You are not logged in.
Thanks for the PKGBUILDs. This is so amazing: At the beginning I had about 300 fps and now 11k xD. Everythings just how it should be. Well ok, Tuxracer (etracer) runs at 20 - 30 fps but still, it is playable now.
:D:D
Akii
Offline
olovram wrote:Amm, when was thiss going t extra?
1.6 final isn't out yet and it's not even in testing. considering unusable 3d performance on pre-965 chips due to no tiling in stable releases (requires patched libdrm, intel driver from git and a patched kernel) this should make it to testing first anyway.
so i'd say xorg 7.5 in extra is still a bit away...
Well, thats pretty bad u_u
Offline
I'm not sure if it's been mentioned already, but there's a whitespace character that breaks stuff in the first package, on line 1 right before the pkgname variable.
EDIT: my mistake--it must have been the way I copied/pasted from the browser
Last edited by Ronin-Sage (2009-02-16 00:23:34)
Offline
Just FYI: some of the versioned PKGBUILDs are outdated by now. You can check http://xorg.freedesktop.org/releases/individual/ and bump the version numbers on your own.
Offline
Tiling patches for GMA945 and below went upstream [1]. Did anyone measure the performance?
Also, please remember that a bug not reported is a bug non-existent.
Offline
I'm a noob in some areas--when he says upstream kernel, he means the stable vanilla/official kernel, 2.6.28.6, right? So it might be as simple as building the latest packages noted on the first page, replacing the version numbers to be the latest, and doing a pacman kernel upgrade?
Thanks.
EDIT: Looks like '.6 isn't in the Arch repos yet, so I'll have to take a good look at kernel-building in the wiki.
Last edited by Ronin-Sage (2009-02-18 18:57:01)
Offline
Actually, I think he means at least 2.6.29-rc5. But you need to compile the entire stack to take advantage of it (PKGBUILDs are in this thread).
Offline
I see, thanks.
Offline
Tiling patches for GMA945 and below went upstream [1]. Did anyone measure the performance
I tried a mixture of current+rc+git of the stack on the rc5 kernel. I am not sure whether all the fixes were in, but Jesse Barnes's fence register management was. See http://lkml.org/lkml/2009/2/13/346
Unfortunately, the results aren't great for me. Desktop effects are not fun and glxgears reports under 300.000 fps, while I get twice as much with the current Arch stack.
Offline
Skimming through recent git commits, it appears that some intel-related fixes have been applied to xf86-video-intel master branch (2.5.99.2 pkgbuild in this thread is pretty old) and drm master as well as drm-next branch of the kernel. Perhaps it is necessary to compile the entire stack from git. mutlu_inek, how does your dmesg and Xorg.0.log look like? Does it scream bloody murder?
edit: Jesse Barnes: "You'll need to upgrade your kernel as well, with the execbuffer tiling patches. You should be able to use Eric's drm-intel-next branch (from git://git.kernel.org/pub/scm/linux/kernel/git/anholt/drm-intel.git), along with git master of xf86-video-intel and libdrm and get tiled rendering (be sure you use UXA as your accelmethod though)." So yeah, I am blind.
Last edited by fijam (2009-02-19 11:53:57)
Offline
having upgrading all packages here to the latest version number as well as the git-packages currently results in no working compiz or kwin effects for me but the White Screen of Death right now. glxinfo says direct rendering is enabled though and composite still works in xfce.
I tried out the git packages for libdrm and the intel driver in aur, but they no longer work. So unless someone can post working package builds for those (and presumably the intel-next kernel branch? I'm clueless about git pkgbuilds) I guess we'll just have to bide our time.
Last edited by b9anders (2009-02-19 15:14:45)
Offline
Great news, y'all(yes I'm from the southern US), I finally got some noticeable stability/performance boosts, using the kernel-2.6.29rc(4?) and the git PKGBUILDS(with updated version #s) for graphics packages. I can get ~1000fps with glx, without tweaks, and pSX(Sony Playstation emulator) seems to work nicely. I'm in the process in correctly configuring my iwl3945 kernel stuff(was disabled in my old manifestation of '.29), so I'll post back when the new build finishes.
Offline
Skimming through recent git commits, it appears that some intel-related fixes have been applied to xf86-video-intel master branch (2.5.99.2 pkgbuild in this thread is pretty old) and drm master as well as drm-next branch of the kernel. Perhaps it is necessary to compile the entire stack from git. mutlu_inek, how does your dmesg and Xorg.0.log look like? Does it scream bloody murder?
I had to use xf86-video-intel-2.6.1. The git branch has a memory leak that fills memory and swap within seconds, locking up the system. IIRC I checked the logs and it looked fine. I will check again with the git driver after the leak has been fixed.
Great news, y'all(yes I'm from the southern US), I finally got some noticeable stability/performance boosts, using the kernel-2.6.29rc(4?) and the git PKGBUILDS(with updated version #s) for graphics packages. I can get ~1000fps with glx
Which chipset is this on?
Offline
On my 945GM-equipped laptop, I can get just under 1000fps with glxgears with this current setup, but for some reason, logging out/in roughly halves the glxgears report on the next run. In other words, I get the good performance upon rebooting, but logging out, then in botches it somehow.
