Home Phoronix Phoronix Forums X.Org Videos From FOSDEM 2008

Radeon IRC Logs For 2009-6-21

Search This Log:


rainbyte: hi
airlied: glisse: whats cpu ib flush?
airlied: don't need to force the cpu to flush caches, its either coherent or uncached
AndrewR: strange ...it is possible what kms-support broken for non-KMS on my motherboard? Because i can use ddx from this branch only in KMS mode, normal mode (say with old 2.6.27 kernel) hard lock my machine at X startup. With "DRI 0" and "AccelDFS 0" too.
AndrewR: master branch works fine ...
AndrewR: it was RenderAccel broken in kms-support branch. With "RenderAccel 0" this branch works in non-KMS mode too.
airlied: yeah its bound to have some divergance, am trying to align things so I can merge them
AndrewR: airlied, np, next reboot here, i want to see if my hang in KMS mode with AGP enabled was actually the same .... (reboot)
AndrewR: airlied, no, i just hit kernel BUG somewhere at linux/include/mm.h (?) and everything freezes ... (PCI mode works)
AndrewR: http://pastebin.ca/1468586 something like this. But i have nearly all kernel debugging turned on, so it may interfere (create this bug)
AndrewR: kernel from kernel.org, up to c277331d5fbaae5772ed19862feefa91f4e477d3 (mm: page_alloc: clear PG_locked before checking flags on free)
ernstp: cool, Xv and EXA seems to work OK with Radeon 4770, rv740, didn't know it was avaliable
ernstp: only issue is that the cursor is a bit corrupted
suokko: Where is ErrorF function/macro defined? (xf86-video-ati is using it for errors)
JyZyXEL: what are the package names for the radeon drivers in debian?
JyZyXEL: im interested if upgrading from stable to testing would give me newer version
JyZyXEL: and performance boost
predictable: Does anyone know how to install latest radeon drivers. Restricted doesn't list anything..
JyZyXEL: oh looks like i would have to go to unstable to gain any newer version
predictable: Anyone?
suokko: predictable: radeon driver isn't retricted
suokko: And you should have very near the latest stable if you are using spring release of your distro
predictable: suokko: well is there away to get newer drivers to support better 3d besides what auto. comes with ubuntu 9.04
suokko: predictable: There is some development drivers in launchpad but they aren't stable yet.
MostAwesomeDude: Why do people always think that we're trying to hide the "good" 3D?
nanoki: Because conspiracies are exciting?
MostAwesomeDude: I guess so.
nanoki: People see them everywhere there days.
MostAwesomeDude: I mean, the only conspiracy I can see is the Growling Stomach conspiracy.
suokko: Because people lost the 3D functionality with fglrx support ending
MostAwesomeDude: But I think that'll go away if I get some food.
nanoki: Sometimes they're right but the amount of misfires is so significant that net result is waste of time.
MostAwesomeDude: suokko: No, they didn't. They didn't have to do their system upgrades.
suokko: I know but someone forgot tell it to all U users who just saw notice in update-manager that newer version is there
MostAwesomeDude: It's totally possible to chill on an Ubuntu LTS or RHEL or Fedora 9 or last year's OpenSUSE and get systems that still receive security updates and have working fglrx.
nanoki: MAD: Bon appetit.
suokko: So U's fault not to notice all fglrx users about danger ;)
nanoki: They did notice them.
MostAwesomeDude: suokko: Not really. I don't care about fglrx beyond its ability to show me what a particular card can do.
nanoki: The end users are at fault for updating despite warnings.
predictable: no im not thinking anyone hides anything im new to linux. and I can't get my geforce 5500 fx card to work with resolutions higher than 640x480 so I put in an old also radeon 7000 card
MostAwesomeDude: Well, from our POV, they're at fault for not using open drivers. From AMD's POV, they're at fault for not using a supported distro.
predictable: Just wondering if theres newer drivers for it like hardware manager showed for geforce card, that quote never worked right... Does for my 9 series geforce but its down stairs hooked up to hdmi theater
MostAwesomeDude: Really, "OMG frxlg dont work on lfs/gentoo/arch" is just stupid. AMD doesn't support it, doesn't go through extra hoops to support it.
nanoki: True.
AbortRetryFail: wasn't fglrx created because some of their business customers demanded linux drivers?
MostAwesomeDude: predictable: For that old card, everything should work great.
MostAwesomeDude: AbortRetryFail: Largely.
mjr: predictable, radeon driver should work fine, no proprietary support for that card exists nowadays
AbortRetryFail: considering they were targeted at the FireGL line.
MostAwesomeDude: AbortRetryFail: FireGL and Radeon X driver. FGLRX. :3
nanoki: ARF: Likely, that's why they'll probably push its equivalent for Solaris soon enough.
AbortRetryFail: I actually use the damned thing, but i want to switch to Arch linux. fglrx won't work with the latest kernels and xorg
predictable: mjr: won't these work i heard they would. http://support.amd.com/us/gpudownload/linux/Pages/radeon_linux.aspx?type=2.4.1&product=2.4.1.3.39&lang=English
AbortRetryFail: ooh silly question: I know zaphod mode isn't really supported on the radeon driver, but can i get the same effect with two cards?
nanoki: doesn't personally see that big differences between GNU/Linux distros.
predictable: any also what drivers come with ubuntu is there a command I can type in to see what drivers im using now?
AbortRetryFail: nanoki: i use slackware right now. The difference between having to deal with dependencies and whatnot and having the distro do it for you is pretty big.
predictable: and
MostAwesomeDude: predictable: *Please* do not try fglrx with that card. Your chipset is in the r200 family; we have complete support for it, including full 3D.
mjr: predictable, #ati for the properietary driver (but no, they don't support older cards anymore with that)
MostAwesomeDude: Bending over backwards to make fglrx work will only end in pain and suffering.
predictable: ok
AbortRetryFail: r200 support rocks. My ex-girlfriend has a radeon 7500 that runs all kinds of native games and wine.
MostAwesomeDude: AbortRetryFail: Multicard is roughly as supported as it's going to get, from the DDX side. Xserver needs more work though.
nanoki: ARF: Dunno, Gentoo was mostly trivial for me. The issue was that the multilib implementation sucked but it's hard to find a GNU/Linux distro in which it wouldn't. :)
AbortRetryFail: yeah x86-64 multilib is a pain.
