Home Phoronix Phoronix Forums X.Org Videos From FOSDEM 2008

Radeon IRC Logs For 2009-12-19

Search This Log:


spstarr: [22767.972076] sysprof: loaded (1.1.5)
spstarr: great
spstarr: hmm
spstarr: Could not open performance counter: Function not implemented
spstarr: i wonder if im missing something
spstarr: oh
spstarr: CONFIG_PERF_COUNTER not set
airlied: MostAwesomeDude: from TTM pov shuodl be the same as r100, from CP mostly the same, different microcode,
airlied: output setup is quite similiar
MostAwesomeDude: airlied: Yeah, I remember thinking that there wasn't that much code involved.
airlied: I think 2D hw engine is the same
airlied: not tex rect so not much of EXA composite can be done
MostAwesomeDude: Is UMS r128 already EXA-enabled?
MostAwesomeDude: I should find a third monitor.
cxo: people are always throwing crts aways on kijiji or craigslist
MostAwesomeDude: While I'm at it, I should try to nail an r500 that doesn't have these goddamn CP bugs.
MostAwesomeDude: Ooh, I could always surprise my parents for asking for a monitor for Christmas.
MostAwesomeDude: When one hasn't had a list for the past decade, one's parents become very eager to buy things for oneself.
spstarr: oh PREF_EVENTS is new name
spstarr: looks like sysprof module needs update for .32+
airlied: MostAwesomeDude: I think it has some EXA done
MostAwesomeDude: airlied: Hm. Mixed pixmaps would surely help, if it could do the TTM thing.
MostAwesomeDude: I'll look into it. I mean, that's really the oldest chipset we want to have in the KMS set, right?
airlied: unless someones wants to do mach64
MostAwesomeDude: :C
cxo: what is with all the sci-fi movies this christmas
spstarr: if anyone wants to use sysprof you need to change the following code:
spstarr: collector.c:
spstarr: +/*
spstarr: #ifndef __NR_perf_counter_open
spstarr: #define __NR_perf_counter_open 336
spstarr: #endif
spstarr: +*/
spstarr: attr->size = sizeof(*attr);
spstarr: - return syscall (__NR_perf_counter_open, attr, pid, cpu, group_fd, flags);
spstarr: + return syscall (__NR_perf_event_open, attr, pid, cpu, group_fd, flags);
spstarr: use the new syscall name it seems to work
spstarr: with 2.6.32+
spstarr: profiles mesa+second life
spstarr: i suspect this is going to be a *HUGE HUGE HUGE* log file generated if it doesn't crash sysprof in a few hours
soreau: spstarr: What is so difficult about submitting a patch or using a pastebin for goodness sakes
spstarr: soreau: its for people searching on google :)
soreau: Flooding the channel just makes a mess and gets lost, effectively accomplishing nothing
spstarr: this channel is logged and you can find good tidbits on the log
soreau: Yea, like weeks later
soreau: In any event, that is not the correct way to go about things
soreau: it's just plain annoying
soreau: *sigh* I guess you couldn't expect anything less from a second-life user ;)
soreau: and the only one complaining about egg cooking temperatures
spstarr: :P
spstarr: over 3 million samples so far, oh boy..
spstarr: I am very curious to see the results of this profile
spstarr: stops the profile
spstarr: 9.6 million samples
spstarr: interesting
spstarr: airlied: drm_clflush_pages == self: 13.99% total: 13.99%
spstarr: its getting hit hard
spstarr: but i note this is with the intel i965_dri i will test with radeon today
spstarr: airlied: I also will be testing irq stalls today with sysprof, this should help you figure out why im getting stalls
spstarr: it almost seems drm_clflush_pages initially is quite low in total
spstarr: but after a 9.5 million samples goes to such a high % something seems interesting
gimzo: mesa doesn't build for me since a few days ago, says: gmake[3]: ../../../../../src/glsl/apps/compile: Command not found
AndrewR: gimzo, autogen.sh helps for me ...
gimzo: AndrewR, thanks
DanaG: http://anandtech.com/weblog/showpost.aspx?i=662
DanaG: hmm, DisplayPort audio... is that a known to-do for radeon?
airlied: I think its similiar to HDMI audio, don't have an DP audio hw to test
DanaG: Nor do I.
DanaG: I just have hdmi.
gimzo: on gallium glxgears locks up and reboots the system
MostAwesomeDude: It's a special technique. :3
airlied: sounds like r300g = i965g ;-)
airlied: looking at i965g makes me realise how little of 965 I understand
MostAwesomeDude: gimzo: Which chipset?
gimzo: r400
MostAwesomeDude: airlied: I think i965g's around that same place as r600g right now. :3
airlied: MostAwesomeDude: no its actually uses thehw
MostAwesomeDude: gimzo: Which r400?
gimzo: RV410 [Radeon X700 Pro (PCIE)]
MostAwesomeDude: airlied: Ah. TBF you can run softpipe on r600+ right now.
MostAwesomeDude: gimzo: Oh, you've got the same one as me. Wonder why you're having issues...
gimzo: MostAwesomeDude, it worked ok a few weeks ago when I last checked
MostAwesomeDude: gimzo: It worked a few hours ago when I committed.
gimzo: It didn't work ok a few days ago, I just thought it's temporary so I didn't say anything
MostAwesomeDude: Hm. You've got all the new, fancy stuff?
MostAwesomeDude: Does glxinfo work?
gimzo: yes
gimzo: but anything that actually tries to draw anything locks up
MostAwesomeDude: That's very bizarre.
gimzo: what new stuff? I have xf86-video-ati a few weeks old and kernel 2.6.31.6-166.fc12.x86_64
MostAwesomeDude: Sounds like you're all prepped. Does RADEON_SOFTPIPE=1 work?
gimzo: let's try (if I dissappear then it doesn't)
gimzo: MostAwesomeDude, softpipe works
MostAwesomeDude: gimzo: Good. Does r300 classic work?
gimzo: yes
MostAwesomeDude: Hmm.
MostAwesomeDude: Well, does progs/trivial/clear work? If so, then we might be onto something; if not, then I'm stumped.
gimzo: it works
MostAwesomeDude: Hmmmmmmm.
