Xgl and i810

Message

alanic · Post by **alanic** » Thu Feb 16, 2006 6:02 pm

Hello,

I wanted to start a new thread because when you search for "Xgl i810", all Xgl related threads come up as a result because of the use flags in quoted codes. And I couldn't find any useful information for i810 anywhere.

Let me write down my situation:
I have an Intel chipset on my laptop and I use the i810 driver with the i915 kernel module. Here is the output of lspci:

Code: Select all

00:02.0 VGA compatible controller: Intel Corporation 82852/855GM Integrated Graphics Device (rev 02)

I managed to build Xgl and it is running well, but there isn't much hardware acceleration as far as I can see, because no matter what combination of -accel parameters I give to Xgl, it runs quite flickery and bad to the eye. Even scrolling stuff inside windows is slow. It is the same if I don't give any parameters to Xgl.
I have direct rendering true in my normal X sessions. But I think Xgl needs some kind of extension that the current driver does not support. Does anybody know how we can list the possible parameters to -accel?

So, if anybody has a more successful experience, or any knowledge whatsoever that may lead to a smoother Xgl on i810 driver it would be very useful.

By the way, if you are using KDE, you need glxcompmgr. It is not in Hanno's latest tar's, try his older tar's.

BlackEdder · Post by **BlackEdder** » Thu Feb 16, 2006 6:06 pm

http://en.opensuse.org/Xgl wrote:Intel / open source driver "i810"
Driver does not accelerate blits from back buffer to front buffer which makes it very slow. The driver will soon be updated and should then work OK.
Driver has neither pBuffer nor FBO support. When using a composite manager all windows are rendered in software and only compositing is hardware accelerated. Astonishingly, this works well enough for most use cases. FBO support is supposed to be added to the driver in the near future.
XVideo YV12 surfaces are hardware accelerated (when no composite manager is used), but due to a bug in the driver the video will miss one of the color channels, leading to false greenish/purple colors. This has to be investigated.

I only seem to get a black window. sometimes with a refresh it becomes blue, but the xterms are greenish and I can't see what I'm typing. Did you also have this problem?[/quote]

alanic · Post by **alanic** » Thu Feb 16, 2006 6:14 pm

About the opensuse quote:
You know what, actually there were two patches lying somewhere in these forums about intel stuff. I didn't try them yet, and I'm not sure if I have time for them, but maybe they might work. I hope that "soon" will be soon:)

Black windows: honestly, I never had my windows be all black no matter what I do. When I run glxcompmgr for the first time, most of the windows go black because they don't update their contents for that time being. But after I write something in them or put them under a window and back they are normal again. Is this what you have?

BlackEdder · Post by **BlackEdder** » Thu Feb 16, 2006 6:22 pm

Hmm they go black when I start compiz and I can get it better, but not in any way usuable. I actually don't use gnome, but understood that glxcomp has only limited functually compared to compiz. Is this true?

I think the fixed i810 drivers will take some time, but maybe you are in luck

alanic · Post by **alanic** » Thu Feb 16, 2006 6:30 pm

With glxcompmgr I could run wobbly, shadow, fade(terribly slow) and zoom(sweet). I couldn't try too much, but kde's virtual desktop thing does not agree with rotate and I couldn't run rotate. It might only be a key combination problem. If it was fast, glxcompmgr would be the best thing on earth for me since it runs with KDE;)

edit:

I just received this email from Hanno:

There is a kde-window-decorator. If you use my ebuilds, grab the latest
overlay and build compiz with USE="kde"

edit:

it requires qt4. let's see if it'll work with kde 3.5.1

irondog · Post by **irondog** » Thu Feb 16, 2006 6:37 pm

So, I can run XGL on i810?

alanic · Post by **alanic** » Thu Feb 16, 2006 6:55 pm

This might help you decide:
shot (click on the image twice to resize)

It's not usable because it's not fast enough though.