MostAwesomeDude: predictable: Quickest way is probably "LIBGL_DEBUG=verbose glxinfo". You should see "Mesa R200" or something similar in the GL_VERSION and GL_RENDERER strings.
MostAwesomeDude: nanoki: Gentoo's got the most painless multilib IME.
AbortRetryFail: Fedora seems pretty trouble-free as well.
suokko: predictable: yes. The default drivers for your card are the best possible. I remember trying fglrx when it still supported r200 cards but it was worse than mesa r200 driver :)
nanoki: MAD: With its pre-built category-based 32-bit compat packages?
AbortRetryFail: lol @ pre-built + gentoo
MostAwesomeDude: nanoki: The only two packages that don't work right with those are Flash and Sun Java.
nanoki: You have like a dozen semi-related programs that you have to install to get one.
nanoki: And to update one you need to update all.
MostAwesomeDude: But really, multilib is silly hax.
predictable: In that Libgl has an error in 2 places is that ok?
nanoki: Works on Windows? ;)
MostAwesomeDude: predictable: Could you pastebin it?
predictable: also I never saw r200 I saw like OpenGL renderer string: Mesa DRI Radeon 20061018 AGP 4x x86/MMX/SSE2 NO-TCL with 1.3 mesa 7.4
predictable: Mostawesomedude: sure
AbortRetryFail: no tcl?
MostAwesomeDude: nanoki: If you were going to do a true build-from-source multilib, you'd have to compile most of your system *twice* and every time you update you would need to build your updated packages twice too.
nanoki: Also I've heard Solaris has a nice multilib implementation too. Mostly GNU/Linux that can't do it sanely.
MostAwesomeDude: predictable: Oh, Radeon 7000 might be r100, not r200. I thought it was r200, though. Hm.
predictable: mostawesomedude: http://pastebin.com/m2b66eeac
suokko: I think r200 started with 8500
predictable: Yes I know its an old card, I know nvidia 5500 fx is old 2 but i know its faster. But the support for fx cards 5200 especiall 5500 suxors
suokko: and ended to <9500
nanoki: MAD: It used to be liked that, afaik mostly dropped because it created double package maintenance required.
predictable: I need to buy a new card for this linux machine, but ive had linux 3 or 4 days so im virtually new and learning
nanoki: Hmm, like that even.
predictable: <--- old machine for linux lol
predictable: 2.8 ghz standard core 1 gig memory
predictable: heh
MostAwesomeDude: predictable: Okay, yeah. Your 3D works fine. You're on an r100, which is no speed demon, but compiz should work.
predictable: compiz did work but some stuff was limited unlike on geforce card down stairs.. rain/water ect couldn't work and the animations such as airplane ect .. but its no biggie. Just learning with what I have
predictable: If I put linux back on the faster machine i have to go through hdmi for tv and over lay problems with tv and adding modlines ect..
suokko: Many compiz stuff requires newer features that aren't supported by radeon 7000
predictable: seems it would freeze from time to time. So i brought it to this old machine lol
predictable: suokko: the fire works though so it looks like rain should 2 lol thats what was oddly enough
MostAwesomeDude: predictable: Many of the compiz plugins aren't quite correctly marked for OGL version, so they won't behave right.
MostAwesomeDude: Just see what works and what doesn't.
predictable: mostawesomedude: I haven't been in the computer world really since I was a teenager so most knowlege has been lost.. ogl is open gl right? I remember my voodoo first 3d card was like glx or something.. What does radeon use
predictable: I know I sound terrible uninformed, but its been years really messing with a pc.
MostAwesomeDude: predictable: Yeah, GL.
MostAwesomeDude: Radeons aren't directly GL like older cards, but r100 is pretty simple to program.
MostAwesomeDude: r200 is a bit tricky, but that driver was written a while ago and is self-documented.
MostAwesomeDude: r300-r500 are complex little beasties though.
MostAwesomeDude: And r600+ are mostly shaders.
AbortRetryFail: it's all mircosoft's fault with directx :(
predictable: so all newer radeons use gl?
MostAwesomeDude: And now I'm ranting.
MostAwesomeDude: predictable: Well, we translate GL into chipset-specific setup.
predictable: I mean like back then it was a toss up open graphical library vs direct x and ogl always was better until newer direct..
zhasha: predictable: they use a universal setup
zhasha: both pipelines (D3D/OGL) are the same
predictable: kk
zhasha: a few minor differences aside
predictable: I feel like an idiot now a days lol
zhasha: and AbortRetryFail, it's not MS' fault
AbortRetryFail: yeah i know :)
AbortRetryFail: OpenGL didnt do what they wanted it to do back then. They actually supported OpenGL for a while.
MostAwesomeDude: WGL
MostAwesomeDude: They didn't really try to get rid of it until Vista.
zhasha: MostAwesomeDude: what makes you think they're trying to get rid of it?
predictable: all I know is ogl games were so easy to wall hack lol
mjr: well, they offered a lackluster opengl implementation on top of d3d on vista, and while they allowed the use of third party opengl drivers, that'd get rid of the desktop gimmicks
MostAwesomeDude: zhasha: They removed the fast paths for WGL and there's no longer any GL drivers shipped as part of the system.
zhasha: MostAwesomeDude: no, they added a slow path
mjr: (I think the latter has been ameliorated though, but I'm not sure)
zhasha: they didn't take away the fast paths
AbortRetryFail: I guess that sucks if you use GL programs on windows.
zhasha: the 3D driver can implement a native OpenGL implementation through opengl32.dll if they so please, however if they don't, you will have a slow OpenGL 1.4 implementation atop D3D
predictable: well on visita I use the newest direct x I didn't even know gl was still supported even in xp I had games all requiring direct x only older games q3 ect I think talks about ogl
AbortRetryFail: yeah they really push DirectX on the developer side.
predictable: I remember playing duke nukem with 3dfx was 3dfx like ogl?
AbortRetryFail: glide? that was it's own thing.