MostAwesomeDude: Clearly, I need to test 64-bit more often.
MostAwesomeDude: If you can find anything in progs/trivial that locks up, then I'll take a look at it.
MostAwesomeDude: I bet there's some little thing that's unsigned when it should be uint32_t.
gimzo: ok, on it
eosie: MostAwesomeDude: I am running on 64bit all the time, and unsigned is of the same size on both architectures
eosie: *long* isn't
Pallokala: I wonder if this is a good message: "r600_cp: Failed to load firmware "radeon/R600_rlc.bin""
Pallokala: comes up during booting 2.6.33-rc1
Pallokala: rv630 as chip here
chithead: you need to install firmware. if you have radeon drm compiled in, firmware needs to be in initramfs and/or built into the kernel
Pallokala: the other firmware needed is there (RV630_pfp.bin and RV630_me.bin)
Pallokala: and the kernel does not carry the rlc-file
chithead: indeed, kernel does not accept any new firmware
chithead: you have to install it via your distro's packages, or download it manually if your distro is too slow
Pallokala: yes, thanks
Pallokala: I think I will manage from here.
gimzo: MostAwesomeDude, only tri-fbo-tex-mip locks up
gimzo: all others work (at least don't lock up, haven't checked if output is correct)
gimzo: MostAwesomeDude, openarena locks up after intro animation, xbmc crashes, but doesn't lock up the system
rcbarnes: I was wondering if anyone has expeience with the radeon 4350 and the radeonhd driver, particularly compiz, xcompmgr, second life, and general openGL linux apps.
rcbarnes: I want to drop my nvidia binary-driver card, but I don't want to lose any significant functionality.
Zajec: radeon and radeonhd are DDX drivers
Zajec: 2D, modesetting, Xv
Zajec: and you mentioned 3D applications which needs 3D, so on DDX driver
Zajec: 3D driver for new ATI GPUs i called r600 and is part of Mesa
Zajec: it runs Compiz, KDE4 effects, openarena really fine
Zajec: but more advanced games and Wine are "not yet"
Zajec: maybe you will find something on http://wiki.x.org/wiki/RadeonProgram
rcbarnes: Zajec, what's the stability situation? Can I depend on rock-solid driver execution for several weeks between reboots?
darkbasic: rcbarnes: I'm doing a badblocks check by 24 hours, using mesa git master (glsl enabledf) , libdrm git master, xf86-video-ati git master and linux 2.6.32 + radeon-drm-testing and playing fretsonfire sometimes
rcbarnes: darkbasic: what chipset does your card have?
darkbasic: rcbarnes: 3870
darkbasic: rcbarnes: rv670
rcbarnes: darkbasic: and you get complete stability with that driver?
darkbasic: rcbarnes: only with latest driver snapshots, but this may change in future
rcbarnes: Hm. Looks like I've more waiting to do before switching cards, then.
Zajec: rcbarnes: i experience random locks up sometime
Zajec: rcbarnes: once a week
rcbarnes: Good to know. Well, thanks for the info, darkbasic & Zajec.
Zajec: rcbarnes: will work on that lock up later :)
Zajec: so should be fixed not so long :)
adamk_: OK, I'm back.. And
adamk_: And sl_pp_context_create is still undefined.
adamk_: Running strings on src/glsl/pp/libglslpp.a as cxo suggested does show sl_pp_context_create... Five times, to be precise.
hnsr: holy bananas, OpenGL 2.x/GLSL by default now with r[67]00 ?
adamk_: It's supposed to be, yes.
hnsr: updates at once
phoenix64: are there any plans for r600g btw?
hnsr: sweet it works, though at 0.2 fps :D
adamk_: So how can I disable the GLSL code at compilation time?
adamk_: So dmesg shows that R700_rlc.bin is being loaded, and I'm using the latest Mesa from git (with 2.0 working on this particular installation), but it still says IRQ's are not enabled...
adamk_: Hmm... Maybe I should update libdrm.
adamk_: Well that didn't do it.
Zajec: hm, maybe Mesa rebuild? don't know
adamk_: I just rebuilt it this morning, that's why I have opengl 2.0 now.
rehabdoll: whats the cmd-line option to boot without kms again? i get a black screen with .2
rehabdoll: hd4770
adamk_: nomodeset
rehabdoll: thanks
Zajec: adamk: do you use KMS?
adamk_: I am on this machine (where I'm getting opengl 2.0).
nessundorma: adamk_, mesa uses libdrm headers, I think, so you should recompile mesa everytime you recompile libdrm
adamk_: Well recompiling Mesa is easy enough...
adamk_: OK, so I updated libdrm, reinstalled it. Then recompiled Mesa. Still get the IRQ's not enabled message.
rehabdoll: humpf, kms is broken for me with 2.6.32.2, boots fine with it disabled. dunno what happens with it enabled, no signal to screen and nothing in the logs. Seems no daemons are ever started
adamk_: rehabdoll: Is your kernel built with fbcon support?
edgecase: the thing is i'm using 2D, no compiz, and it's terrible show
edgecase: slow
rehabdoll: yes, same config as 32.1
adamk_: edgecase: Pastebin your /var/log/Xorg.0.log file.
adamk_: rehabdoll: Sorry, that was my one idea.
rehabdoll: if i was missing fbcon, X should atleast start
rehabdoll: i dont know if the kernel panics, or what
adamk_: Time to setup a serial console or netconsole :-)
edgecase: ok updating stuff, will paste in a minute
osiris: MrCooper: I've tested the gearbox on rv380 and it works (I've just pushed two patches for blit, didn't check before them)
edgecase: adamk_ http://70.47.139.174/Xorg.0.log
adamk_: Well that all looks normal to me, but I haven't used a first generation radeon in a while.
Zajec: adamk: do you use KMS?
edgecase: adamk_, it seems like certain things have a delay, of almost a second
adamk_: Zajec: Didn't we go over this? :-) I am using KMS on this particular setup (where IRQs aren't working).
