AMD Zacate E-350

[curaga]

I think that's the UMS path, dri2 should not call that function...

[neckname]

./src/radeon_driver.c
    if (info->ChipFamily >= CHIP_FAMILY_CEDAR) {
        xf86DrvMsg(pScrn->scrnIndex, X_INFO,
                   "No DRI yet on Evergreen\n");

I think that's the UMS path, dri2 should not call that function...

Hi, thanks for your reply.
I tried to find out if you are right, and I found this (on my own system, not the E-350):

glxinfo|grep -i dri

OpenGL renderer string: Mesa DRI R600 (RS780 9614) 20090101  TCL DRI2

Compiled driver:

xf86-video-ati-6.14.1_build # hexdump -C src/.libs/radeon_driver.o |grep -C 2 yet
0000c2e0  65 63 21 0a 00 55 6e 61  62 6c 65 20 74 6f 20 61  |ec!..Unable to a|
0000c2f0  6c 6c 6f 63 61 74 65 20  63 70 20 72 65 63 21 0a  |llocate cp rec!.|
0000c300  00 4e 6f 20 44 52 49 20  79 65 74 20 6f 6e 20 45  |.No DRI yet on E|
0000c310  76 65 72 67 72 65 65 6e  0a 00 44 69 72 65 63 74  |vergreen..Direct|
0000c320  20 72 65 6e 64 65 72 69  6e 67 20 66 6f 72 63 65  | rendering force|

I guess that part of the sourcecode is being compiled, so "XF86DRI" seems to be set.
radeon_driver.c:

[...]
#ifdef XF86DRI
[...]
    if (info->ChipFamily >= CHIP_FAMILY_CEDAR) {
        xf86DrvMsg(pScrn->scrnIndex, X_INFO,
                   "No DRI yet on Evergreen\n");
        return FALSE;
    }
[...]

This is the message we saw in the xorg.log and there is no other apperance of this warning in the sourcecode.

If I am reading the glxinfo output correctly, I'm running DRI2. Is this sufficent or do I have to check something else to see if it is DRI or DRI2?

[neckname]

Missed the 1 minute editing-limit: I think I have mixed up DRI/DRI2 and UMS/KMS
So.. you think this part of the code is only called when not using KMS, i.e using KMS will work, not using KMS won't work?

[curaga]

Yes, I think the official position was not to make UMS accel at all for the new chips.

[neckname]

Haven't yet had the chance to play with the notebook again, just wanted to say thanks:

Thanks.

[BlackStar]

To anyone with an E-350: we have uncovered the secret of undervolting, with good results. This processor has a lot of voltage headroom and you can reduce power consumption and temperatures by a few degrees both at idle and under load.

This post details how to undervolt your E-350 by modifying pstate vids (safe) and this post explains how to enable coolsense for HP dm1z netbooks under linux:

http://aceracpi.googlecode.com/svn/trun ... acer_ec.pl
sudo ./acer_ec.pl := 0x5A 2
sudo ./acer_ec.pl := 0x5B 2

[BlackStar]

The correct link for acer_ec.pl is http://aceracpi.googlecode.com/svn/t..._ec/acer_ec.pl

[ariel]

Nope that does not help for Debian 6.0 aka squeeze. I wrote a test script that did not create symlinks but changed the package to match the new xvba-va-driver package (but would work with any xvba-video update package too). The libva 1.0.1 in squeeze seems to be too old. No correct vainfo output at least.

Instead of installing xvba-video manually you can just install in in Debian testing/sid:

apt-get install xvba-va-driver

Ubuntu natty has 0.7.7 instead of 0.7.8, maybe file a bug. You can fetch of course the debian package for it:

http://packages.debian.org/sid/xvba-va-driver

Best get of the old xvba-video package if you installed it already:

dpkg --purge xvba-video

it does nothing anyway when you don't use libva from

http://www.splitted-desktop.com/~gbeauchesne/libva/

which uses a completely different packageing. It is really sad that those packages are not compatible. Also why does it need to be in non-free? It would be more logical when it would be in the normal repository just like vdpau-va-driver as the xvba sdk is available - so source could be released too...

Hi Kano - I recently built an htpc with amd fusion / e-350. I installed ubuntu natty amd64, and built my own xbmc with vaapi support

I manually installed from splitted desktop:

xvba-video_0.7.8-1_amd64.deb
libva_0.32.0-1+sds2.tar.gz

and the latest catalyst 11.5 from ubuntu's x-updates ppa.

Whenever I try to play any h264 content in xbmc with "vaapi accelaration" enabled in it, it will crash X. If I disable, xbmc can play any content with no problem (a no acceleration...)


You seem to have gotten vaapi to fully work on linux. Did you do this in natty/amd64? any caveat like a missing link or something else?

What do you mean when you say that the packaging is broken - can that be affecting the install I have?

Thanks

[jaszhou]

Hi Kano - I recently built an htpc with amd fusion / e-350. I installed ubuntu natty amd64, and built my own xbmc with vaapi support

I manually installed from splitted desktop:

xvba-video_0.7.8-1_amd64.deb
libva_0.32.0-1+sds2.tar.gz

and the latest catalyst 11.5 from ubuntu's x-updates ppa.

Whenever I try to play any h264 content in xbmc with "vaapi accelaration" enabled in it, it will crash X. If I disable, xbmc can play any content with no problem (a no acceleration...)


You seem to have gotten vaapi to fully work on linux. Did you do this in natty/amd64? any caveat like a missing link or something else?

What do you mean when you say that the packaging is broken - can that be affecting the install I have?

Thanks

I tried Catalyst 11.3 + CatlystHacks libva + xvba from splitted desktop following this link:

http://wiki.cchtml.com/index.php/Ubu...talyst.2Ffglrx

I haven't got a chance to try it on Natty yet. I hope the same solution works for Natty.

[ariel]

I tried Catalyst 11.3 + CatlystHacks libva + xvba from splitted desktop following this link:

http://wiki.cchtml.com/index.php/Ubu...talyst.2Ffglrx

I haven't got a chance to try it on Natty yet. I hope the same solution works for Natty.

Thanks for the quick reply. It seems that on natty there is a packaging problem/conflict between splitted desktop (and also catalyhacks' ppa) version of libva and xbmc.
It is described in this post:
http://ubuntuforums.org/showpost.php...8&postcount=20

Still trying to figure out how to solve this conflict without having xbmc removed...