AbortRetryFail: only worked on 3dfx's cards.
zhasha: I think the pipeline in itself was pretty standard
zhasha: Carmacks engines so far I've seen are all OpenGL based
MostAwesomeDude: There is a tdfx driver in Mesa...
zhasha: it would suck nuts if MS cut it off completely
zhasha: They might support DX too now, I don't know
predictable: ok back
predictable: yeah 3dfx was glide wrappers or something
suokko: predictable: anyway: I got my keyboard back into together. Here is stable updates to video drivers: https://edge.launchpad.net/~ubuntu-x-swat/+archive/x-updates (If you realy want to try but there isn't much updates for r100 cards there)
MostAwesomeDude: radeon-rewrite provides DRI2 for r100, but other than that, there's been no change in that driver for years.
suokko: And if you want to try all the latest development version: https://launchpad.net/~xorg-edgers/+archive/radeon-kms You might have to use livecd to test because compatibility with Jaunty isn't the best
stikonas: suokko: what's wrong with KMS under jaunty?
suokko: stikonas: It is hard to get working so livecd iso might be easier option to test
predictable: suokko: so more or less I won't see any differ right?
suokko: stable updates shouldn't show any difference but radeon-kms could have big difference
predictable: I just wanted to try to get more fps or try to play that nexuiz kinda sloppy with this old card
suokko: But it all depends how well your card is supported with KMS
predictable: what is kms if I can ask?
stikonas: kernel modesetting
nanoki: zhasha: Hmm, having vendors ship WGL sounds sane, Microsoft devs might lack the know-how to do it properly anyway.
stikonas: predictable: but it of beta quallity now
predictable: so whats the differ. in those drivers than the newer stable ones?
stikonas: predictable: when kernel manages your video card it is possible to add more features to drivers but new code always introduces bugs
suokko: I don't know if radeon 7000 provides enough power to run nexuiz. I have to turn all settings down to minimum with my r200 card
nanoki: predictable: Part of moving some features from userspace to kernelspace, mostly.
nanoki: That is, KMS is that.
predictable: so is it basically updating the kernal to add more features than the software stand alone? Im not sure if I follow you guys are tech savy and im a noob idiot lol
predictable: or is it just adding new features to xorg.conf than what stand alone offers?
predictable: like differ. code
nanoki: Mostly re-organizing features, of thinks. The benefits are bigger though. As in, better power managemen and it works as part of DRI2.
nanoki: Me thinks even.
suokko: predictable: no. It is new way of writing video card drivers for linux. That should provide better performance for graphics demanding games
nanoki: Have the right stuff done in the right place and the end result works better.
predictable: suokko: k
predictable: suokko: be nice if I could figure out how to get past this geforce 5500fx low 640x480 reso in nvidia manger its a faster card than this one.. but that card dont seem to be supported worth a crap
suokko: predictable: You might try different versions of the closed source drivers
suokko: One of them might work with your card
suokko: You can download them directly from N
nanoki: I doubt an opensource ATi community is the best place to get information on an nVidia case with closed drivers. :)
predictable: suokko: well I tried manually installing older version and same thing.. then i got into errors about 2.4 and 2.6
suokko: I only own one ati card so I don't know about nv stuff
predictable: I could careless what card I use as long as ones faster than the other at the moment.. if this card was faster than geforce id stick with it but these are the options lol
genady12: hey
nanoki: predictable: Mostly closed driver on Linux are a bit of a pain and nVidia gives roughly zero support for the opensource effort...
genady12: can someone help me setting a proper resolution?
predictable: nanoki: yeah
predictable: well I have to run thx for info guys have a good day
genady12: nanoki, ?
nanoki: Oh, Gods. Army propaganda from television.
nanoki: s/from/on/
nanoki: genady12: He needed help with his Geforce. :(
genady12: I am trying to change resolution to 1920x1080 with 2nd display connected to ati 7200
genady12: s/7200/7500
suokko: suspects that having 2 display with one of them at HD is too much for that card
genady12: I need only one
genady12: its working with windows :(
nanoki: bridgman? ;)
bridgman: I can't help with NVidia either, sorry ;)
suokko: You should five some more background info so helping is easier: Which distro? Can you pastebin your Xorg.log file?
nanoki: Oh, right. If it's not a hardware thing, nothing then. :)
bridgman: actually I'm still reading backlog
bridgman: there's a lot of backlog
nanoki: bridgman: Mostly meant since you've traditionally done pretty good guesses at hardware limitations on ATi regarding resolutions. ^^
genady12lap: http://pastebin.com/m69469ff6
genady12: nanoki,
bridgman: reads the backlog again
nanoki: Sorry, I'm on a cell phone.
JamesLast: Hi, i have an radeon-modesetproblem.
bridgman: no worries, I'm mostly waiting for the coffee to kick in
suokko: genady12: (II) RADEON(0): Max desktop size set to 1600x1200
bridgman: at some point the words start to form sentences
suokko: That is because of some bad code in video-ati
genady12: (II) RADEON(0): Modeline "1920x1080"x0.0 138.50 1920 1968 2000 2080 1080 1083 1088 1111 +hsync -vsync (66.6 kHz)
JamesLast: Laptopscreen comes up with an whitescreen, after closing the lid and opening everything works
JamesLast: when i start with linus-kernel this workaround doesnt function, only with fedoras kernel.
suokko: brb, I'm going to look the code if there is easier fix than compiling custom ddx driver
JamesLast: bug-report with radeontool dumps is at http://bugs.freedesktop.org/show_bug.cgi?id=19459
genady12: suokko, are you talking with me?
suokko: yes
genady12: ok thanks!
suokko: genady12 You can set Virtual option in Xorg.conf to force ddx to use larger displayes than 1600x1200
suokko: http://pastebin.com/m759dc5d7 like this
suokko: Of course you need to change the resolution to 1920 1080 :)
genady12lap: lets try
suokko: http://pastebin.com/m2e6e056c __change_page_attr_set_clr seems like very expensive call with r200 KMS
bridgman: MostAwesomeDude; I always wondered what fglrx stood for ;)
suokko: That oprofile output was taken when I just resized gnome-terminal (I wanted to know what made Xorg took about 60 % CPU time while doing resize)
genady12: ok lets see if it helps
suokko: I as also reading the bug report about radeon performance problems if clearing the new window area was enabled. Isn't that clearing possible without reading anything from screen? I thought that maybe it could be done in GPU using shaders or possible just upload a new cleared pixmap/texture from main memory.