BlackEdder · Post by **BlackEdder** » Thu Feb 16, 2006 6:57 pm

I remembered I use dri provided by the kernel, this probably causes the problems with xgl. I can't seem to get direct rendering to work otherwise though

Code: Select all

drmOpenDevice: node name is /dev/dri/card0
drmOpenDevice: open result is -1, (No such device or address)
drmOpenDevice: open result is -1, (No such device or address)
drmOpenDevice: Open failed
drmOpenDevice: node name is /dev/dri/card0
drmOpenDevice: open result is -1, (No such device or address)
drmOpenDevice: open result is -1, (No such device or address)
drmOpenDevice: Open failed
[drm] failed to load kernel module "i915"
(II) I810(0): [drm] drmOpen failed
(EE) I810(0): [dri] DRIScreenInit failed. Disabling DRI.

HazarT · Post by **HazarT** » Thu Feb 16, 2006 8:12 pm

Same problem here!

I have tried the 2.6.15 kernel DRM i915 driver as staticly compiled, module, and even tried the x11-dri modules that you can emerge. All setups give this error en xorg.x.log

Code: Select all

drmOpenDevice: node name is /dev/dri/card0
drmOpenDevice: open result is -1, (No such device or address)
drmOpenDevice: open result is -1, (No such device or address)
drmOpenDevice: Open failed

I'm beginning to think that it could be a problem with udev not updating the device filesystem /dev, when the module is inserted. I have also tried to manually create the device nodes, but without any luck.

Any help would be greatly apreciated.

evolusion · Post by **evolusion** » Thu Feb 16, 2006 10:53 pm

Well you're getting further than I am with my i810 anyway. Everything compiles fine, but when I try to start Xgl, regardless of options specified, I get the following

Code: Select all

X Error of failed request:  BadLength (poly request too large or internal Xlib length error)
  Major opcode of failed request:  158 (GLX)
  Minor opcode of failed request:  1 (X_GLXRender)
  Serial number of failed request:  93
  Current serial number in output stream:  94

Any helps/tips would be mucho appreciated

irondog · Post by **irondog** » Fri Feb 17, 2006 8:15 am

alanic wrote:This might help you decide:
...

lol

HazarT · Post by **HazarT** » Fri Feb 17, 2006 8:07 pm

evolusion wrote:Well you're getting further than I am with my i810 anyway. Everything compiles fine, but when I try to start Xgl, regardless of options specified, I get the following
Code: Select all
X Error of failed request:  BadLength (poly request too large or internal Xlib length error)
  Major opcode of failed request:  158 (GLX)
  Minor opcode of failed request:  1 (X_GLXRender)
  Serial number of failed request:  93
  Current serial number in output stream:  94
Any helps/tips would be mucho appreciated

Look at my output:

Code: Select all

laptop ~ # Xgl :1
X Error of failed request:  BadLength (poly request too large or internal Xlib length error)
  Major opcode of failed request:  159 (GLX)
  Minor opcode of failed request:  1 (X_GLXRender)
  Serial number of failed request:  93
  Current serial number in output stream:  94

It is the exact same error message. I think it's because DRM is not enabled. So we ought to rename the title of this thread to. "Xorg 7.0 DRM on i810" - because that's the first step. We can't have Xgl before DRM is working.

alanic · Post by **alanic** » Fri Feb 17, 2006 9:40 pm

I wrote:

it requires qt4. let's see if it'll work with kde 3.5.1

Mesa currently does not compile on i810. Hanno's ebuild tries to compile an i830 driver that's not there. If you remove it, it still doesn't compile:

Code: Select all

intel_ioctl.c: In function `intelGetLastFrame':
intel_ioctl.c:49: error: `I915_PARAM_LAST_DISPATCH' undeclared (first use in this function)
intel_ioctl.c:49: error: (Each undeclared identifier is reported only once
intel_ioctl.c:49: error: for each function it appears in.)