Zajec: ok, sry, missed that :)
Zajec: adamk: did you post dmesg?
adamk_: I did not, but I will in a minute.
edgecase: adamk_, any idea what I should look for? Xorg using cpu, apps using cpu, ? what likely config options can I change?
edgecase: I found that XAA had drawing glitches, so i'm using EXA for example
adamk_: Zajec: http://pastebin.com/m2a94be6c
adamk_: edgecase: Sorry, I really don't know. All I can tell from the log file is that 2D and 3D acceleration is enabled.
Zajec: adamk: looks fine :/
darkbasic: has someone problems applying today's drm-next patches to linux 2.6.32?
BioTube: darkbasic: I tried last night to no avail
BioTube: drm-next will go on top of 2.6.32, but pulling 2.6.32.2 on top of that ruins it
darkbasic: BioTube: what about 2.6.32 + drm-radeon-testing?
darkbasic: *2.6.32.2
edt: anyone here running kde4 and the latest mesa? If so are you getting strange icons (a document with a ? on it) in place of the correct icons. It happening here with one desktop icon (1 of 5) and with 5 icons on the task bar?
jinzo: hello, I'm running Arch Linux with radeon driver on a HD4570 (r600 if I'm not mistaken ) and I'm getting some weird problems
BioTube: darkbasic: same thing
jinzo: I don't know if they're related to the driver, but first thing that poped out was the KMS
jinzo: as KMS can't be set on this card yet, I'm wondering if I should disable KMS explicitly, or can I leave it like that ?
edt: kernel is 2.6.32.2 + drm-radeon-testing with conflicts use the testing version of the files
BioTube: jinzo: 2.6.32 stock supports r600 KMS
edt: jinzo what makes you say KMS cannot be set?
jinzo: also, is there a known bug ( I searched, didn't found anything proper tho ) that the driver crashes, when user hits the [enter] key ( any app, from OOo to FF )
jinzo: edt, I'm still on 2.6.31, so I'll need to upgrade if I'll wish to have that
edt: or merge drm-next or drm-radeon-testing
jinzo: but from the information I got ( arch wiki / gentoo wiki ) I'm not sure, if I should disable KMS explicitly or not
edt: which is what I did while on .31
jinzo: and it seems, that it only crashes when the system is running for some time, but it's really weird, quite constant crashes then, triggered with [enter] button
jinzo: also, I read that the radeonHD driver is "deprecated" in favor to radeon driver developement, but couldn't find any additional info on this
edt: jinzo I have an r500 (hd3300 via 790gx chipset) and use xf86-video-ati lots and lots of development is happening on this driver for KMS/3D.
edt: for r500-r600
jinzo: cool
jinzo: I'm looking for a stable driver
edt: I also run arch on a couple of boxes. One is a ascer netboot which is using .32 without problems
jinzo: I can leave without advanced stuff (eg. composition )
jinzo: edt, .32 from arch repos or custom built ?
jinzo: ( I still need to figure out how packaging exactly works in arch =
edt: first from repos then from aur without changing the options set by the build
edt: from repos suspend did not work. It does with the aur kernel
jinzo: cool.
jinzo: I'll stick to the repos, don't need suspend that much.
edt: the acer uses 3d with a intel 915 chip. The other box is using .32 too with an r300
jinzo: Also, I don't like AUR too much, and I think I missed an option to install the packages so they'll be updated thru pacman automagicially
edt: I would go for .32.2 with an up-to-date mesa 7.6 (switch to 7.7 when it gets out of rc) with xorg 1.7
edt: and think you would do ok
jinzo: ok, thanks for the advice
jinzo: also, how's with 64 bit vs 32 bit support ? any big differences/cavetas ?
arsenm: jinzo: you can use yaourt to do that
edt: this box runs gentoo with git libdrm, mesa, xf86-video-ati and is usually quite stable - though I do not stress 3D much (I use kwin and run a few test games with each update...)
edt: this box is 64 bit, the other two are 32
jinzo: arsenm, woha, thanks
jinzo: I'm thinking of going gentoo too, and a bit more bleeding edge software, but first I'll play with arch a bit
jinzo: it's nice to have packages installed so quick for a change :D
gatzby: Hi I am using the radeon hd 3200 in my laptop, and am trying to connect it to an external tv. I tried the VGA port and the HDMI port. On both attempts, when I go to System -> Administration -> Display the tv is seemingly correctly detected, but no picture comes up. Is there something I need to do to get TV-out working?
chithead: gatzby: you may need to use kms and pass atomtvout option to the radeon kernel module
gatzby: is there a page you know of that describes how to do that offhand? if not I'll just google it
chithead: "modinfo radeon"
gatzby: thank you
darkbasic: does someone uses gentoo? why today's mesa ebuild defaults with +gallium use flag?
chithead: gatzby: for pre-kms setup, read http://www.x.org/wiki/radeonTV
chithead: darkbasic: gentoo specific issues best discuss in #gentoo-desktop
darkbasic: chithead: a more generic question: if I enable gallium in mesa, is it possible to choose which driver to use (the gallium one or not)?
darkbasic: *the gallium one or the classic one
chithead: I think gallium will be used if enabled
Zajec2: osiris: ping
osiris: Zajec2: pong
Zajec2: osiris: i get same issue with bloom on r600
Zajec2: osiris: are you aware of that? could you try to fix this maybe?
osiris: Zajec2: openarena?
Zajec2: osiris: yes
Zajec2: i refer to some blit work and http://bugs.freedesktop.org/show_bug.cgi?id=19366
osiris: Zajec2: well, for r600 major part of the r300_blit.c file would have to be rewritten
jinzo: goddamn, laptop freezed again
jinzo: so, I updated it to .32 kernel and never radeon drivers ( damn, KSM is sweet )
osiris: Zajec2: and since I don't know r600 GPUs at all yet, it will be pretty hard for me to do it
jinzo: and I'm getting some errors when I try to run glxgears, I googled those errors and found out that appreantly I need a newer mesa
jinzo: what mesa version do I need ?
Zajec2: osiris: ok, thanks for info
jinzo: also, the freeze is happening after I hit the return key, woot.
jinzo: error: http://pastebin.ca/1720441
Zajec2: osiris: do you think Wine and Sims 3 may be using blit?