MostAwesomeDude: bridgman: Wait, you didn't know?
bridgman: you would be amazed what I don't know ;)
MostAwesomeDude: I mean, that's just what I assumed; there was one version that printed "FireGL/Radeon X" into the Xorg log.
suokko: (thatwas about problems caused by disabling: 107_fedora_dont_backfill_bg_none.patch)
genady12: suokko, ubuntu started failsafe thing
suokko: What does your Xorg.0.log say this time?
genady12: uploading
damjan: hi all, a quick question ... I want to see if updated drivers for ubuntu 9.04 will improve video on my hd3200 (currently the video stuters) .. so I herad there was some repository with faster updates for the driver
suokko: Here is stable updates to video drivers: https://edge.launchpad.net/~ubuntu-x-swat/+archive/x-updates
damjan: thanks
genady12lap: genady12@silent: ~ $ xrandr
genady12lap: Screen 0: minimum 320 x 200, current 1024 x 768, maximum 1920 x 1080
genady12lap: VGA-0 connected 1024x768+0+0 (normal left inverted right x axis y axis) 477mm x 268mm
genady12lap: 1920x1080 59.9 +
genady12lap: how do I set it?
suokko: System->Preferences->Display :)
genady12lap: only 1024x768 there
suokko: xrandr -s 1920x1080
genady12lap: Size 1920x1080 not found in available modes
suokko: Are you still in Failsafe mode?
genady12lap: no
suokko: I'm not good with xrandr but maybe man xrandr helps
suokko: Also Xorg.0.log could have some useful information
genady12: (II) RADEON(0): Max desktop size set to 1920x1080
damjan: xrandr --output VGA-0 --mode 1920x1080
genady12lap: suokko, xrandr --output VGA-0 --mode 1920x1080
genady12lap: worked
suokko: good :)
genady12lap: can I set it perepentaly ?
genady12lap: permanenty*
genady12lap: err permanently
suokko: http://wiki.debian.org/XStrikeForce/HowToRandR12
suokko: I think there might be way to do it
genady12lap: ~ $ glxgears
genady12lap: 333 frames in 5.0 seconds = 66.552 FPS
genady12lap: heh
suokko: What does glxinfo |grep renderer say?
suokko: (Is it still hw accerlated?)
genady12lap: yes
Ximik: мужики, есть кто русский?
Ximik: #ati
suokko: Funny experiment: Makeing all ttm ovjects uncached make performance about same that itwould be in 486 machine. It also causes corruption. (Don't try this at home)
nanoki: suokko: The performance bit sounds like to be expected. 0:) The other thing sounds like a bug. :/
suokko: Real bug is that linux kernel takes a lot of processing time for setting cache control for pages
suokko: So maybe ttm would be better of reserving large pages of kernel memory at once and then just rotating that memory between applications
suokko: I don't understand why cache control headers have to be set so often in ttm. But maybe there is good reason
AndrewR: suokko, and you get same oprofile results with nopat kernel option?
suokko: I haven't tried
suokko: # CONFIG_X86_PAT is not set
suokko: Seems like default for me
suokko: TG has a nice mm document in their website that says unbinding memory is fast operation unless cache control is changed ... If understood TTM correctly cache control is changed every time when page is freed.
biker_rat: Does anyone her know what a Firegl V3250 is equivalent to in a radeon series?
nanoki: wagers bridgman would :)
nanoki: rats.
nanoki: Signs that you're a nerd, part 16537: you make remarks to channel backlog which you are sure no one ever will bother noting but you yourself find somehow amusing.
nanoki: (and yes, the recursive self irony is not lost on me)
fcami: nanoki: I was about to answer and prove that staying connected to IRC long enough was the key to getting answers
fcami: (by effectively not telling him because he's gone)
nanoki: Neat, you knew the answer? What was it then?
fcami: http://ati.amd.com/de/companyinfo/press/2006/4854.html <= I know for a fact that V3100, V3200, V5000, V5100, and V7100 are R400s.
fcami: since the V3250 is included in that list, it is most probably the same arch.
nanoki: Right, thanks. Good to know. :)
fcami: nanoki :)
fcami: of course, that's not a proof - it could still be R500. not sure, that chip was rare IIRC.
nanoki: Yeah, will search for more press releases back home to make sure. The idea of using them was a good one at the very least. ^^
fcami: hehe, thanks. :)
fcami: I wish hardware manufacturers would adopt a reliable, easy to follow naming scheme.
fcami: and keep their hardware on spec as well.
nanoki: Or maintain an up-to-date chart on what means what on the Internet.
nanoki: In either case, AMD/ATi is probably free to use these ideas completely royalty-free if they so wish. ;)
fcami: yeah, the chart idea is pretty cool, there is already one on Wikipedia, but it is somewhat incomplete.
bridgman: fcami; FireGL/FirePRO numbering is different from consumer
bridgman: first digit indicates "positioning" (low, medium, high, ultra-high etc.)
fcami: hi bridgman :) yes, I know. My V7100 salutes you.
bridgman: next digit is essentially generation
bridgman: say hi to your V7100 for me
fcami: hehe, will do next time I open the case :)
bridgman: it gets confusing because so many people copy and paste specs from other cards when making their web pages
bridgman: so the web pages are basically placeholders but the information is wrong
bridgman: most people don't read the specs anyways, I guess ;)
fcami: is a spec-addict
bridgman: here is a very comprehensive description of the 3250
bridgman: http://uk.ts.fujitsu.com/rl/servicesupport/techsupport/ProfessionalPC/Workstation-Graphics/Graphics_Cards/ATI%20FireGL%20V3250.htm
bridgman: but some of the information is obviously wrong; it can't be an X600/RV380 *and* have 16 pixel processing pipelines
bridgman: X600 is a 4 pipe AFAIK
fcami: I'm not even sure putting a X700-like fan on a X600 makes sense either
fcami: hmm, 256 bit ram. can't be X600 IIRC.
bridgman: it says 128 bit further up ;)_
fcami: haha. what did you say about copy / paste ?
bridgman: we do try to overbuild the firegl cards though
Veerappan: probably wouldn't do to have the expensive workstation cards die an early death due to heat... :)
fcami: they're pretty good, as far as I'm concerned. I'd rather have more metal and less fan power sometimes (I modded my V7100 and 8800), but they're good out of the box.
bridgman: yep, I don't like relying on fans either... unless you have 2 and a reliable way to know when one of them has failed
nanoki: Heh, the card sizes are approaching ridiculous imo for at least consumer cards. :)
fcami: bridgman: s/2/2 whisper-quiet ones/
nanoki: I already ended with problems with my ATi card partially overlapping my SATA connectors on my motherboard.
