Home Phoronix Phoronix Forums X.Org Videos From FOSDEM 2008

RadeonHD IRC Logs For 2008-7-26

Search This Log:

1 Search Results For Hello
Jump To First Result


yangman: dmb: works fine here with DRI and EXA enabled
dmb: yangman, yeh, wondering about XAA though
dmb: because i need to use 2 X servers
dmb: exa doesn't work very well on the second server
yangman: I don't see why it wouldn't
dmb: and it basically has no 2d accel
dmb: which is unusable
dmb: yangman, drm only works on one server
dmb: so the second server has no drm
yangman: oh, multiple server
yangman: yeah, I've no idea then
mikkoc: hm, lots of commits in radeonhd
mikkoc: what's the difference now between radeon and radeonhd?
mikkoc: does 2d work together with 3d now?
yangman: mikkoc: yes
dmb: yangman, do you still expirence screen corruption?
dmb: I do
yangman: which kind?
yangman: I guess it doesn't really matter. the answer is yes
yangman: and also minor corruption when running 3d apps now, but it goes away when I switch VT and back, just like the other one
mikkoc: yangman: so, regarding 3d, which is better now? radeon or radeonhd?
yangman: mikkoc: they use the same 3d backend
yangman: I suppose the hooks in radeon is more mature, but "better" is a really loaded term
mikkoc: eheh
dmb: stability is quite weak with XAA
dmb: and dri
mikkoc: yangman: were you referring to this when you said radeon is more mature than radeonhd? :D
mikkoc: http://i36.tinypic.com/96f09k.png
mikkoc: after activating desktop effects: http://i38.tinypic.com/wcgxgj.png
mikkoc: i guess radeonhd needs a bit more time?
ndim: mikkoc: Just last night, the quick-and-dirty-2D branch has been merged into radeonhd master - which means that everybody tracking master is testing that code for the first time.
mikkoc: yea, that's why i tried radeonhd
mikkoc: saw all those commits
udovdh: ndim, does this new code also work for rv630?
udovdh: (could be a reason to reboot this workstation also into 2.6.26)
udovdh: if it works for rv630, which changes in xorg.conf should I make for exa/xaa?
Fuddl: udovdh: i guess not, since 2d acceleration needs the 3d engine on r6xx GPUs, AFAIK. and for these cards the specifications haven't been released, yet
udovdh: ok, so no reason yet...
udovdh: thanks Fuddl
Fuddl: udovdh: it was a guess, so don't rely on my information ;)
udovdh: oh. hmmm
udovdh: anyone? ;-)
BigBrain: Well, I can at least confirm that the 2d accel relies on the 3d engine for >=r600
BigBrain: and the specification part is right, too ;-)
ndim: udovdh: Not being on the "inside" crew, I'd think Fuddl is right.
udovdh: ah, ok.. thanks
ndim: I guess, some people we know are busy working on the R6xx 3D stuff behind closed doors still under NDA, and will push the commits to the public git repos once the R6xx 3D specs have been made public.
Fuddl: as i know my luck, this will probably happen while i'm on vacation and offline ;)
ndim: 14:24:15< daniels> in the 2d side, you have mach64, r128/radeon/r200/r300/r400, and r500/r600
ndim: 14:24:30< daniels> in the 3d side, you have mach64, r128, radeon/r200, r300/r400/r500, and r600
ndim: Hmm. I'm not sure about the "2d and r600" part is correct, but apart from that, it is probably accurate.
udovdh: I will be online all the time
udovdh: except for september I guess
udovdh: so before september?
udovdh: ;-)
Fuddl: oh damn, i'll be offline in september again, d'oh! :)
udovdh: even if I were online, I could not see the video of the radeonhd boxes at home... ;-)
Fuddl: suggests using a webcam
ndim: KVM-over-IP.
Fuddl: lol
Fuddl: udovdh: i can confirm there's no 2d acceleration on r6xx, yet
udovdh: Fuddl, ok, thanks
udovdh: then we continue the wait ;-)
Fuddl: .oO( "chuck norris doesn't sleep - he waits" )
Fuddl: hm, the black screen on mplayer start/quit still exists. i hoped this was fixed since version 1.2.1
Simon-: invoices ATI for the last hour of his time
Simon-: fglrx corrupts /proc/slabinfo badly... for some reason blacklisting fglrx has decided to stop working and I didn't notice it was still being loaded while trying to bisect things :/
n0body: Hello, I can't build radeonhd anymore. I get exactly the same error: http://www.phoronix.com/forums/showpost.php?p=40269&postcount=13
Fuddl: n0body: i had that, too an hour ago. i was missing some additional development drm/dri packages
n0body: Fuddl: Great! I'll have a look and install a few packages, thanks.
