6.0 in testing

erm67 · 2009-03-14 14:18:36

Demon wrote:

And guess what? I've disabled composite and I'm using EXA instead of XAA now, and after a couple of hours X uses around 30 MiB! It started at 25 MiB. That's a difference, considering it was using 50 MiB at start before...
Edit: Yes, I'm using 6.12 too.

Well ok I always thought that compiz and KDE special effects were a waste of resources, without composite they do not work.

Slikkie · 2009-03-14 17:18:41

Lazer wrote:

(...) btw, and a little OT, how do you manage having just a few packages from the testing repo? I didn't want to enable the repository (I have no problems with other packages) so I downloaded the xorg packages manually, but it's a pain to update them. Is it not possible to tell pacman to use testing just for some packages?

I do pacman -Suy to get a list of anything that can be upgraded, then I answer the "Proceed with installation" question with no, and install the packages in the list that I like with pacman -S

MilesRdz · 2009-03-14 19:45:16

Does anyone know when xf86-video-ati 6.12.0 will hit the testing repository (guess not)?

Last edited by MilesRdz (2009-03-15 14:50:50)

AlphaWolf · 2009-03-14 22:06:34

blasse wrote:

SoftVision wrote:
AlphaWolf wrote:
I have some Questions about the new Intel Driver which comes with 1.6 and about the 2.6.29:
1) When 2.6.29 is stable and in the repos: Do I have to pass something like vga=xxx to the kernel to get my desired framebuffer resolution or will it be automatically? The best I can get without some ugly hacks (915, fbvesa etc...) is 1024x768 while my X11 resolution is 1440x900.
I'm on Fedora 10 at the moment at I had to pass vga=792 to be able to use the Plymouth boot loader in Fedora. I have a resolution of 1440x900 but I wasn't able to get it (though that could be fixed in newer kernels because I'm on 2.6.27 and Intel support was said to be experimental at the time).
2.6.27 doesn't have KMS support, just as 2.6.28. The question is to we need vga=xxx argument with 2.6.29?

I tested it today. You do not need the argument.

Dr4go · 2009-03-15 02:52:32

hoschi wrote:

Here is my .config:
Important: Don't compile any framebuffer-driver in the kernel, but framebuffer-support itself
[...]

Just curious and for the 2.6.29 kernel release interesting: Why is that important? Is it only important for MacBooks or also for other intel based kms capable hardware?

Thanks in advance.

psych · 2009-03-15 08:09:03

It seems to be important for all intel hardware... if you want to use KMS !

Kms does not run at my Dell Laptop (X4500) with any framebuffer driver in kernel.

erm67 · 2009-03-15 08:28:37

It is maybe possible that others with ATI chipsets try if it possible to drag a window over an opengl window (glxgears) with composite enabled? I am just curious to know if the problem is only with r200-r300 chips or composite is broken for all ATI chips.

bender02 · 2009-03-15 08:44:52

Dr4go wrote:

Just curious and for the 2.6.29 kernel release interesting: Why is that important? Is it only important for MacBooks or also for other intel based kms capable hardware?

Yes, it's for any intel hardware. However, I tried it on my GM965 (X3100) with 2.6.29-rc7, and the console support wasn't so good - the last line was screwed up and redrawing the screen was tearing. So I switched back to vesafb, I guess I'll need to wait some more until the i915 driver has matured some more.

hoschi · 2009-03-15 20:44:41

psych wrote:

It seems to be important for all intel hardware... if you want to use KMS !
Kms does not run at my Dell Laptop (X4500) with any framebuffer driver in kernel.

The reason is simple, with KMS their is only one driver for Framebuffer and X11. If you enable KMS you can read a small warning about that in the description of the kernel.

<edit />
Seems that I can really feel lucky with my MacBook.

x86_64 Testing xf86-video-intel 2.6.99.902-1
Any experience with 2.7 series? I'am afraid about problems with KMS, on the other hand the new driver maybe increase the performance under X11?

Last edited by hoschi (2009-03-15 20:47:48)

dcrabs · 2009-03-15 21:05:20

I have the 2.6.99.902-1 running on a 2.6.28 kernel and so far it is fine. Will try again with the latest kernel rc release but I have to compile it again.
There is still a huge memory leak with compiz enabled, though.

AlphaWolf · 2009-03-15 21:42:15

With Kernel 29-rc8 and KMS enabled my xorg does not start:
Fatal server error:
Cannot run in framebuffer mode. Please specify busIDs for all framebuffer devices

I had do uninstall xf86-video-fbdev. He still tries to load fbdev but can't find it (there are warnings about that in the log). You still don't need an xorg.conf. He automatically detects KMS and load DRI2 / UXA.

Furthermore you shouldn't load your DRI driver (i830, i910 as module). It worked for me but if you have em inside your kernel he switches the mode earlier (+ you get the cute penguins (with the vanilla kernel)).