fcami: which is essentially my V7100 setup : huge HR03 heatsink, 12cm fan in front, 9cm fan in back.
fcami: and case airflow.
bridgman: yep, that sounds good
nanoki: (Sure, bad planning ahead from motherboard manufacturer but can't the cards be miniaturizised at all? :/ )
bridgman: sure, but cost goes up as size goes down
bridgman: look at mainframes
nanoki: Meep. :) Point taken.
nanoki: At this rate you'll run into cases not being big enough for display cards in a few years though. :p
bridgman: nah, it's just a question of keeping power under control
bridgman: size is more or less a function of the amount of heat we need to dissipate
nanoki: Well, I didn't mean the size as in the size of the fan.
nanoki: Ohh, right.
bridgman: so here's the way I interpret the v3250 number
bridgman: 3 = entry level
bridgman: 2 = r3xx/4xx generation
bridgman: 50 = larger ram than base
nanoki: (The thing is that cards range further and further along the PCI plot and would eventually hit the front of the case :)
bridgman: v3200 is X600 128MB
bridgman: v3250 is X600 256MB
nanoki: PCI slot even.
conso: hi! I'm using the radeon driver on my notebook. The framerate in games seems pretty ok and also compositing is working fine, but the graphics in some games are just plain wrong. A lot of textures are just black. When playing eduke32 i can only see something when shooting.
conso: is that the normal behaviour of the radeon driver, or am I missing an option?
nanoki: Which distro and are you using out of the box drivers?
conso: I'm using arch-linux. I'm currently on my desktop, but I'lll get my laptop now for more info
TCW: conso, one needs the old Duke3D CD to play eduke32?
conso: aye
TCW: I hope I still have it... thanks for remmebering me :)
conso: i always had one in my rack, ^^
TCW: Yes... "had" could be the case here :)
Veerappan: I'd let you know how my testing goes, but my Duke CD's are all 1200+ miles away
conso: try out the high resolution pack, it's really amazing what those guys have done!
TCW: hooray, found it!
TCW: :)
conso: ok, back to my radeon-problems: my package-manager tells me I'm using xf86-video-ati 6.12.2-2
bridgman: games will be mostly affected by the mesa driver, not the radeon/ati driver
conso: but i also tryied the svn-drivers and had the same problem
nanoki: Mesa version counts more.
conso: 7.4.2
nanoki: Hmm, I'm slow. :)
conso: why that? :D
bridgman: I'm probably closer ;)
bridgman: he means I answered first
bridgman: you might want to try a newer Mesa
conso: ok, I'll look what yaourt can give me :)
suokko: And what hw is that? None cares if it is r100 or r500 card that has problems ;)
conso: -shoud be r300
bridgman: I think the driver does a pretty good job of exposing what is supported, doesn't it ?
conso: radeon mobility 9500 iirc
TCW: conso, the hrp pack exe is a zip archive?
conso: probably
mcgreg: hmm I think I'm an update-junkie ;) I check at leats twice a day if there has bene any updates on the git tree of ati-radeon, drm and mesa ;)
nanoki: bridgman: Yeah, via renderer string.
conso: ok, my package manager has got mesa-git
TCW: conso, ehrm... in the eduke32 zip are only exes?
conso: i could try it out :)
Veerappan: mcgreg: you're not alone. I only do a git pull daily, but I refresh the relevant cgit.fd.o pages regularly.
nanoki: (As in, iirc there were some stuff which clearly show the card is too old there)
conso: tcw: where did you get it?
TCW: conso, http://www.eduke32.com/
conso: strange. what distribution do you use?
suokko: joins the club
TCW: conso, why does that matter?
suokko: I do git pull and check the changes with gitk ;)
conso: well, maybe there is a package for it ^^
TCW: conso, arch has a package for it?
conso: back to my radeon vs games problems: should i download mesa-git and libgl-git with my package manager or could it be there is a problem with missing ram on my graphics-card or something like that and I would only need an option in xorg?
otaylor: conso: it's very, very, unlikely that an option will fix a bug
conso: tcw: kind of, you can install it with yaourt on arch
TCW: conso, btw. regarding your radeon issue... same problem with many games?
conso: otaylor: i wasn't shure it is a bug, but ok. I'll go for mesa-git ^^
conso: tcw: I only tried some, but must gl-based games hat that problem, more or less disturbing
conso: even things like trackballs and teeworlds have issues
mcgreg: I saw here the 21.june just passed ... it was the longest day on this year.
conso: aye
conso: and it was cold and rainy here ^^
TCW: conso, I am on a radeon x1600 and played some OpenGL games the last days, absolutley no issue. You have a much older card so I really wonder why you have issues but I don't. Older card / chip means better supported btw.
nanoki: Have a nightless night around there?
Veerappan: conso: going to try out trackballs as soon as I can install it, and I'll let you know how it works out.
MostAwesomeDude: bridgman: I never knew that about FireGL cards. Now the numbers make sense.
mcgreg: I wish they days would be as long as this one every day
conso: Veerappan: thx
mcgreg: MostAwesomeDude: one of your post in the phoronix forum about whining about the current situation regarding fglrx and free drivers ... a good post :)
mcgreg: some ppl doesnt seem to understand a but of anyhting but are whining all tghe time and over and over again.
Veerappan: well, first level is made up of red/black tiles, but not sure if that's intentional. trying out teeworlds next
conso: mesa-git is building on my laptop :)
mcgreg: heh
conso: Veerappan: on my desktop, all tiles look red and it looks way better that way
conso: (desktop has nvidia card)
Veerappan: k, well I can confirm that I'm getting red and black tiles which blend in with the background way too well... but I don't have screenshots of correct output to compare to.
suokko: Veerappan: Try mesa swrast :)
Veerappan: and teeworlds crashes the moment I try to join a server.
Veerappan: k. I'll try to figure that out
conso: i'll install trackballs here, to give a proper screenshot
MostAwesomeDude: mcgreg: If I do one thing well, it's StarCraft. If I do two things well, it's StarCraft and flames.
