diogo: hey, I was using the F9 on my laptop (ATI Xpress 1150/Dell vostro 1000) and using the radeon driver but had problems to select in blender (used ati 6.8.0 and mesa 7.1) there has been sometime since that time is it still a problem occuring on RS48x chips and radeon + mesa dri?
diogo: someone >> blender + xorg 1.5rc+mesa 7.1rc + radeon 6.8
diogo: (64 bits)
buggs: diogo, wait some time (hours)
diogo: hours? damn too busy?
buggs: at home i could give it a try
buggs: but i was more relating to the general idelness of the channel
diogo: ok... I'm downloading F9 again to try it... I'll comeback later then
kriko: ~time Time_for_sex
N00dl3z: Arse. I thought I'd tracked down my 2.6.26 related radeon/X hangs (nearly 5 days uptime rather than the usual < 2) but then it just went again.
kriko: wrong channel
nha: kriko: awww
kriko: I'm joking with a bot, nothing to be ashamed of :)
kriko: just missed the proper window
diogo: Hi.... http://imagebin.org/23459 take a look here using F9 and got this problem on my video card + blender
diogo: does anyone knows how to fix it (if DRI is disable than it works)
gentoofu: anybody know anything about the RM-70 processors?
TobiasTheCommie: an rm-70 is a rocket launcher
joecool: is it possible to scale the r500 down with radeon?
joecool: for a laptop... this thing is still a pressue cooker
joecool: powerplay i guess?
joecool: hrm, i see.. alex has an overlay for it
joecool: heh.. i asked that same question in #radeond about 12 days ago
joecool: ty google
joecool: gentoofu: i'm bad with memory too :(
redeeman: joecool: if you bored you could try the new bicubic XV scaling
joecool: i'm running the latest master right now
redeeman: im not sure it's merged
joecool: unless i had to do something special with mplayer to use it
redeeman: you must use xvattr to set it to use it
redeeman: but i don't think it's merged
redeeman: i think mostawesomedude has it in a branch
joecool: where's his branch?
redeeman: i have zero clue
joecool: i don't see it on freedesktop
joecool: unless he goes by another name
redeeman: do not know
joecool: redeeman: how do i set it even
joecool: you're not giving me much to work on
redeeman: that, i can find, 2 sec
redeeman: just have to grep my log :)
joecool: ah csimpson
joecool: that helps
joecool: but i could patch..
joecool: .... but i'm lazy too
redeeman: he talked about pushing yesterday
redeeman: so maybe he did actually merge
joecool: he's got his own branch
joecool: for it
redeeman: i really must fix my box so i can get the R500 in
joecool: ok... time to kill X and see how this goes
onestone: MostAwesomeDude: ping
MostAwesomeDude: onestone: Pong.
MostAwesomeDude: Moar Xv improvements?
onestone: MostAwesomeDude: yes. http://dev.compiz-fusion.org/~onestone/bicubic_patches.tar.gz
onestone: MostAwesomeDude: the last optimization that could be possible is to use a FP16 helper texture instead of FP32. But I will ask agd5f or bringman how much faster it could be.
MostAwesomeDude: onestone: Yeah, that may help on the r3xx.
MostAwesomeDude: Which reminds me.
RTFM_FTW: well its less bandwidth intensive
MostAwesomeDude: We should investigate writing r3xx stuff. I can't test without a card, though. I've got to go to the recycling center tomorrow; I'll see if there's any old Radeons sitting around.
MostAwesomeDude: RTFM_FTW: IIRC we shouldn't be re-uploading the texture each time.
onestone: MostAwesomeDude: I don't have a r300 but I think that I will try to write the current r500 code as arbfp again to make the GPU shader analyzer generate r300 code
RTFM_FTW: sourcing from the data
MostAwesomeDude: onestone: That would work quite well.
onestone: MostAwesomeDude: but it's not that easy to convert the shader analyzer output to driver code. It resulted here in a very expensive nearest neighbor filter here ;-)
MostAwesomeDude: onestone: Yeah...
MostAwesomeDude: I'm not completely confident on my abilities with r3xx shader code, but then again, can't be worse than my r5xx skills.
onestone: MostAwesomeDude: And my patches show that a human head can mutch better optimize code ;-)
MostAwesomeDude: Well, *your* human head, at least. ;3
onestone: Once you have working code it's easy to optimize it, because you see if you changes break something ;-)
RTFM_FTW: mmm optimizing shader code is a fun experience
onestone: MostAwesomeDude: the last patch took more than 8h only because I made stupid mistakes, and I discovered some swizzling restrictions on r500
RTFM_FTW: mmm do the open source docs discuss the swizzling limitations
onestone: But my 160x120 test video helped a lot
onestone: RTFM_FTW: no idea
RTFM_FTW: heh I'm surprised you guys haven't started on a shader compiler for R3xx, R4xx, R5xx yet...