Zajec2: osiris: i tried Sims 3 in WIne and that was veeery slow and with broken rendering
Zajec2: i wonder if that slowness can be cause by blits?
osiris: Zajec2: probably
Zajec2: hm
Zajec2: good I at least know issue :)
agd5f: MostAwesomeDude: http://cgit.freedesktop.org/~agd5f/xf86-video-ati/log/?h=r128-support
osiris: airlied: any suggestions how to modify the common radeon texture code to enable tiling? maybe new function for vtable that would store the texture data, and for r300 we would use the 3d blit
jinzo: let me ask in here too, I found out this log: http://people.freedesktop.org/~cbrill/dri-log/index.php?page=dri-devel-2009-11-19.log , the user has the same problem as I do, now I'm trying to figure out, what's "newer" mesa
jinzo: how new exactly, if anyone knows ?
Ex-Cyber: should xf86-video-ati 6.12.4 be able to work with R600 KMS, or does that still need git?
Zajec: I can see radeonhd treats (usMinEngineClockPLL_Output / 2) as minimum engine clock (and later it checks PowerPlayInfo table
Zajec: is that something... safe?
Zajec: worth using in KMS?
MostAwesomeDude: agd5f: You rock.
agd5f: Zajec: I'm working on a patch to pull in the power states from the power tables
agd5f: Ex-Cyber: you need git
Zajec: agd5f: do you have most done? will it take you much more time?
Zajec: agd5f: should I start this and maybe finish before you? ;)
agd5f: Zajec: another day or so I guess
Zajec: ah, ok :)
Zajec: so I'll drop my idea of porting that :)
Zajec: well, then I'm boring again :/
agd5f: Zajec: I have the table defs for r6xx+ as well
Zajec: agd5f: radeonhd has that RE i believe
Zajec: unless you has that magical flags :)
Zajec: that wasn't RE
agd5f: Zajec: might as well use the real thing :)
Zajec: agd5f: what about releasing HDMI IRQs (airlied?)?
agd5f: Zajec: I wrote that. It's awaiting IP review
Zajec: uh, how simple register noumber can be waiting IP :(
Zajec: i'd guess that faster than they review it :P
Zajec: agd5f: can you tell if it's single IRQ like "HDMI state changed" or detailed like "HDMI bits per sample changed" and similar?
agd5f: Zajec: the tricky part is that the audio chip is licensed, so some of the IP isn't ours
Zajec: agd5f: well, maybe... but what this patch actually contains except register and value enabling one more IRQ?
Zajec: agd5f: thanks I didn't choose law but IT ;)
agd5f: Zajec: the irq stuff is probably ok, but the rest is still needs review
jinzo: woha, using -git "stack" and stuff works great. Thanks guys.
Zajec: agd5f: well... do we need anything else than this IRQ?
Zajec: i thought Christian wrote... everything?
Zajec: what is BLIT in meaning of GPU? I don't think http://en.wikipedia.org/wiki/Blit_(computer_terminal) makes more sense to it
Zajec: and I see some "r300: minor blit rework"
Zajec: which I would love to see on r600
agd5f: Zajec: I started on that for r600, but got sidetracked
Zajec: agd5f: blits?
agd5f: Zajec: copying rectangles
agd5f: basically what exa copy or xv or bo moves do
Zajec: blit is just copying rectangles? it's that something like EXA operation?
Zajec: hehe ;)
agd5f: exactly the same
Zajec: agd5f: why currently it is so slow?
Zajec: don't we do this on hardware, or...?
agd5f: Zajec: it's software
agd5f: right now
Zajec: so is this just like changing one function to use what EXA does currently?
agd5f: you basically have to fill in the GL blit hooks with a hw implementation
agd5f: which no one has done yet
jinzo: I'm sorry if I'm repeating myself, but on gentoo wiki I read that radeonHD is "dead" and I should use radeon driver instead, is that true ?
MostAwesomeDude: jinzo: Close enough.
agd5f: Zajec: it's not used very often so it hasn't been a high priority
Zajec: agd5f: well, enabling Bloom in OpenArena slows my FPS to 1, just like normal Sims3
Zajec: agd5f: so I hope if blit fixes OpenArena it can fix Sims 3 as well ;)
agd5f: Zajec: right. some stuff uses it, but not that often
jinzo: MostAwesomeDude, thanks, so stuff from radeonHD is getting merged/redone in radeon ?
MostAwesomeDude: readpix is already accelerated thanks to TTM; drawpix needs blitter.
Zajec: agd5f: osiris said it may be used in Wine + Sims 3, so I really want to believe this for now :)
MostAwesomeDude: Gallium just grew a nifty new builtin blitter. :3
MostAwesomeDude: Guess I should put more time towards r600gl.
Zajec: jinzo: radeonhd developed some stuff that we mostly move to KMS ratrher than radeon
MostAwesomeDude: *r600g, even.
Zajec: jinzo: similary to what we do with radeon -> KMS :)
jinzo: cool, what's the ETA on compleating that ? ( roughly )
MostAwesomeDude: jinzo: Already done; only a few special cases are left.
taiu: readpixels is the same as always
jinzo: woha, cool.
Zajec: jinzo: most still in .33 hopefully
osiris: agd5f: well wine is using it alot, because glBlitFramebuffer is implemented on top of it.
Zajec: jinzo: I believe to have PM layout written and agd5f is writing/porting PowerPlayInfo parser
Zajec: jinzo: HDMI ported..
Zajec: jinzo: so 33 should be feature-full :)
jinzo: KMS is really neat, I didn't tought that it's soo good untill I tested it
Zajec: osiris: great, that makes me believe I'll get Sims 3 soon :)
jinzo: cool, PowerPlayInfo is all that's needed for Power management ?
Zajec: jinzo: no
mentor: Would it be better to have glblitframebuffer do something else?
