Testing latest -intel in Ubuntu

[faibistes]

Everything smooth here

Indeed, it was -sarvat2. No font corruption, I'm experiencing no issues right now. Well, it could be faster, when I activate expo or scale on compiz it drops to 10 fps. No big deal, but compiz feels a bit sluggish overall (funny that dragging a wobbly window only drops to 15 fps).

In fact everything is a bit weird performancewise. If I activate the "test" compiz plugin, in order to see the fps, the performance is way worse. glsxgears with "test"= 30fps, glxgears without "test"=300 fps, even if I drag the wobbling glx gears window.

Should I disable (or enable) any feature in compiz?

[Sarvatt]

Should I disable (or enable) any feature in compiz?

yeah, all of it unless you like the huge memory leak in UXA right now (hint: cat /proc/dri/0/gem_objects and look at the second line)

[Vermind]

yeah, all of it unless you like the huge memory leak in UXA right now (hint: cat /proc/dri/0/gem_objects and look at the second line)

Uhh, I have been using compiz all the time. The bytes seems to be growing... If I disable compiz, will the UXA memory leak not occur? how about in 3D apps?, say SDL

[Sarvatt]

Uhh, I have been using compiz all the time. The bytes seems to be growing... If I disable compiz, will the UXA memory leak not occur? how about in 3D apps?, say SDL

Only seems to be 3d compositing for me. Metacity compositing is fine as are 3d apps. I hear kwin compositing is fine too but I dont use KDE. I've actually had the amount of memory leaked wrap around to a negative number because it was so large the leak made the font corruption problem alot more apparent with how much it was causing swapping thats for sure.

Some good info on the recent changes here-

http://anholt.livejournal.com/41132.html

[Vermind]

I commented on the post there...

Some good info on the recent changes here-

http://anholt.livejournal.com/41132.html

saying that the background + icons become mangled on resume from hibernate.
And he replied:

If you have a regression, please git-bisect and report a bug for the broken commit. Otherwise, it probably won't get fixed.

I should also report the X freeze on suspend-resume using KMS. Where do I do that?

[morgwai]

Hi folks,
I have a problem with compiling latest -intel in Jaunty and noticed there are some experts here, so maybe you will be able to help me:
I provided in my PPA packages for Jaunty from the display-port branch as I have machine with intel chip and display port. Unfortunately since Remove XF86DRM_MODE which was later marged into display-port branch I can not any more use the patch 105_no_modesetting.diff and compilation fails.
Can you tell me how to work around this? Does the driver need some simple patch or do I have to upgrade some other components (xserver, linux kernel) to be able to compile this driver?

I did a brief research through this thread to find some solution as you probably discussed this issue before, but it's quite long and there are a lot of other details discussed, so I'd be grateful if you could answer me.

Many thanks

Morg

[Sarvatt]

I'm seeing some consistant performance increases lately at least

http://global.phoronix-test-suite.com/index.php?k=profile&u=robert-11148-16739-5615

morgwai, comment out the patch in debian/patches/series to remove the patch. you most likely need to update libdrm though at the very least and that needs some special care on jaunty thanks to needing to replace the older linux-libc-dev headers. there have been alot of changes since 2.6.3 that it sounds like you are basing your debian/ from though.. if anything copy the jaunty drm-snapshot from xorg-edgers to your PPA first and let it build/get published before reuploading. I would grab xserver-xorg-video-intel from edgers as well and use the debian directory from that as your basis instead of 2.6.3.

[Craig73]

@Savantt - does your latest build of the kernel include Eric's latest commit (your build seems to be 2 days before the commit referred to by his blog, and his blog talks about 8xx 3d fixes)

[Sarvatt]

Yes, the whole reason I made it was to merge drm-intel into it before it made it into linus' tree, the fixes go there first

http://git.kernel.org/?p=linux/kernel/git/anholt/drm-intel.git;a=summary

[Sarvatt]

Morgwai, looks like it'll be a bit harder than that for displayport support. There are updates to the kernel you have to grab

http://cgit.freedesktop.org/~keithp/drm-intel/log/?h=drm-intel-next

as well as needing the different driver (and possibly drm). It's experiemental right now and they are trying to get it working in #intel-gfx on irc.freenode.net.