Alienfreak · 2009-03-16 00:31:15

2.6.99 doenst bring any advantages in my case.
KMS still jsut doesnt work with UXA in X11 ^^

Oh an Demon... could you try to use a SPOILER for your pages over pages of xorg.log? ^^

Last edited by Alienfreak (2009-03-16 00:32:11)

AlphaWolf · 2009-03-16 01:14:32

hoschi wrote:

psych wrote:
It seems to be important for all intel hardware... if you want to use KMS !
Kms does not run at my Dell Laptop (X4500) with any framebuffer driver in kernel.
The reason is simple, with KMS their is only one driver for Framebuffer and X11. If you enable KMS you can read a small warning about that in the description of the kernel.
<edit />
Seems that I can really feel lucky with my MacBook.

x86_64 Testing xf86-video-intel 2.6.99.902-1
Any experience with 2.7 series? I'am afraid about problems with KMS, on the other hand the new driver maybe increase the performance under X11?

I do not understand this part. Could you explain it more detailed, please?:)

Dr4go · 2009-03-16 02:01:24

Thanks for the answers pals!

My goodness... I got everything working using Arch x86_64 with a custom built v2.6.29-rc8 kernel (without all fb drivers, but with fb support as mentioned previously - kms activated by default) using an xfce4 desktop + compiz fusion.

This new stuff in the testing repo (especially the xf86-video-intel v2.6.99.902-1) is working fine and to my pleasure (again) with outstanding performance (using uxa). I even got compiz working with direct rendering (loose binding deactivated), which seems to give a bit more performance than the indirect rendering + loose binding.

exa although isn't working anymore as it should, because using xfce4 (without compiz) the fonts are somehow tiny. With compiz activated I only get garbage on the screen (looks a bit like the graphics adapter would be broken).

Oh... I forgot... I'm also using the Intel GM965 chipset with the x3100 graphics adapter.

Two questions came up reading the latest posts:

1) Because I read it for the second time: Should kms users provide an xorg.conf to the system or should we simply delete (or move) it somewhere (i.e. not use it)?

2) When using it... which modules should be loaded? Currently I have the following loaded: "record", "extmod", "dbe", "dri", "dri2" and "glx".
In addition I'm asking for the module "GLcore", because it was initially in my xorg.conf, but when recreating a new xorg.conf using "Xorg -configure" and xorg-server 1.6.0 it wasn't there anymore.

Again thanks in advance.

EDITed by Drago: removed the stupid dri question. Thanks alphawolf!

Last edited by Dr4go (2009-03-17 00:44:07)

AlphaWolf · 2009-03-16 02:51:56

@Drago
I was talking about the kernel configuration. You can compile your Intel kernel DRI driver either as module or right into the kernel. You should choose the latter one.

triton · 2009-03-16 03:16:24

I upgraded to the new xf86-video-intel 2.6.99.902-1 from [testing] and my X stopped working. I had to downgrade to 2.6.3-2 to get a working X.

This is in a MacBook 3.1, Intel X3100. Kernel package version is2.6.28.7-2.

2.6.3-2 - http://rafb.net/p/w4LzTo94.html
2.6.99.902-1 - http://rafb.net/p/02vaSr38.html

psych · 2009-03-16 16:44:34

Ok...
KMS runs fine with 2.6.29-rc8 and intel 2.6.99.902-1 ...
The next thing... Plymouth... any ideas?

psych · 2009-03-16 17:00:52

KMS works not as good i thought :-( ....
Xorg causes very high CPU usage with some programs... creepsmash (java game) is unplayable now...

AlphaWolf · 2009-03-16 18:17:23

I noticed that, too. Java apps are performing worse (with kms enabled ). X becomes very laggy.

Last edited by AlphaWolf (2009-03-16 18:17:43)

hoschi · 2009-03-16 20:43:12

AlphaWolf wrote:

hoschi wrote:
psych wrote:
It seems to be important for all intel hardware... if you want to use KMS !
Kms does not run at my Dell Laptop (X4500) with any framebuffer driver in kernel.
The reason is simple, with KMS their is only one driver for Framebuffer and X11. If you enable KMS you can read a small warning about that in the description of the kernel.
<edit />
Seems that I can really feel lucky with my MacBook.

x86_64 Testing xf86-video-intel 2.6.99.902-1
Any experience with 2.7 series? I'am afraid about problems with KMS, on the other hand the new driver maybe increase the performance under X11?
I do not understand this part. Could you explain it more detailed, please?:)

GNU/LINUX use a model of two graphics-drivers, one in the kernel for the framebuffer (vt1-6, uvesafb or intelfb) and one for X11 (vt7 i710, i915). Info here, xf86-video-intel is NOT the driver, this is just some important userspace-code for X11. It is very cool that the developers were able to run LINUX in this crazy way, with changing the drivers in runtime and suspend and other things, you see GNU/LINUX is really stable if you can sth. like this.