Zajec: jinzo: we still need some little stuff
jinzo: ah, cool nevertheless
xAFFE: hi folks, I'm trying to get my tv out running, it works so far, btu the screen is very distorted with green and strange stripes. I tried to set tv_standard to pal or ntsc, but it does not work. any ideas?
jinzo: that's the only feature I'm in need right now ( still have to see, how the git sources are regarding stability )
edgecase: yikes openareana trashes things
edgecase: kill -9 now i have no mouse cursor
edgecase: ok vtswitch fixed it
mentor: Fullscreen 3D apps should run on a different vt/workspace
edgecase: can they?
edgecase: windowed after a minute i get
edgecase: [335933.214097] [drm:radeon_cs_parser_init] *ERROR* cs IB too big: 16523
edgecase: [335933.214107] [drm:radeon_cs_ioctl] *ERROR* Failed to initialize parser !
edgecase: solution is don't grab the purple object
mentor: edgecase: Dunno
mentor: edgecase: I suspect the answer is complicated
edgecase: cool it runs tho,
edgecase: and crashes with KMS don't leave you logging in via ssh to reboot :)
edgecase: just check dmesg, carry on your business
MichaelLong: I still experience this probleme here, https://bugzilla.redhat.com/show_bug.cgi?id=531825 on .33-rc1 any news on it in the meantime?
ghepeu: hello
ghepeu: osiris, are you there?
ghepeu: I see that you merged r300-blit into master
ghepeu: it works and it's fast, but I'm seeing glitches
osiris: ghepeu: hi
osiris: ghepeu: what app?
ghepeu: openarena, compiz with the gaussian blur enabled
ghepeu: http://img696.imageshack.us/img696/5359/schermatar.png
ghepeu: this is a semi-transparent window with the gaussian blur enabled
ghepeu: before it as ok, but unbeareably slow (1 fps or less)
ghepeu: now it seems it's not composing the window with the correct other windows
osiris: ghepeu: could you enable the if-zeroed printfs in r300-texcopy.c and r300-blit.c and provide me with the output?
osiris: ghepeu: create a bug report for it
ghepeu: ok, give me a few minutes
evil_core: IT would be hard to add pageflip for r500?
ghepeu: Ohm{}nis, I'm using KMS
osiris: ghepeu: no hurry, I'm not sure I'll have time to look at it today
ghepeu: if you're in my timezone (CET) I can understand why :)
osiris: yeap :)
ghepeu: how do you take a screenshot in openarena?
osiris: no idea
evil_core: /screenshot prboably like in quake ;)
rehabdoll: ghepeu: what font is that on your desktop?
ghepeu: the one in 'FreeNAS'? Comics, from the sharefonts package
rehabdoll: nice, must have missed it
rehabdoll: oh, nonfree :/
ghepeu: yes, the package was removed from my distro for that reason
basilgohar: How can I get an Xorg log file when the system doesn't get to the GDM login screen?
basilgohar: I am running Fedora 12, and I want to get it to work with my Radeon X1600 AGP, and it starts fine using KMS, but then when it comes to displaying the GDM login screen, it just stops, system locks...
basilgohar: And I booted with another video card, and there was no Xorg.0.log.old
tlp: Wow.
tlp: WoW works
tlp: ah, crap. expect it crashes
tlp: except*
Ex-Cyber: well, this is odd... I have KMS working pretty nicely now, but xorg still sets an invalid mode on my monitor by default ( https://bugs.freedesktop.org/show_bug.cgi?id=24419 )... the weird part is that console does seem to be using the "invalid" 75Hz mode just fine
twnqx: did i mention that with 2.6.32.1 + drm-radeon-testing resume doesn't work?
tlp: http://tlp.lickwid.net/wow-r600foss.jpg
tlp: unfortunately it crashes exactly there.
tlp: seems to render fine though
agd5f: Ex-Cyber: you should send that patch to the xorg ML
Ex-Cyber: agd5f: I'm not 100% convinced that it's the right thing at this point... it turns out that both KMS console and Windows can both do 75Hz (although they have rather different hsyncs; that's what I just checked)... still, I guess it's better to pick a suboptimal mode than a non-working one
Ex-Cyber: I'm not using that patch now since I updated to server 1.7; I'm back to specifying an equivalent modeline as preferredmode in xorg.conf
agd5f: Ex-Cyber: does the mode work ok with radeon git master or the latest drm bits? I recently reworked the pll algo and it seems to have helped a number of previously problematic monitors
adamk_: Which reminds me... I was going to test that for you by trying the native 1600x1200 resolution that caused flickering.
rnoland_: sigh... the mis-handling of drm handles is a pain
rnoland_: in order not to have to fix libdrm and DDX drivers... I implemented a hack to deal with it....
rnoland_: but the code is all much cleaner when handles are all of the right type.
rnoland_: lots of casting in libdrm etc is avoided....
ghepeu: osiris, if you're still here, this is what openarena prints out: http://pastebin.com/m25768937
ghepeu: and this is the output of compiz: http://pastebin.com/m3518c4a6
ghepeu: I'll open the bug in a few minutes
adamk_: agd5f: And it looks good. No flickering at 1600x1200 any more.
adamk_: I'll go ahead and close that bug.
Ex-Cyber: agd5f: I don't think I have latest drm bits, but I am running radeon master (well, I assume that's what the Gentoo x11 overlay 9999 ebuild grabs) and if I remove xorg.conf I don't get a display after starting X
adamk_: Is there some reason the console now shows a white border on larger monitors in areas that it doesn't access due to the fact that a smaller monitor is attached...
adamk_: Does that sentence even make sense?
Ex-Cyber: whereas if I have a modeline specifying the timings of the first detailed mode and specify that as preferredmode (since I don't know of any way to tell xorg to use that specific EDID mode), it does work fine
agd5f: adamk_: the console output is limited to the smaller of the two heads
adamk_: agd5f: Right, but previously the entire monitor would still be black, even the part on the larger monitor that was limited. Now the area "outside" is all white.
adamk_: I can grab a snapshot if you like.
agd5f: adamk_: weird. maybe something with the fbdev stuff
adamk_: I don't remember exactly when it changed... When I was tracking 2.6.31 + drm-next, it was definitely fine.
adamk_: http://adam.visualtech.com/DSC00290.JPG is what I'm referring to, btw.
adamk_: I see this on my computer at work, too.
adamk_: Not a huge deal, but definitely off-putting.
rnoland_: adamk: you shouldn't be running that linux crap. ;)
adamk_: Heh... I like testing bleeding edge stuff.
rnoland_: adamk: you can test that patch that i just posted..
agd5f: adamk_: maybe fbdev doesn't paint the fb
adamk_: rnoland_: Which reminds me, r600 from Mesa git doesn't work on FreeBSD any more.
rnoland_: sigh, what is broken now...
adamk_: rnoland_: sl_pp_context_create is undefined in r600_dri.so and swrast_dri.so
rnoland_: i just updated a couple of days ago...
adamk_: The X server has no issues loading r600_dri.so, so AIGLX is just fine.
adamk_: But direct rendering is broken.