MostAwesomeDude: Actually, I'm kind of lame at SC. But my flames are great.
mcgreg: heheh
mcgreg: I love SC too
mcgreg: playing it regulary recently
mcgreg: but you definitely werent flaming. it was a good and fair post. and the ****ing truth as well.
conso: weee, just compiled dri2proto-git and glproto-git. mesagl is the last to come. I'm really fuzzy ^^ hopefully, games will look better after that
conso: veerappars: progress? thx again
Veerappan: unfortunately not yet. teeworlds crashes on level join, and I'm looking for info on using the mesa sw rasterizer
Veerappan: also on the phone
suokko: Veerappan: Yo ushould check MAD's blog
Veerappan: k
airlied: glisse: so Linus solved your page problem I'd guess, probably not the route we generally want to take though.
conso: here's a foto of how trackballs looks on my nvidia-card: http://img520.imageshack.us/img520/7052/bildschirmfotok.png
suokko: conso: SAme with my r200 card
suokko: So r300 has bug there
conso: ok
conso: for those who want to test eduke32 but don't have a a cd at hand: ftp://ftp.3drealms.com/share/3dduke13.zip
TCW: conso, everything black here too :)
conso: :/
conso: and you are using mesa-gl?
conso: can you see something when shooting?
TCW: http://paste.debian.net/39867/ <-
TCW: conso, no, Debian sid current
TCW: and yes, I see something when shooting
conso: finished installing mesa-gl. I guess reboot isn't necessary?
conso: will now test trackballs with mesa-git
TCW: too sad, Return to Castle Wolfenstein, Nexuiz, Urban Terror, tremolous, all wroking great here
conso: nexuiz is a little slow on my old laptop, but i don't have any glitches in those games, afaik.
conso: it's just that the duke is so ultra-cool and has much better single-player imho
TCW: " tcw: I only tried some, but must gl-based games hat that problem, more or less disturbing" <- you said that all other games have that issue too?
TCW: s/all/most/
conso: aye. it's a while back that i tried trebulous and urban terror. nexuiz was painfull slow, so I didn't check a lot, but I don't remember any glitches in those games
conso: so I probably played those games with an older driver/mesa
conso: in the last weeks i only tried some games, like duke-nukem, trackballs, teeworlds and hedgewars and they all seemed to suffer from simmilar problems to me
conso: i just recalled my experience with those games when you named them
conso: sorry, if i confused things
TCW: conso, you use the HR pack?
conso: aye
TCW: conso, me not at the moment... I hoped different texture set may fix the issue too :)
conso: nope, it doesn't ^^
conso: ok, i just tried eduke32 with fresh mesa-git and nothing changed.
conso: may a reboot change something?
TCW: conso, most likely no
conso: ok
conso: so, thx for the bug report. I wouldn't have been abled to write it that precisely
TCW: conso, you did file a bug?
conso: nope, not yet
conso: before entering this channel, i still thought there was something wrong with my xorg.conf
TCW: conso, but you inted to?
conso: if you think it helps
TCW: conso, " so, thx for the bug report. I wouldn't have been abled to write it that precisely" <- so what did you mean with that? :)
conso: http://paste.debian.net/39867/ <-
TCW: that's no bug report, that's a message I get from several other games too... but they work nevertheless
conso: ah, ok
TCW: it was just for the channel :)
conso: :/
TCW: conso, I am almost sure it is a bug in eduk32 with radeon
conso: ok, so it should be reported to them
TCW: conso, at least the project seems to be a very active one... so it may be a very good idea to contact them
TCW: conso, the the comments http://happypenguin.org/show?Duke Nukem 3d&start=10
TCW: see the*
TCW: url with whitespaces... damn http://happypenguin.org/show?Duke%20Nukem%203d&start=10
conso: i just went to #eduke32 :)
conso: BUILD_NOFOG=1 eduke32
TCW: hmm... the hints in the comments did not help :/
conso: works
TCW: conso, oh?
TCW: conso, reason?
suokko: fogcoord extension is missing in r300 driver?
TCW: conso, does not work here either... and in addition to "almost everything black", the remaining textures like the weapon and the ammo and stuff are garbled
conso: strange
TCW: conso, or was it a build option?
conso: no, it's runtime
conso: and works perfect for me :/
conso: a little slow, though ^^
TCW: $ md5sum duke3d.grp
TCW: 981125cb9237c19aa0237109958d2b50 duke3d.grp <- conso, would you be so kind and compare the md5sum?
conso: 22b6938fe767e5cc57d1fe13080cd522 duke3d.grp
TCW: conso, what version? But... I got it fixed here without that BUILD_NOFOG=1 option... nano ~/.eduke32/eduke32.cfg and changed ScreenBPP = 32 to 8
TCW: to be correct, I did change that value to 16 but after a testrun I saw it was changed to 8 instead
conso: as long as it works ... i have the atomic edition. some old cd i got frome a market back in the days
conso: the HRP is pretty amazing, isn't it?
TCW: conso, I didn't play around with the HRP pack yet
conso: have fun ^^
suokko: OS fixed function fog is broken
TCW: conso, how do I setup the HRP stuff in linux? Obvioulsy I don't want to click on setop.exe ;)
DanaG: Just extract the stuff, I believe.
suokko: TCW, Run it in dosemu/dosbox ;)
DanaG: Duke3D had a really amazing game engine for its time... you could even see your own reflection!
DanaG: Eat that, HL2! =þ
TCW: suokko, not an option :-P
suokko: I think that some 3D games used really cool stuff before all the acceleration stuff started. Now it is again starting to be possible to use unique techniques
conso: dosbox won't let you use HRP ;)
suokko: TCW: Do you have mesa source in your HD?
TCW: suokko, "in my HD"?
suokko: harddriver :)
TCW: on my hdd?
TCW: HardDiskDrive :)
conso: back in the days, duke3d was the only shooter I needed and liked. It's really amazing. I wished duke nukem forever wouldn't have been cancled. It would be a reason to update my old pcs
suokko: yes
TCW: suokko, yes... but from last week. Why?
suokko: Can you go to prgos/tests
suokko: and type make fog
suokko: and then run the fog :)
suokko: What does it look like?
TCW: suokko, oh... I can't do that right now
conso: maybe I can
TCW: the source I have is from git last week, the stuff I _use_ right now are debian packages
suokko: conso: You need mesa source code for it. If you have it then it is easy
suokko: But yo ucan run the test from git with debian packages :)
TCW: suokko, oh? It does not need to match?
conso: i just build from git, but yaourt (something like an package-manager with build-script support) already deleted the sources :(
suokko: No. prog directory just includes standard alone opengl code :)
conso: is it prog or prgos?