Fuddl: n0body: i think it were these (debian, sid) packages i had to add to the regular build dependencies of the debian package of version 1.2.1: x11proto-gl-dev, libdrm-dev, mesa-common-dev, libdrm-dev, x11proto-dri2-dev, x11proto-xf86dri-dev, libgl1-mesa-dev
n0body: okay, libdrm-dev is installed, trying again...
n0body: Thanks for you help, now it works :)
Fuddl: but there's still one inconsistency between src/rhd.h and src/rhd_driver.c, since DMAForXv is declared between #ifdef USE_DRI, but it's always referenced in if (rhdPtr->ChipSet < RHD_R600) { rhdPtr->DMAForXv = TRUE; ...
udovdh: Fuddl, if this is a programming `error` (feature?) please let the devs know.
Fuddl: udovdh: i would guess it's a left-over from the 2d-quick'n'dirty merge into master
udovdh: indeed some sort of placeholder
udovdh: but it should be looked into
Fuddl: [x] reported on ML
cire: is this a known issue: after some hours or minutes of work (KDE 3.5.9, radeonhd 1.2.1, xorg 1.4.2, dual head, HD3650) my mouse pointer gets weird. It is no longer an arrow, but still changing it's appearance when moving over window borders for resizing etc. Also when resizing it is not the usual symbol (2 arrows).
cire: I saw that it was altered as if it was divided vertically, and afterwards both parts swapped
cire: like /\ to \/
cire: my flatmate also has this problem
cire: (same system)
cire: <=> gets to >=<
ndim: cire: If that is the 1.2.1 release, that is quite old. If it is a more recent git snapshot, telling us "1.2.1" is a little unspecific (there have been 272 commits in master since 1.2.1 :)
cire: it is the one currently in debian unstable (sid)
cire: how do I get the correct version?
ndim: Can't remember. "dpkg something".
cire: ii xserver-xorg-video-radeonhd 1.2.1-2
cire: this is a debian version number
cire: I think
ndim: http://packages.debian.org/changelogs/pool/main/x/xserver-xorg-video-radeonhd/xserver-xorg-video-radeonhd_1.2.1-2/changelog
ndim: "Sun, 11 May 200"
Fuddl: ... or dpkg -s | grep Version ;)
cire: in changelog, last change was april, 14th by Matthias Hopf
ndim: cire: Yupp. 273 commits in master since that one.
cire: so the Debian packages are a bit... outdated?
ndim: Feel free to file a bz at freedesktop.org or mail the list, but my guess is that 273 commits later, the devs probably won't be able to track the issue down.
Fuddl: cire: compared to the upstream repository yes. but regarding the latest upstream release the package is up-to-date
ndim: cire: A lot has happened since the release, but there has not been any point in time in which a snapshot would have been 100% useful as release.
cire: okay. Since there's a lot of progress at the moment, I'll just wait until a new package comes out.
ndim: I am shipping git snapshots in Fedora, but the issues are difficult to track.
ndim: Fuddl: Have you noticed the most recent commit? :)
Fuddl: ndim: yes i have - quite fast reply, wow! ;)
Fuddl: ndim: so you're maintaining radeonhd for fedora?
cire: if a new version comes out, and my problem remains, I'll write a bz. In Fact I can work with what I have at the moment. Thanks.
ndim: cire: Thanks for the report.
cire: Thanks for your support
Fuddl: cire: i built^Wugly hacked a debian package from master/HEAD this morning, if you're interested
cire: I am :)
ndim: Writing hardware drivers sucks. You never know when to release, and you can't just have a regression test suite to run which tests all known previous bugs.
Fuddl: cire: are you on amd64?
cire: no
ndim: Fuddl: Yes, I am responsible for breaking the Fedora radeonhd package.
Fuddl: ndim: i enjoyed that at university ;) ... but it was nothing compared to a graphics card driver...
ndim: I started it back when radeonhd was the only driver working for my R5xx and I was impatient to wait for radeon to catch up.
cire: Fuddl: Do you only have an amd64 binary?