Txt_file: how can I change the dpi of the xserver permanantly? I tried with "Option "DPI" "n x n"" in xorg.conf and "xrandr --dpi x" but both won't work correctly.
adamk_: rnoland_: I think this is since the GLSL code was turned on by default.
rnoland_: so, build is fine, but load is broken?
adamk_: agd5f: No not likely a problem with KMS, but a change in the fbcon code?
adamk_: rnoland_: Yep.
adamk_: rnoland_: Tried i386 and amd64, too.
rnoland_: ok, updating now....
adamk_: bbiab
agd5f: adamk_: probably related to glsl-pp-rework-2 branch that got merged a few days ago
gellmar: hi
gellmar: reinstalled the radeon driver according to wiki
gellmar: glxinfo shows the correct renderer but no DRI
gellmar: glxgears give 2100 fps instead of 3600
gellmar: what can be wrong?
gellmar: no error msg in xorg.0.log and dmesg
mentor: Gosh A minute and a half
mentor: Patient fellow
mattst88: I was going to answer too.
mattst88: people like that don't deserve help.
mattst88: oh, he's back.
gellmar: tried setting 0666 in dri
mattst88: run 'LIBGL_DEBUG=verbose glxinfo' and check the first couple lines for errors
mentor: gellmar: What makes oyu think 3600 is the correct figure?
gellmar: I checked it yesterday
gellmar: with Yes in DRI
gellmar: and R600 as renderer
gellmar: but the driver version was not the last one...
mattst88: is he responding to me, or did he ignore my suggestion?
mentor: Unknown
gellmar: now...
gellmar: mattst88: the same output as for ordinary glxinfo
mattst88: hmm. :\
mattst88: even if DRI is working, you should get extra lines like 'libGL: OpenDriver: trying /usr/lib64/dri/tls/r300_dri.so'
gellmar: http://pastebin.com/m653b07d9
mattst88: 3rd line.
gellmar: mattst: this is an output with LIBGL_DEBUG=verbose
mattst88: it doesn't seem to think so.
gellmar: that;s why I am asking
mattst88: are you sure there weren't any typos? try it one more time.
gellmar: look at the command
gellmar: LIBGL_DEBUG=verbose; glxinfo
gellmar: right?
Droste: without ";"
gellmar: ok...
gellmar: will paste an output
gellmar: aha...
gellmar: thanks...
mattst88: :)
gellmar: that ; gave me incorrect output
gellmar: and I could not understand why
gellmar: setenv LIBGL_DEBUG=verbose did not work...
Wizzup: export
mattst88: `LIBGL_DEBUG=verbose glxinfo` only sets LIBGL_DEBUG to verbose for that one command, and nothing more.
mattst88: I don't know where the ; came from. I gave you the command without it.
gellmar: it was my typo
gellmar: :(
gellmar: decided not to copy
gellmar: but to type
gellmar: and one more question
gellmar: in wiki, mesa and radeon are configured to /opt/xorg
gellmar: and what about libdrm and xorg-macros?
mattst88: no idea. I use Gentoo's live ebuilds. They handle all that.
gellmar: okay...
mattst88: got to run. best of luck
gellmar: I am simply writing a script for automated install
gellmar: and once I broke X today debugging that ;)
Droste: which distro?
gellmar: ubuntu 9.04 x86_64 with 2.6.32 custom kernel
Droste: hm ok, i don't know :-) in opensuse it's all in /usr, but i don't think this will help you :-D
adamk_: gellmar: Without the output of 'LIBGL_DEBUG=verbose glxinfo' there's not much we can tell you about your direct rendering issues.
gellmar: and that typo made me crazy :)
Droste: I think he already figured out whats wrong :-)
gellmar: yes...
gellmar: load misconfiguration
gellmar: caused by the rests of yesterday's install
gellmar: now will reboot and report a success :)
gellmar: success!!!!
gellmar: :)
gellmar: 3800 fps
gellmar: the only problem the desktop manager reports it can not turn the effects on
Eddi|zuHause: hm... i gave the radeon driver another try on opensuse 11.2, but when starting civilization 4 via wine i get "r300VertexProgUpdateParams:Params exhausted"
gellmar: with dri off it worked
gellmar: but anyway thanks - opengl is more necessary for me than that effects
adamk_: gellmar: compiz? Or the KDE desktop effects?
gellmar: compiz
gellmar: found a tip for compizconfig
gellmar: dunno if it works
adamk_: gellmar: Bring up a terminal, run 'compiz --replace &' and pastebin the output.
gellmar: no whitelisted driver found
_Groo_: adamk_: do you know if kde devs fixed kwin for 4.4 beta 2? radeon its useless witn 4.4 beta 1
gellmar: will now add to whitelst
adamk_: _Groo_: No idea.
_Groo_: :P
_Groo_: as usual the radeon devs dont use kde lol
adamk_: What makes you say that?
