You are not logged in.
http://lists.freedesktop.org/archives/x … 00430.html
Hi guys, xf86-video-intel 2.2.0 is out there and a new release to the arch mirrors would be appreciated. Alternatively, it would be great to compile the driver by ourselves (or even try the latest git, as users with the X3100-965GM still get bugs fixed almost on a weekly basis) and copy manually the new intel_drv.so into the xorg driver directory.
I only have a doubt: since quite a few development libraries / tools are required, is there an easy method to get those dependencies satisfied with a few commands or do we have to run through the nightmare of finding it by ourselves? (pretty much like you do in debian with "apt-get build-dep xserver-xorg-video-intel")
We might post a new article to the Wiki explaining step-by-step how to do this as the new Intel driver has still a long way to go and we should give everyone the opportunity to compile a git version easily.
Regards,
Fabio
Last edited by fabio (2007-11-23 09:50:18)
Offline
just flag the package out of date (if not done already) and wait till the dev find some time to update it.
The depends and makedepends are in the PKGBUILD (I suggest you start reading uppon ABS).
Offline
fabio, I set up the testing abs using this thread http://bbs.archlinux.org/viewtopic.php?id=7154. I then edited the PKGBUILD and changed the release version to 2.2.0 and added the md5sum from the link you posted. I commented out all the patches to the intel driver leaving the patches to mesa. The package built quite easily.
Still testing it out. Noticed that it loads EXA as the default acceleration method. I changed it in xorg.conf to load XAA as compiz seems to run faster with it. The driver seems quite stable on 965G chipset.
Now if only someone can write a PKGBUILD for dri2 from http://wiki.x.org/wiki/DRI2 (Kristian Høgsberg) mesa git repo we may be able to test out some interesting effects with compiz and exa.
Last edited by kishd (2007-11-23 15:01:23)
---for there is nothing either good or bad, but only thinking makes it so....
Hamlet, W Shakespeare
Offline
We shouldn't be using XAA anymore as it contains bugs (see Xv with Compiz) that probably will never get fixed properly. It is true that probably now it still results slower than XAA but in the future we should get past these problems. Anyhoo, Mesa 7.1 (from git) should contain vertex shader support for 965G which seems quite interesting. We can get the chance to compile that as well and see how it works under Arch. As soon as I get out of the office I might give it a try.
EDIT: can you tell me how Xv works under Compiz? Still getting crash or tearing? It probably should be still broken using XAA, but if you revert back to the default EXA it should theoretically be fixed.
Last edited by fabio (2007-11-23 11:23:29)
Offline
Xv seems to work fine with EXA and Compiz no crashes. However the general display is a bit slow with EXA. Will fool around with trying to get the latest mesa and rebuild the driver.
---for there is nothing either good or bad, but only thinking makes it so....
Hamlet, W Shakespeare
Offline
fabio, I set up the testing abs using this thread http://bbs.archlinux.org/viewtopic.php?id=7154. I then edited the PKGBUILD and changed the release version to 2.2.0 and added the md5sum from the link you posted. I commented out all the patches to the intel driver leaving the patches to mesa. The package built quite easily.
I dit the same, but on my system the new driver it crashes X. I'm not using compiz or anything like that, just plain xfce 4, single monitor.
Kind regards, enrique
Offline
kishd wrote:fabio, I set up the testing abs using this thread http://bbs.archlinux.org/viewtopic.php?id=7154. I then edited the PKGBUILD and changed the release version to 2.2.0 and added the md5sum from the link you posted. I commented out all the patches to the intel driver leaving the patches to mesa. The package built quite easily.
I dit the same, but on my system the new driver it crashes X. I'm not using compiz or anything like that, just plain xfce 4, single monitor.
Which chipset are you running and which error message do you get when it crashes?
Offline
enrique, Are you using testing? I have testing enabled and my system is upgraded using all the testing packages.
---for there is nothing either good or bad, but only thinking makes it so....
Hamlet, W Shakespeare
Offline
Tried the new intel driver with EXA and the 865G chipset. ( my other computer). Works fine except it is still slower than XAA with compiz-fusion. Xv seems fixed.
---for there is nothing either good or bad, but only thinking makes it so....
Hamlet, W Shakespeare
Offline
Anybody knows if this version of the drivers uses framebuffer compression? it should help to consume less power...
I answer to myself:
Jesse Barnes (55):
Enable framebuffer compression (use Option "FrameBufferCompression"
Last edited by Damnshock (2007-11-24 00:23:17)
My blog: blog.marcdeop.com
Jabber ID: damnshock@jabber.org
Offline
enrique, Are you using testing? I have testing enabled and my system is upgraded using all the testing packages.
The repository of testing packages it's quite small and the packages on it don't seem to have any relation with the Intel driver (e.g. Mesa, DRI, Xorg etc are not there).
As far as the framebuffer compression is concerned, consuming less power is indeed a good thing especially on laptops, but is there any known drawback for the use of this functionality?
Fabio
Last edited by fabio (2007-11-24 11:39:39)
Offline
kishd wrote:enrique, Are you using testing? I have testing enabled and my system is upgraded using all the testing packages.
As far as the framebuffer compression is concerned, consuming less power is indeed a good thing especially on laptops, but is there any known drawback for the use of this functionality?