Xgl doesn't either:

Code: Select all

../../../../GL/mesa/.libs/libGLcore.a(s_arbshader.o): In function `_swrast_exec_arbshader':
s_arbshader.c:(.text+0x4e): multiple definition of `_swrast_exec_arbshader'
../../../../GL/mesa/.libs/libGLcore.a(lt1-s_arbshader.o):s_arbshader.c:(.text+0x4e): first
defined here

*holds his current /usr/bin/Xgl tight*

HazarT · Post by **HazarT** » Fri Feb 17, 2006 9:49 pm

alanic wrote: Mesa currently does not compile on i810. Hanno's ebuild tries to compile an i830 driver that's not there. If you remove it, it still doesn't compile:
Code: Select all
intel_ioctl.c: In function `intelGetLastFrame':
intel_ioctl.c:49: error: `I915_PARAM_LAST_DISPATCH' undeclared (first use in this function)
intel_ioctl.c:49: error: (Each undeclared identifier is reported only once
intel_ioctl.c:49: error: for each function it appears in.)

I solved this problem by adding I915_PARAM_LAST_DISPATCH with the value 3 to the kernel heders. You can also just upgrade to the newest linux 2.6.16-r1 kernel, which has the correct version kernel headers. I'm just installing it now. Let's see if the DRM works then

-------

EDIT: DRM works with the newest kernel headers. Just install 2.6.16-r1 kernel, and DRM works out of the box

Code: Select all

laptop ~ # glxinfo
name of display: :0.0
display: :0  screen: 0
direct rendering: Yes
server glx vendor string: SGI
server glx version string: 1.2
server glx extensions:
    GLX_ARB_multisample, GLX_EXT_visual_info, GLX_EXT_visual_rating,
....
....

EDIT: Updated the XGL_HOWTO wiki to reflect this.

HazarT · Post by **HazarT** » Fri Feb 17, 2006 11:18 pm

Regarding the i830 compile error when compiling mesa, i have sent a patch to hanno. The reason is that the i830 and i915 has merged into one-and-same driver.

Patch:

Code: Select all

--- mesa-6.4.3_alpha20060216.ebuild     2006-02-17 23:51:27.000000000 +0100
+++ mesa-6.4.3_alpha20060216.ebuild.new 2006-02-18 00:09:36.000000000 +0100
@@ -104,7 +104,7 @@
                add_drivers ffb
        fi
        if use video_cards_i810; then
-               add_drivers i810 i830 i915
+               add_drivers i810 i915
        fi
        if use video_cards_mga; then
                add_drivers mga

evolusion · Post by **evolusion** » Sat Feb 18, 2006 5:19 pm

DRM works with the newest kernel headers. Just install 2.6.16-r1 kernel, and DRM works out of the box

Does that mean you have xgl working then and the previous error has gone? If so, how is the performance?
I'm having trouble getting DRI working (tried the 2.6.16-r1 vanilla sources) and am wondering if it's worth the bother of figuring it out just now...

HazarT · Post by **HazarT** » Sat Feb 18, 2006 5:43 pm

Xgl is running yes. But when i start compiz i get the same black windows problem as Nvidia people does. So it's not really usable yet.

Edit: After i have updated to the latest hanno snapshot, mesa, xgl, and friends, it seems that i can't get drm working again. Even after re-emerging =mesa-6.4.2-r1, the xorg-server and xorg-i810 driver... So i think once again, that we are in the same boat.

yamakawa · Post by **yamakawa** » Sun Feb 19, 2006 6:38 pm

Since this is the 1st page, I believe we should include the following links for convienience.
Otherwise we would be lost!