Droste: I'm using radeon and kde 4.4 beta1 right now and it works. or do you mean the effects?
fredrikh: _Groo_: it is fixed in beta2
_Groo_: Droste: i have an evil rs485 (aka 200m), and beta 1 cant activate any compositing.. worked like a charm in 4.3.x
_Groo_: fredrikh: good to know fredrikh, do you use kubuntu by any chance? any way to test it with a ppa or so?
fredrikh: in addition opengl compositing is enabled by default for r600/r700 on a fresh install, if you have mesa 7.7
Droste: ah ok, I don't use the effects and never tried :-)
fredrikh: no, regular debian
_Groo_: fredrikh: im using mesa 7.8 with latest dri2/ddx/kms/mesa
_Groo_: Droste: you should, kwin is very good
Droste: _Groo_ but my hardware is not :-D
_Groo_: Droste: i have an evil 200m, your hardware cant be any worse then mine
Droste: I think my old cpu and low ram are the real problems :-)
_Groo_: Droste: its gpu accelerated, give it a try
_Groo_: fredrikh: is klipper fixed too? he doesnt show me any pasted text since i upgraded to 4.4
Droste: ok, no compositing with kde 4.4 beta 1 :-D I will try it with beta2 :-)
_Groo_: Droste: yeah beta 1 broke radeon real bad
gellmar: is here
gellmar: Checking for Xgl: not present.
gellmar: xset q doesn't reveal the location of the log file. Using fallback /var/log/Xorg.0.log
gellmar: No whitelisted driver found
gellmar: aborting and using fallback: /usr/bin/metacity
stikonas_: Droste: it is enough to recompile kwin from trunk
stikonas: if you want to get working desktop effects with beta1
_Groo_: stikonas: can you post a svn link, i can try it, missing 3d a lot
Droste: stikonas_: I'm waiting for the opensuse packagers to put it in the factory repository :-) I don't like compiling my stuff.
stikonas: http://websvn.kde.org/tags/KDE/4.3.85/kdebase/workspace/kwin/
Droste: (only doing it with radeon drivers, because no kms sucks :-D)
stikonas: _Groo_ you have to checkout http://websvn.kde.org/tags/KDE/4.3.85/kdebase/workspace in order to compile
_Groo_: thanks stikones, gonna compile it asap.. btw was it a checking bug? aka, wrong radeon opengl parsing detection from kwin?
_Groo_: stikonas: k
stikonas: I think command would be svn co http://anonsvn.kde.org/home/kde/tags/KDE/4.3.85/kdebase/workspace
_Groo_: stikonas: thanks
airlied: adamk_: I think it was black by luck, I need to explicitly make it black ;-)
airlied: osiris: yeah probably some vtbl entries needed to do tex tiling
fredrikh: _Groo_: it works for me, but i never noticed that it was broken
adamk_: Ahhh... yeah, I think black just looks better in this case. The white is kind of glaring.
_Groo_: fredrikh: it is in 4.4 beta 1 with mesa 7.7/7.8
fredrikh: klipper doesn't have anything to do with mesa though :)
_Groo_: fredrikh: ahh you were talking about klipper sorry
_Groo_: fredrikh: well in beta 1 , klipper is always empty no matter what i do, nevers shows the clipboard
Droste: funny, my klipper is copying everthing I highlight... thats also not really helpfull :-)
Droste: _Groo_: did you used the preview feature for the windowbar with radeon before kde4.4 beta1?
fredrikh: _Groo_: someone committed a "Fix nothing added to history when history was empty" 2 weeks ago, so if that's your bug, it's fixed
_Groo_: fredrikh: nice to know
_Groo_: Droste: preview feature for the windobar?
fredrikh: Droste: that's the default behavior, but this is getting a bit off topic for this channel ;)
Droste: yeah sorry :-) it's a little different from default behaviour... but you'r right, offtopic :-)
Droste: _Groo_: don't really know what it's called in the english version... previews of the windows of the programs opened in the bar at the bottom (/top/left/right wherever you have yours positioned)
_Groo_: Droste: ahhh i dont use it, i prefer the feature that shows the window when you mouse over it
Droste: yes :-) that's the feature I meant :-D last time I tried to use it (kde 4.2) it uses mainly cpu cycles not really accelerated. Is this better with current radeon?
Droste: because this really is a helpfull feature besides all the eyecandy :-)
Droste: _Groo_: ping
_Groo_: Droste: i dont know, i need to recompile kwin to test it in 4.4, ill tell you later :D but the show over mouse is much better, then preview
Droste: ok thanks :-)
_Groo_: Droste: try it, its very cool
Droste: I will
_Groo_: fredrikh: dependency hell, workspace need kdelibs, etc etc.. gonna wait for proper packages.. i dont have that much time to build beta 2 from scratch
gellmar: no...no way
gellmar: starting compiz without checks works, but cairodock doesn't
gellmar: any way to enable xgl with radeon
airlied: xgl has been dead for a long time
gellmar: did enable the composition extension to xorg.conf and put radeon to all whitelists...
gellmar: compiz claims no xgl,then no whitelisted driver found
[Enrico]: gellmar: that warning is harmless
gellmar: then reverts to metacity
[Enrico]: the xgl one i mean
gellmar: so the stuck is in whitelist?
[Enrico]: i guess
[Enrico]: i don't use compiz since kde 4.2
gellmar: and opengl cairo dock also works with black background
[Enrico]: but radeon should already be in the whitelist
[Enrico]: gellmar: here cairo-dock works like a charm
adamk_: gellmar: You can bypass the whitelist each time you start compiz by doing this: "mkdir -p ~/.config/compiz; echo SKIP_CHECKS=yes >> ~/.config/compiz/compiz-manager"
gellmar: i thought about it
gellmar: but I wonder if ogl applications will work with it
adamk_: If you have DRI2 working, yes.
gellmar: i can live without effects,but i want to use opengl for more than compiz
gellmar: there was a message dri2 is ignored
gellmar: setting i mean
[Enrico]: gellmar: are you sure you are using radeon driver and not fglrx ?
gellmar: but dri2 is loaded
gellmar: surf
gellmar: *sure
soreau: gellmar: Which distro?
gellmar: new system,no fglrx ever set
gellmar: ubuntu jaunty with 2.6.32 kernel
gellmar: 64bit
[Enrico]: gellmar: you might want to try karmic, more recent driver
soreau: You need to enable kms with radeon.modeset=1 to have dri2 working then
[Enrico]: [and kernel]
soreau: Oh yes, it will only work with karmic
soreau: jaunty is too old for dri2
soreau: Well, if using 2.6.32.. and you have latest libdrm, mesa and ddx compiled with kms support it should still work
gellmar: should I build a custom kernel for that?
[Enrico]: gellmar: no it should just work in karmic, but well you can install the .32 kernel from ppa if you wish
gellmar: mesa 7.7rc3
gellmar: kernel 2.6.32
gellmar: libdrm 2.4.16
gellmar: all from sources
soreau: dmesg|grep -i kms
gellmar: nothing
soreau: thats not right
soreau: grep for modeset, not kms
gellmar: okay...will check in 10min
amarks: nexuiz is not rendering correctly for me with git master, lots of black areas
gellmar: now on the phone
gellmar: btw, how to enable kms in radeon?