None that I know of
My blog: blog.marcdeop.com
Jabber ID: damnshock@jabber.org
Offline
but it's not enabled by default, is it?
edit: just used the PKGBUILD and files from the testing package, commented out the first 4 patches and built the package without problems, and framebuffer compression is enabled by default
Last edited by baze (2007-11-25 16:32:46)
Offline
I'm running 2.2.0 now with Mesa 7.0.1 (standard from core) and I have to say EXA with Compiz is almost unusable for the slowness. I've already read from freedesktop.org that EXA doesn't bring the same performances as XAA yet, but this is a big step backward not just some fps less. I'm wondering if the slowness I'm experiencing is due to outdated components (Mesa, DRM) or it's the actual status of the EXA framework. Being the latter, I'm surprised they decided to adopt it as default already.
Offline
I've installed Mesa 7.0.2 (mesa, libgl) and recompiled xf86-video-intel with Mesa 7.0.2. I had to hack the PKGBUILD of Mesa and put two .h files under /usr/include to let it compile. Leaving FramberCompression enabled gives me corrupted screen (no fonts...?), disabling it works.
Now EXA is much faster with this version of Mesa, but still output of Xvideo looks crappy under Compiz (at least it works, with XAA it crashes the application).
A bit disappointed by this driver release, maybe the big job is being done for Mesa 7.1 and Intel driver 2.3.0?
EDIT: with EXA and FramebufferCompression enabled the fonts disappear. Maybe because I'm using the packages *-lcd and the autohinter enabled. Unfortunately the impression that EXA worked smoothly with Mesa 7.0.2 lasted no more than a few minutes. In fact after stopping the X server and removing kernel modules i915 and drm, I've got a sluggy and jerky output again. Switched back to XAA.
Can somebody try to use EXA with FramebufferCompression enabled to verify if it's just an issue with my font configuration? This way we can isolate at least one bug.
Last edited by fabio (2007-11-26 00:30:29)
Offline
Tried on intel 965G but got the following error in Xorg.0.log
(**) intel(0): Framebuffer compression enabled
(WW) intel(0): Allocation error, framebuffer compression disabled
---for there is nothing either good or bad, but only thinking makes it so....
Hamlet, W Shakespeare
Offline
Which chipset are you running and which error message do you get when it crashes?
From lscpi:
00:02.0 VGA compatible controller: Intel Corporation Mobile 945GM/GMS, 943/940GML Express Integrated Graphics Controller (rev 03)
From Xorg.0.log:
(WW) intel(0): PRB0_CTL (0x0001f001) indicates ring buffer enabled
(WW) intel(0): PRB0_HEAD (0x0401489c) and PRB0_TAIL (0x000148b8) indicate ring buffer not flushed
(WW) intel(0): Existing errors found in hardware state.
Error in I830WaitLpRing(), timeout for 2 seconds
pgetbl_ctl: 0x7ffc0001 pgetbl_err: 0x0
ipeir: 0 iphdr: 2000011
LP ring tail: 148c0 head: 1489c len: 1f001 start 0
eir: 0 esr: 0 emr: ffff
instdone: 4081 instpm: 0
memmode: 306 instps: 800f00c4
hwstam: ffff ier: 0 imr: ffff iir: 0
Ring at virtual 0xa7a6c000 head 0x1489c tail 0x148c0 count 9
0001481c: 00000000
00014820: 00000000
00014824: 00000000
00014828: 00000000
0001482c: 00000000
00014830: 00000000
00014834: 00000000
00014838: 00000000
0001483c: 00000000
00014840: 00000000
00014844: 00000000
00014848: 00000000
0001484c: 00000000
00014850: 00000000
00014854: 00000000
00014858: 00000000
0001485c: 00000000
00014860: 00000000
00014864: 00000000
00014868: 00000000
0001486c: 00000000
00014870: 00000000
00014874: 00000000
00014878: 00000000
0001487c: 00000000
00014880: 00000000
00014884: 00000000
00014888: 00000000
0001488c: 00000000
00014890: 00000000
00014894: 00000000
00014898: 00000000
0001489c: 00000000
Ring end
space: 131028 wanted 131064
Fatal server error:
lockup
I also tried Openbox, and it also crashes. The best part is that X can't be restarted after it has crashed, only a reboot fixes it.
Kind regards, enrique
Offline
enrique, Are you using testing? I have testing enabled and my system is upgraded using all the testing packages.
Nope, just changed the pkgbuild to build 2.2.0
Kind regards, enrique
Offline
I also tried Openbox, and it also crashes. The best part is that X can't be restarted after it has crashed, only a reboot fixes it.
Disable all the options you have in xorg.conf under the device section and try again. If you don't have any I can't be that helpful as the only Intel card I have is a 965G.
Last edited by fabio (2007-11-26 10:44:33)
Offline
With Compiz-fusion and EXA enabled Xv output is extremely slow and jerky. Can someone test it? Running on 965GM.
Offline
With Compiz-fusion and EXA enabled Xv output is extremely slow and jerky. Can someone test it? Running on 965GM.
Some dirty fix.Add to your xorg.conf this line under Device section:
Option "MigrationHeuristics" "greedy"
This works only if you have EXA enabled. Hope it helps.
Offline
fabio wrote:With Compiz-fusion and EXA enabled Xv output is extremely slow and jerky. Can someone test it? Running on 965GM.
Some dirty fix.Add to your xorg.conf this line under Device section:
Option "MigrationHeuristics" "greedy"
This works only if you have EXA enabled. Hope it helps.
Still quite jerky for me.
Offline
somebody on the xorg mailinglist said something about
option "NoExaComposite"
or sth like that, which should speed up a little bit.
Offline
Any update on this guys?
Offline
If you want this package i have built it for myself.
I have been running it for 2 weeks + now, no complaints.
Last edited by B (2008-01-10 18:45:19)
Got Leenucks? :: Arch: Power in simplicity :: Get Counted! Registered Linux User #392717 :: Blog thingy
Offline