TCW: http://paste.debian.net/39868/ <- suokko
suokko: Do you have glut installed?
suokko: (or freeglut because it is debian?)
TCW: suokko, oops... installing it now :) I was almost sure it was installed already... but ... :)
TCW: suokko, so fog is running. What do you want to know?
suokko: Does it render the boxed with different colors?
TCW: suokko, hum... hard to say :)
TCW: I'd say no... almost everything white, I can see some different shaded artifacts in some boxes, the middle row is darker than top and bottom
suokko: http://img145.imageshack.us/img145/1159/fog.png
suokko: For me fogcoord is broken but fixed pipeline works
TCW: nope, looks totally different here
suokko: So fixed pipeline fog is broken for you
conso: and me
suokko: Is there bugzilla report about that?
TCW: Some output of this program requires GL_EXT_fog_coord <- I get this line as output (and some other lines too)
suokko: osiris_: Do you have any good guess why DRI2 r200 EXT_fog_coord is broken? :)
suokko: TCW: But it should at least draw 3 rows of boxesin top without fogcoord
TCW: suokko, yes, 3 rows
TCW: ha,,, HRP running :)
suokko: and all white
TCW: suokko, more or less white, yes
conso: i have an pretty old hr-pack here, I'll download the new one later. It's still amazing, if you ask me. The game looks a lot younger
suokko: TCW can you make a report about fixed funtion fog in r300 is broken? (I don't have hw so can't help testing it)
TCW: suokko, fixed function?
DanaG: hah, eduke32 just segfaults for me.
TCW: DanaG, HRP or default?
DanaG: HRP.
conso: TCW: BUILD_NOFOG=1 deactivates fixed function fog, so that's what to report, I guess
TCW: DanaG, did it heare too, uppercase files? rename them to lowercase
TCW: conso, but I don't need that option, so I am a bit confused what to report
DanaG: Which files?
DanaG: and how can I programmatically fix the case of all the files?
TCW: DanaG, the one it is mentioning... game.con iirc
TCW: eduke.con was it here
DanaG: ls: cannot access *.con: No such file or directory
DanaG: i.e. I see no .con file anywhere.
TCW: DanaG, in the DUkePlus directory?
DanaG: Dukeplus? I don't have dukeplus.
DanaG: Just HRP.
TCW: DanaG, huh?
TCW: d3d_hrp/DukePlus/eduke.con <- DanaG
DanaG: hmm, I must have the files in a separate place -- none of those exist.
DanaG: Where exactly did you get eduke32?
DanaG: And HRP?
TCW: DukePlus is a folder in the HRP I downloaded
DanaG: Odd... I know I have models of some sort instead of sprites... yet I can't seem to find the HRP.
TCW: DanaG, models instead of sprites? Here it still looks like sprites
TCW: the monsters
TCW: DanaG, and whyt do you mean with you can't find HRP?
TCW: what even
DanaG: I'm looking for it in the game dir... and in my random-downloads dir; it's not in either place.
TCW: DanaG, did you download the HRP stuff?
DanaG: I'm sure I did... yet I can't find it.
TCW: http://www.duke4.org/files/hrp/dn3d_hrp-4.0(321).exe <- this is the one I downloaded
DanaG: Oh, and stupid eduke32 site has only Windows builds, and not an immediate link to source.
DanaG: Instead, they have a link to a link.
Veerappan: yeah, a link to a directory tree you can get the source at
DanaG: oh, I see... I downloaded the HRP as a zip, and just stuck it in a game dir.
DanaG: It worked before... but now it doesn't. Odd.
TCW: crap... I broke duke3d :)
DanaG: oh hey, it works again.
DanaG: Just had to compile a new version.
TCW: I did set in options from software rendering to OpenGL, now the gfx are all mixed up (just squares, no text in the menu) I can't recocnize the option names to revert and don't find the settings file... could one help me to find the right option to revert back? :)
TCW: it was in the video settings submenu
DanaG: okay... how do you use the HRP?
TCW: ./eduke32 -game_dir DukePlus -forcegl <- DanaG, no idea about the forcegl option, but it worked this way
DanaG: I still see no DukePlus anywhere.
TCW: DanaG, look in the DukePlus.bat file
DanaG: Here's what's in my HRP file:
TCW: conso, help? :)
DanaG: hrp_art_license.txt, hrp_readme.txt, autoload dir{duke3d.grp dir{maphacks,zip, duke3d_hrp.zip}}
DanaG: s/,zip/.zip/
DanaG: i.e. two more zip files within the one zip file.
DanaG: Maybe I should read... the readme.
DanaG: "Then run the installer (the file you downloaded/got from a friend or whatever) by double-clicking it. Then just follow the onscreen instructions."
DanaG: HAH.
DanaG: There IS no installer.
TCW: http://paste.debian.net/39870/ <- DanaG, that's a ls -lR of my complete setup
DanaG: Where'd dukeplus come from? That's not part of HRP.
TCW: DanaG, see the download link above
TCW: http://www.duke4.org/files/hrp/dn3d_hrp-4.0(321).exe <- this is the one I downloaded
DanaG: I was using the zip version.
DanaG: That's stupid... the zip version is missing things!
TCW: that is a zipped-ese
TCW: exe
TCW: DanaG, zip of HRP? I did not see a zip anywhere
DanaG: That's really bad manners, though... shafting the linux users by making them think they should use the zip version... and omitting things from that zip version!
DanaG: http://hrp.duke4.net/download.php
DanaG: "Separate download version; also included with current HRP v4 installer"
TCW: DanaG, btw. why do you need any installer? I just did merge the built eduke32 stuff and the HRP stuff from the exe-HRP-download
DanaG: oh, and the dukeplus.rar has eduke32 windows version in it.
DanaG: I don't need any installer, but I don't expect downloading the "zip version" of a thing (because I shouldn't need an installer!) to OMIT things.
TCW: and yes... in there are prebuilt win32 binaries... ignore them
DanaG: That's also bad manners.
DanaG: =þ
TCW: I did delete them
TCW: DanaG, does it start now?
DanaG: nope, re-adding HRP makes it segfault again.
TCW: DanaG, compare to my ls -lR, does it look the same?
TCW: note the EDUKE.CON
DanaG: oh, in dukeplus?