[Enrico]: gellmar: in the grub config in the kernel command add radeon.modeset=1
soreau: If its disabled by default in the kernel config, you would have to use that
[Enrico]: soreau: in ubuntu it is
soreau: I know
soreau: Thats why I said it
gellmar: and no modules other than kernel are required to get it working?
soreau: right
gellmar: will check it now
amarks: gl 2.0 shaders in nexuiz seem to have bad rendering on r700, if i disable it in nexuiz settings, works fine
eosie: MostAwesomeDude: I don't see how is your provoking vertex setup different from my patch that it deserved such rewrite
MostAwesomeDude: eosie: I'm not re-emitting the entire rasterizer state, and I'm also not modifying the rasterizer CSO.
eosie: right
eosie: MostAwesomeDude: concerning my texcompare patch, it's not being forgotten, is it?
MostAwesomeDude: eosie: No, it hasn't been forgotten. I'm out of town right now; you'll have to wait about eight hours.
cxo: "Don't look to be running Unigine Heaven on Linux with the ATI Radeon Mesa driver or any other Mesa3D drivers anytime soon."
cxo: suckage
bridgman: the vmware slides talked about some of the work that needs to be done in order to get mesa ready for GL 3.x
airlied: http://dri.freedesktop.org/wiki/MissingFunctionality
airlied: not the "piles and piles of work" bits
airlied: note
gimzo: is there an idea when can it be expected ?
bridgman: yeah, I think that's what vmware said too ;)
airlied: at the current dev rate I'd expect GL3.0 in 2-3 years, just in time to star tGL5.0
bridgman: everyone cried that GL was moving too slowly; now I wish it would slow down a bit ;)
rix_: I am having problems with my ATI graphic card while using OpenGl in virtual machine. Anybody has an idea ? :(
Ghworg: Is there any software that requires GL3 at all? (Apart from that one game)
airlied: not to mention how we can implment some parts of GL3 that are patented
rix_: Opengl is working in my host (ubuntu) but not in windows guest
rix_: Even windows live messenger is not working, it crashes
bridgman: yeah, I'm working with our Legal folks to see if there are any new options for dealing with patents in the open source code but no revelations yet
Ghworg: Set up an office somewhere that doesn't recognise sofware patents and do everything there ;-)
airlied: yeah I think the floating point textures thing is the messy one
rix_: AppName: msnmsgr.exe AppVer: 14.0.8089.726 AppStamp:4a6ce533 ModName: opengl32.dll
airlied: Ghworg: doesn't help
DanaG: And then there's that whole video DRM crap.
gimzo: gather a huge number of people to constantly bug governments to drop software patents ?
DanaG: Did you see the thing about how CRIA owes billions of dollars to artists?
bridgman: rix_; AFAIK 3D in virtual machines is very much a bleeding-edge thing, just starting to work and not in all virt systems yet
DanaG: That would be a good reason to get DRM stuff thrown out.
bridgman: that was nothing to do with DRM - they just didn't pay the bills (allegedly) ;)
bridgman: but what the heck, if it gets rid if DRM I'm all for it
DanaG: Still, it would make a good excuse for "why the hell should we do what restrictions you want, when you go screwing the artists anyway?"
airlied: yeah that works so well in life
bridgman: DRM wasn't to protect the artists, it was to protect the companies that owned the artist's work
bridgman: screwing artists is a time-honored tradition
rix_: bridgman: thanks for the answer
bridgman: rix_; best chance for success would be on a distro-specific forum, I think
bridgman: there isn't really one generic driver-level solution AFAIK
rix_: ok you are right, thanks
MostAwesomeDude: 09f9 and whatnot.
bridgman: 09f9 ?
MostAwesomeDude: First two bytes of the first AACS processing key.
bridgman: ah yes
bridgman: I was trying to relate it to virtualization or screwing artists
MostAwesomeDude: I think it applies to both.
MostAwesomeDude: Someday I'll assume a fake name and hax apart the Win32 chain of trust. Then we can have AES accelerated on our GPUs. :3
MostAwesomeDude: But there will be time for this later.
Trollinator: Does the page http://www.x.org/wiki/RadeonFeature refer to the current git or to the current stable release of radeon?
chithead: it refers to the latest code in git. not to any releases and especially not to what distros ship
Trollinator: perhaps it would make sense to mention that on the page so people like me don't waste your time asking ;)
soreau: I think by the nature of the page, one can assume it is the latest code. I doubt WIP would be happening in a stable release for example ;(
soreau: ;)
Trollinator: well, i wasn't talking about "one", i was talking about "peple like me" ;)
Trollinator: but anyway, i don't case.
Trollinator: care
Trollinator: thanks, bye
felipec: it seems my display size is being detected wrongly... xdpyinfo output is completely wrong, even though xrandr is ok
_Groo_: gotta go, seeya another day
EruditeHermit: soreau, hey
soreau: EruditeHermit: Hi?
EruditeHermit: soreau, how is it going?
edman007: has anyone gotten trem 1.2 to work? http://tremulous.net/ the old version works perfectly, the new version gives me a black screen with a white boarder, and the mouse is the same thing but smaller
soreau: EruditeHermit: You really don't want to know the answer to that question but for the sake of good conversation, it's going ;)
soreau: EruditeHermit: I saw you got a new hdd
soreau: EruditeHermit: How's that goin?
EruditeHermit: just reinstalling stuff
EruditeHermit: soreau, whats wrong?
soreau: jusr irl stuff
EruditeHermit: soreau, well perhaps the new year will be better
EruditeHermit: soreau, lousy year here too
soreau: EruditeHermit: Yea, maybe something about '2010' will inspire people. Without 'two thousand and --', 'twenty --' sounds much more futuristic :)
eosie: the end of the world just behind the corner
soreau: Well it's about time
eosie: r300g locks up in OA after a few minutes of playing with both current mesa master and 3-weeks-old master
eosie: and I swear it worked in kernel 2.6.31
eosie: s/in/with/