Fuddl: ndim: im just curious: by what createria do you package a new snapshot? i'm in the debian games team and game developers often argue to package snapshots which i feel quite unconfortable to handle :/
ndim: Nowadays, radeon and radeonhd have partially disjunct sets of bugs, so I am still maintaining radeonhd.
Fuddl: cire: yep, i only built it for amd64, but i can offer the source package
yangman: http://bugs.freedesktop.org/show_bug.cgi?id=13405 <- is it this bug?
ndim: Fuddl: The last snapshot which made it to F-9 stable is from April :)
ndim: Too many regressions with the newer snapshots.
cire: yangman: no, it isn't
Fuddl: ndim: ah, so you picked the most recent one before the release?
ndim: It was after 1.2.1, that's for sure.
cire: yangman: it doesn't matter on which area my cursor is - when it is transformed it remains in the transformed state.
cire: and nothing is flickering ;)
ndim: Anyway, I'm not sure yet what to expect from the quick-and-dirty-2d merge.
yangman: cire: read the entire thread. people describe it differently
Fuddl: ndim: better try something than doing nothing - keep thinking positive! ;)
ndim: Yeah, people can always fall back to radeon :)
yangman: although if the mouse is constantly corrupted, then it's a different issue
cire: read all the posts - it is not my bug
Fuddl: cire: you can dget http://wwwcip.informatik.uni-erlangen.de/~sibrklei/packages/unofficial/xserver-xorg-video-radeonhd_1.2.1+git-1.dsc and built it if you want to. the package worked for a personal test for me, but it's just ugly and quick-and-dirty hacked together - keep that in mind :)
cire: okay, thank you anyway :)
cire: hi, I have to add something: I now switched to radeon and restarted X (without rebooting). The cursor remains transformed.
ndim: is running b0f470aa now.
ndim: Yay. The same torcs hang.
megahealth: hi
matteo: what's in the quick_and_dirty_2d branch?
airlied: matteo: nothing now..
airlied: its merged into master
matteo: mm
matteo: why i get poor results unless I use ShadowFB?
matteo: i have a radeon X1600
matteo: but a 4870 is coming
airlied: not sure, you got a drm loaded?
matteo: yes
matteo: scrolling in konsole or web pages is slow
matteo: not as in vesa mode, but i can see the lines drawing
airlied: matteo: wierd.. maybe try EXA.. does -ati do the same thing?
matteo: i'll boot the laptop, to check if drm mod is loaded
matteo: yes i use shadowfb or EXA
matteo: i never use XAA
matteo: but now i'm not sure I load the kernel module
matteo: i'll poweron and check
matteo: intel 915 here :P
matteo: airlied: is kernel mode setting a planned feature?
matteo: yes, module loaded
matteo: but unused
matteo: airlied: why the module is unused?
matteo: woah, I added Option "DRI"
matteo: it's damn fast now with EXA
airlied: matteo: a plannet feature for what?
matteo: for the radeonhd driver
airlied: matteo: planned even.. it'll go in -ati definitely.
airlied: its up to radeonhd if they want to use it or not.. I'm guessing not since they can't fork it.
matteo: fork?
airlied: well there won't be two drivers for kms..
MostAwesomeDude: airlied: Speakin' of which, how many hoops do I have to jump through for r5xx KMS, and what's missing from it?
airlied: so userspace can use what it wants..
airlied: MostAwesomeDude: you build it insmod radeon modeset=1 and should get a framebuffer
airlied: the DDX kinda works but there is no 3D driver yet
MostAwesomeDude: airlied: 'k. Is EXA going to be user-side or kernel-side?
airlied: MostAwesomeDude: accel always stay on the userspac side
airlied: anything that requires using registers directly goes in the kernel
airlied: anything that can use IBs or cmdbufs in userspace
MostAwesomeDude: airlied: Okay, so the EXA stuff shouldn't need much changing, then?
airlied: MostAwesomeDude: it requires porting to a memory manager
MostAwesomeDude: Ah.
MostAwesomeDude: Joy....
airlied: and a new command submission routine for lockup avoidance stuff
airlied: but the code shouldn't need many changes.
airlied: just new hooks.
matteo: so it needs kernel support too?
matteo: it's not just a DDX feature?
airlied: memory manager needs kernel support
airlied: as its a kernel memory manager
GerbilSoft: just switched from radeon to radeonhd - 2D seems faster, but any sort of compositing screws up royally
GerbilSoft: ..and now i'm noticing font corruption. yay