But in fact: It is complicated and slow. Why not use one layer in the kernel (KMS) which us allow only one driver for everything! Kernel 2.6.29 intoduce KMS for intel-graphics. No more flickering while switching the vts, no more problems with suspend to ram and high screen-resolutions (MacBook Widescreen 1280x800) on the framebuffer! Thats great!

In fact: It stable, its fast, allow more nice features (widescreen resolutions on framebuffer) and it easy to include new features.

Last edited by hoschi (2009-03-16 20:49:17)

hoschi · 2009-03-16 21:03:25

I installed the new xf86-video-intel 2.6.99

No problems. And a much more performance from 41,1 FPS to 41,3 FPS with 1280x800 in IOQuake3

<Update />
Switching from Framebuffer to X11, or wakeup from Suspend to RAM (S3) kills STDIO
I hope kernel-2.6.29-rc8 or later, or the next release of xf86-video-intel will fix this.

Last edited by hoschi (2009-03-16 21:12:30)

psych · 2009-03-16 21:54:26

AlphaWolf wrote:

I noticed that, too. Java apps are performing worse (with kms enabled ). X becomes very laggy.

Any idea why it lags???
I dont understand this... why does kms influence 2d performance so much... and why only special apps????

Dr4go · 2009-03-17 00:37:00

AlphaWolf wrote:

@Drago
I was talking about the kernel configuration. You can compile your Intel kernel DRI driver either as module or right into the kernel. You should choose the latter one.

Oh... well thanks, I didn't realize that you were talking about the kernel configuration. Just to clarify a few things (to help other users). In order to achieve the faster mode switch you need to build the following into the kernel (not as module):
- Device drivers -> Graphics support -> /dev/agpgart (AGP support) -> Intel 440LX...,
- ... -> ... -> Direct Rendering Manager (...) and
- ... -> ... -> Direct Rendering Manager (...) -> Intel 830M... (clearly with i915 driver selection and enabled modesetting)

Anybody got a hint for the two questions in my previous post?

Thanks in advance,
Drago

dcrabs · 2009-03-17 08:04:57

@Dr4go
I think removing the xorg.conf is fine. I only have one with a device section for testing (UXA,EXA) and the "DontZap" option. Nothing else.
I guess that answers both questions.

bender02 · 2009-03-17 09:18:25

hoschi wrote:

...KMS...
In fact: It stable, its fast, allow more nice features (widescreen resolutions on framebuffer) and it easy to include new features.

Not stable nor fast for me, but I'm looking forward to the moment when it will

Last edited by bender02 (2009-03-17 09:18:47)

Arch Linux

#151 2009-03-14 14:18:36

Re: xorg-server 1.6.0 in testing

#152 2009-03-14 17:18:41

Re: xorg-server 1.6.0 in testing

#153 2009-03-14 19:45:16

Re: xorg-server 1.6.0 in testing

#154 2009-03-14 22:06:34

Re: xorg-server 1.6.0 in testing

#155 2009-03-15 02:52:32

Re: xorg-server 1.6.0 in testing

#156 2009-03-15 08:09:03

Re: xorg-server 1.6.0 in testing

#157 2009-03-15 08:28:37

Re: xorg-server 1.6.0 in testing

#158 2009-03-15 08:44:52

Re: xorg-server 1.6.0 in testing

#159 2009-03-15 20:44:41

Re: xorg-server 1.6.0 in testing

#160 2009-03-15 21:05:20

Re: xorg-server 1.6.0 in testing

#161 2009-03-15 21:42:15

Re: xorg-server 1.6.0 in testing

#162 2009-03-16 00:31:15

Re: xorg-server 1.6.0 in testing

#163 2009-03-16 01:14:32

Re: xorg-server 1.6.0 in testing

#164 2009-03-16 02:01:24

Re: xorg-server 1.6.0 in testing

#165 2009-03-16 02:51:56

Re: xorg-server 1.6.0 in testing

#166 2009-03-16 03:16:24

Re: xorg-server 1.6.0 in testing

#167 2009-03-16 16:44:34

Re: xorg-server 1.6.0 in testing

#168 2009-03-16 17:00:52

Re: xorg-server 1.6.0 in testing

#169 2009-03-16 18:17:23

Re: xorg-server 1.6.0 in testing

#170 2009-03-16 20:43:12

Re: xorg-server 1.6.0 in testing

#171 2009-03-16 21:03:25

Re: xorg-server 1.6.0 in testing

#172 2009-03-16 21:54:26

Re: xorg-server 1.6.0 in testing

#173 2009-03-17 00:37:00

Re: xorg-server 1.6.0 in testing

#174 2009-03-17 08:04:57

Re: xorg-server 1.6.0 in testing

#175 2009-03-17 09:18:25

Re: xorg-server 1.6.0 in testing

Board footer