DanaG: well, that shouldn't have been a factor before, when I didn't even have whatever that dukeplus thing is.
DanaG: ugh, how the hell do you make their thing use gcc -g? Their Makefile is really cryptic.
DanaG: OURCFLAGS? makes me think of "ORC"flags.
DanaG: and USE_OPENAL ?= 1
DanaG: question-equals?
DanaG: http://pastebin.com/f4691c9cd
DanaG: got a backtrace this time.
TCW: just works here
DanaG: Okay, that's odd... turning off compiz makes it not segfault.
DanaG: I can understand the app performing horribly under compiz... but segfaulting? What gives?
TCW: DanaG, DR1 or 2?
DanaG: fglrx, actually. No 3D on R600. :(
TCW: fglrx?!
TCW: You dare to mention that piece of crap HERE?! :)
DanaG: But still... it's not segfaulting IN that driver!
DanaG: Unfortunately, there's no 3D on radeon R600 that I could run eduke32 on.
TCW: *eeks* save does not work?!
TCW: I did save but now I can't load that save game, the load game list is empty
TCW: hum... more or less EWIN ;)
DanaG: heh, Duke runs really really fast. He must be related to Sonic the Hedgehog, or something.
suokko1: Sonic just rolls everywhere. Or at least the old sonic did :)
DanaG: "Select the Game tab and then select DukePlus from the gamedir menu.  Now click Start.  The next time you launch EDuke32, Duke Plus will already be selected."
DanaG: There IS no "DukePlus" in that tab.
TCW: DanaG, as said "./eduke32 -game_dir DukePlus -forcegl"
DanaG: The configuration file "DukePlus/dp.cfg" was not found. Import configuration data from "eduke32.cfg"?
TCW: yes
DanaG: er, I hit no -- and next time I started, it didn't ask me again.
DanaG: I accidentally hit no before you answered.
TCW: DanaG, does not matter that much
TCW: DanaG, worst case you have to change some settings / key bindings again
DanaG: pgup and pgdn just look up and down.
TCW: you don't use the mouse to play?
DanaG: I do play with mouse.
DanaG: But it still looks up and down on pgup and pgdn, for some reason.
TCW: so why bother with pgup / down?
DanaG: That's what dukeplus menu is bound to.
TCW: hm?
DanaG: That's what the problem is... it's doing "look up", not "dukeplus menu"
TCW: I don't get you
TCW: what menu?
DanaG: Read the dukeplus manual.
TCW: why?
DanaG: "what menu?"
DanaG: the manual tells about one.
TCW: DanaG, I did just merge eduk32 with HRP, and start eduke32 now as the DukePlus.bat suggested.... that's it
DanaG: DukePlus is a completely separate mod.
TCW: oh... sorry
TCW: Then I just don't get HRP / DukePlus...
DanaG: okay, apparently "use dukeplus menu" actually MEANS "look up".
DanaG: Because I moved the key... and it still just... looks up,.
DanaG: Miserable failure, this dukeplus.
DanaG: ugh, just segfaulted on falling into the water.
TCW: praise fglrx
TCW: anyhow, I am sneaking into my bed... by
DanaG: no, it didn't do that before dukeplus.
DanaG: http://forums.3drealms.com/vb/archive/index.php/t-34458.html
DanaG: ah, looks like a lot of it was failure to handle 64-bit.
sgcb: is there some reason why dri is disabled when starting multiple X servers?
airlied: sgcb: because DRI doesn't work with multiple X server
airlied: it works now on Fedora 11
airlied: due to kernel modesetting
spstarr: looks at git commits
dmb: airlied, do you know if anything has gone in in the past couple of weeks to increase performance of exa with KMS?
dmb: because for me, its as if exa were disabled
airlied: dmb: gone in where?
airlied: we've mainly spent the last weeks just getting stuff from my tree into the mainline tree
dmb: airlied, i guess drm-next
airlied: not actually writing any code
dmb: oh
dmb: i have a feeling its unusually slower for me then others
airlied: it would be slower than the rawhide kernel, esp on AGP
dmb: this is a mobility card
dmb: pciexp i think
airlied: is it totally unuseable? or just feel slow?
airlied: feel slow is probably where we are at right now
dmb: airlied, pretty unusable, like the same as if no 2d accel was enabled
airlied: paste Xorg log file
dmb: http://pastebin.ca/1469422
airlied: wierd its using greedy
airlied: not sure where that is coming from, remove xorg.conf and try again
dmb: airlied, i think thats from the fact I told it to use XAA (from before), but it still shouldn't effect it should it?
dmb: i'll try it without one
airlied: greedy may also be a problem with the X server
airlied: some distro have dirty hacks in there servers.
dmb: yes (ubuntu)
dmb: brb, restarting x
dmb: airlied, without an xorg.conf, i get cannot open framebuffer, specify busids
dmb: so i just created a simple one
dmb: and told it to use radeon driver
dmb: it still says forcing greedy migration option
airlied: ah mustbe some umbongo bong hits
airlied: you can probably try Option "EXAMigration" "always"
airlied: or "smart" I think
dmb: airlied, should i also tell it to use exa?
airlied: I think I ignore the XAA option in kms now
airlied: so it should always use it
dmb: oh
dmb: i'll try the Option "EXAMigration" "always"
dmb: airlied, still says (II) EXA(0): Forcing greedy migration option
dmb: when using always
dmb: and says
dmb: (WW) RADEON(0): Option "EXAMigration" is not used
airlied: oops
airlied: it MigrationHeuristic
airlied: I always forget those option names
dmb: oh
dmb: airlied, so it would be Option "MigrationHeuristic" "always" ?
airlied: yes try that
spstarr: airlied: it's winter there now? :(
airlied: realises dmb is using kms so migration won't help. doh
airlied: spstarr: well winter isn't really a season over here, its 20 degree C outside, granted it is raining
spstarr: that's not a bad winter
dmb: airlied, weird, it says
dmb: (II) EXA(0): Forcing greedy migration option
dmb: (**) RADEON(0): Option "MigrationHeuristic" "always"
airlied: dmb: don't worry I just realised with KMS it doesn't matter which one you get
airlied: is half asleep
dmb: oh
dmb: airlied, i wonder if it might be the fact that i use 1920x1200, it really needs that performance boost
dmb: (optimization of the path)
airlied: boots G5 with rv515, kms ring failure there.