(I psuedo-tested performance with glxgears, pSX+Final Fantasy VII, and UrbanTerror)
Current setup:
[custom kernel-2.6.29rc5)]
libpthread-stubs-0.1-1
libdrm-2.3.1-3
libxcb-1.2-1
libx11-1.2-1
dri2proto-1.99.3-1
xf86driproto-git-20090218-1
libgl-7.2-1
mesa-7.2-1
intel-dri-7.2-2
inputproto-1.5.0-1
pixman-0.14.0-1
randrproto-1.2.99.3-1
xorg-server-1.5.3-4
xf86-video-intel-2.4.3-1
xf86-input-evdev-2.1.2-1
Looks like I still have some configuring(and learning) to do...
Last edited by Ronin-Sage (2009-02-19 23:40:39)
Offline
That's a pretty confusing mix.
Offline
Okay, dmesg seems to outputs "[ 773.490334] [drm:i915_get_vblank_counter] *ERROR* trying to get vblank count for disabled pipe 0", so I guess it's a vblank(same as v_sync?) issue with i915.
Offline
I downloaded 2.6.29-rc5... I noticed glxgears is now 980 on my i965.
However, I am having a strange problem (Scheduler???) with 2.6.29
I've got CFQ as default, Low Latency Desktop, 1000Hz timer, its strange,
programs pause if i try to launch them .... pause for seconds... the system
is acting strange when any program is run or launched....
the "group CPU scheduler" which is new I disabled it...
i'm investigating.... if I can't fix this i'll have to go back to the 2.6.28 kernel.
-jim
Offline
*sigh* I can't seem to get any bleeding-edge configuration to work either at all or have normal performance, and performance is shoddy with kernel '28 and the current graphics package lineup in extra(?).
I'm giving up for the moment, and actually installing/using Office 2007 and such on my XP partition until the situation gets brighter :\
Offline
The thing that I don' t understand:
Fedora 10 on this laptop (Sony Vaio BZ12XN) gives me 1100+ fps in glxgears and super smooth compiz effects.
Using Arch I have ~300 fps in glxgears and crappy compiz performance.
As far as I know, both distros use the same version of the driver in their stable repos: 2.4.3 and have the same xorg-server: 1.5.3
So what is causing the poor performance in Arch ? mesa ? intel-dri ? I'm lost here...
Offline
Using a new kernel alone or kernel with a mix of old and new software in the stack will not gain any performance. I wish I had the time to set up a sandbox system...
LordChaos73, that's an interesting observation, provided that the rest of the stack is the same and you use the same acceleration method on both. Perhaps Fedora backported some patches?
Offline
I backed off 2.6.29-rc5, at one point my system locked up. I use Meld, a Gtk graphical diff program, to check diffs in my previous kernel config versus old, didn't see anything.
So back on 2.6.28, I tried to compile Mesa 7.3 by hand. got an error:
Mesa 7.3
./configure --prefix=/usr --with-dri-drivers=i965
-UIN_DRI_DRIVER -DDEFAULT_DRIVER_DIR=\"/usr/lib/dri\" dri2.c -o dri2.o
dri2.c: In function 'DRI2Connect':
dri2.c:117: error: 'xDRI2ConnectReq' has no member named 'window'
dri2.c:118: error: 'xDRI2ConnectReq' has no member named 'driverType'
dri2.c:125: error: 'xDRI2ConnectReply' has no member named 'deviceNameLength'
dri2.c:135: error: 'xDRI2ConnectReply' has no member named 'deviceNameLength'
dri2.c:143: error: 'xDRI2ConnectReply' has no member named 'deviceNameLength'
dri2.c:146: error: 'xDRI2ConnectReply' has no member named 'deviceNameLength'
dri2.c:151: error: 'xDRI2ConnectReply' has no member named 'deviceNameLength'
dri2.c:152: error: 'xDRI2ConnectReply' has no member named 'deviceNameLength'
[i965 + 2.6.28 kernel]
installed:
libdrm-2.4.4
xf86-video-intel-2.6.1
dri2proto-1.99.3
*sigh* I can't figure out what im doing wrong.
Last edited by 3lusive (2009-02-20 22:40:05)
Offline
did you update opengl to 7.3 first?
Offline
this is a pretty cool blog from an xorg developer
http://hoegsberg.blogspot.com/
interesting stuff contained within this blog!
Offline
Updated yesterday to xorg-server 2.5.99.903 and latest ligl/mesa/intel-dri git and suddenly ctr+alt+backspace stopped working. Also x server produces a hard lock if I logout of kde and in the meantime some .fdi files are changed (or the HAL daemon is restarted) - the only way to restart the xserver is ctrl + alt + F1 && /etc/rc.d/kdm3 stop.
It's kind of (but not annoying since I rarely restart X server)
My victim you are meant to be
No, you cannot hide nor flee
You know what I'm looking for
Pleasure your torture, I will endure...
Offline
Updated yesterday to xorg-server 2.5.99.903 and latest ligl/mesa/intel-dri git and suddenly ctr+alt+backspace stopped working. Also x server produces a hard lock if I logout of kde and in the meantime some .fdi files are changed (or the HAL daemon is restarted) - the only way to restart the xserver is ctrl + alt + F1 && /etc/rc.d/kdm3 stop.
It's kind of (but not annoying since I rarely restart X server)
So xorg-server 2.5.99.903 solved the error you and I had before? ->
Fatal server error:
Caught signal 11. Server aborting
Offline