http://gentoo-wiki.com/HOWTO_XGL Gentoo Wiki
http://www.hboeck.de/item/282 Hanno's blog
http://dev.gentoo.org/~hanno/ Hanno's overlay
http://svn.hboeck.de/xgl-overlay/ Hanno's overlay repository
http://en.opensuse.org/Xgl OpenSUSE site

rob2687 · Post by **rob2687** » Sun Feb 19, 2006 6:52 pm

HazarT wrote:
evolusion wrote:Well you're getting further than I am with my i810 anyway. Everything compiles fine, but when I try to start Xgl, regardless of options specified, I get the following
Code: Select all
X Error of failed request:  BadLength (poly request too large or internal Xlib length error)
  Major opcode of failed request:  158 (GLX)
  Minor opcode of failed request:  1 (X_GLXRender)
  Serial number of failed request:  93
  Current serial number in output stream:  94
Any helps/tips would be mucho appreciated
Look at my output:
Code: Select all
laptop ~ # Xgl :1
X Error of failed request:  BadLength (poly request too large or internal Xlib length error)
  Major opcode of failed request:  159 (GLX)
  Minor opcode of failed request:  1 (X_GLXRender)
  Serial number of failed request:  93
  Current serial number in output stream:  94
It is the exact same error message. I think it's because DRM is not enabled. So we ought to rename the title of this thread to. "Xorg 7.0 DRM on i810" - because that's the first step. We can't have Xgl before DRM is working.

I get the same errors on a Radeon 7000.

benjaoming · Post by **benjaoming** » Sun Feb 19, 2006 7:32 pm

HazarT wrote: I solved this problem by adding I915_PARAM_LAST_DISPATCH with the value 3 to the kernel heders.

I'm not sure what you mean.. what headers? /usr/src/linux/?

I get the same error, too.

- Benjamin

alanic · Post by **alanic** » Sun Feb 19, 2006 9:13 pm

I'm not sure what you mean.. what headers? /usr/src/linux/?

media-libs/mesa-6.4.3_alpha20060216 is now on the portage tree. I use that mesa right now and Xgl works, even though I couldn't run glxcompmgr with it. I believe I could if I gave enough time.

To compile the source in cvs you can also replace I915_PARAM_LAST_DISPATCH in intel_ioctl.c with whatever value you would put to that header since that macro is not used anywhere else. The resulting binary didn't work though.

Also, x11-drm does not work for me with 2.6.15-r1 as HazarT said, bu the drm in the kernel works to get direct rendering.

benjaoming · Post by **benjaoming** » Mon Feb 20, 2006 9:26 am

Yeah, new kernel 2.6.16-r4 works fine. But Mesa snapshot has broken i915 support (compiles fine). Downgraded to Mesa 6.4.2 and now glx works again. I can even start up Xgl... but not compiz.

- Benjamin

evolusion · Post by **evolusion** » Wed Feb 22, 2006 6:57 pm

I downgraded to mesa 6.4.2-r1, removed x11-drm and compiled it into the kernel (2.6.16-r4 vanilla) and it is a whole lot closer to working for me anyway, so this seems to be the magic combination for now...

My previous error message no longer appears and the Xglx window appears for me, but I'm having some issues with the display inside it, but at least it's progressing =)

SPiRiTs · Post by **SPiRiTs** » Mon Feb 27, 2006 8:37 am

same for me,dri working with

2.6.16 kernel with drm and i915 kernel modules (not x11-drm, too old).

and mesa 6.4.2

mesa cvs compile cleany but doesn't work.

CoffeeBuzz · Post by **CoffeeBuzz** » Mon Feb 27, 2006 7:41 pm

Hey i810 people. I've been trying to get i810/i915 support working with mesa cvs. I don't have a intel card however but I want my ebuild to work for everyone (i've already patched that i915_last_dispatch declaration in so it compiles clean).
So let me get this straight, no one here has had Xgl working with mesa 6.4.3 (cvs)?

All you get is this:

X Error of failed request: BadLength (poly request too large or internal Xlib length error)
Major opcode of failed request: 159 (GLX)
Minor opcode of failed request: 1 (X_GLXRender)
Serial number of failed request: 93
Current serial number in output stream: 94

But if you use mesa 6.4.2-r1 it works? Because DRM in 2.6.15 or xorg7 isnt up to snuff. ?

If you use 2.6.16, mesa 6.4.3 works with xgl?