onestone: RTFM_FTW: I have no printed version of the docs and doing driver development is not easy with only one pc, because you are restarting x a lot
onestone: but kernel core developement (no loadable modules) is much harder
RTFM_FTW: oh I'd disagree :D
onestone: rebooting the whole time
RTFM_FTW: developing a highly optimal driver side "JIT" compiler isn't any easier... in fact AFAIC its considerably less trivial
RTFM_FTW: on a single box you might have a point :D
onestone: even compiz development isn't easy with a single box. you can be happy if you can start a second accelerated x server session
onestone: MostAwesomeDude: do you know how to add a xorg option for the bicubic filtering? I also don't think that it blurs too much even with small scaling ratios
magyar: hi, i got xorg radeon driver ver 6.9 for a x1950 card. Does this module support 3d? also if it does how do I set it up?
leio: 3D if provided (don't know about that), would be provided by DRI drivers, not xf86-video-ati
magyar: leio: i have "libgl1-mesa-dri" installed
MostAwesomeDude: onestone: Yeah, I already have a patch for the Xorg stuff.
MostAwesomeDude: onestone: Also yeah, I noticed that the filter isn't super-strong. But it *does* work a bit.
MostAwesomeDude: RTFM_FTW: The swizzle limitations on r5xx are: no per-channel negate masks, only per-source negate masks. That's it.
MostAwesomeDude: We can even swizzle twice, on textures, so we have tex swizzling too.
magyar: leio: also "man radeon" indicates support for "RV570/R580 Radeon X1900/X1950" in 3d
MostAwesomeDude: magyar: You'll need Mesa 7.1.
redeeman: and xorg 1.5 right?
MostAwesomeDude: redeeman: Yeah, for coompiz.
redeeman: what if i don't want compiz but just 3d?
MostAwesomeDude: redeeman: Xorg 1.4.2 should work.
magyar: MostAwesomeDude: I only have libgl1-mesa-dri-7.0.3-5 in debian lenny
redeeman: im guessing that means my 1.3 won't work :)
MostAwesomeDude: magyar: Yeah, I know. It's not very wide-spread yet, mostly because it's not released yet. ;3
magyar: MostAwesomeDude: so for now I am sol. fglrx is the way to get 3d
MostAwesomeDude: magyar: Yeah. Just wait for a few weeks.
MostAwesomeDude: Really, right now, it's just a matter of Brian deciding how many more buggies to fix before tagging 7.1, I think.
MostAwesomeDude: So sometime next week.
MostAwesomeDude: And then Debian should pick it up quickly.
redeeman: but lenny is freezing up
redeeman: so it probably won't make it
magyar: backport :)
MostAwesomeDude: TBH the entire assembly of Xorg 1.5 and Mesa 7.1 and etc. will probably not be in Lenny.
MostAwesomeDude: But that's why I'm pegged to "testing" and not "lenny".
redeeman: MostAwesomeDude: how come mesa stuff needs x updates?
magyar: MostAwesomeDude: i've noticed that the radeon manpage has an option "Option "DRI" "boolean"" which is off by default
MostAwesomeDude: redeeman: Nope, X needs Mesa updates.
MostAwesomeDude: magyar: That's the radeonhd driver, not the radeon driver.
redeeman: MostAwesomeDude: but how come newer mesa don't work on all X ?
redeeman: if it's that way
MostAwesomeDude: redeeman: Because certain things have changed.
redeeman: ah okay
MostAwesomeDude: That's all.
MostAwesomeDude: The biggest change is in AIGLX.
redeeman: i guess ill just get 1.4
redeeman: well i don't need aiglx
magyar: MostAwesomeDude: I typed "man radeon"
MostAwesomeDude: redeeman: Well, the changes around it, like the software rasterizer...
MostAwesomeDude: magyar: The option in radeon doesn't do anything. DRI is enabled by default.
magyar: bahh, never mind
magyar: The default is off for RN50/ES1000 and on for others
magyar: do you guys have any tips as to which options should one use in xorg.conf?
MostAwesomeDude: magyar: Here's what's in mine:
MostAwesomeDude: No special options.
MostAwesomeDude: At all.
MostAwesomeDude: Well, PowerPlayMode and BicubicFiltering, but those aren't in the main driver.
magyar: k, thanks
redeeman: joecool: ^^
MostAwesomeDude: onestone: 'k, (finally) gonna go test some codes. BRB.
MostAwesomeDude: onestone: Back. Looks solid.
MostAwesomeDude: Um, out of curiosity, is there any kind of video out there that would REALLY emphasize the difference that the filter makes?
redeeman: what algorithm was used before bicubic?
MostAwesomeDude: redeeman: Bilinear.
redeeman: MostAwesomeDude: on simpsons it's easy to see difference
MostAwesomeDude: redeeman: Or, rather, separable 2D bilinear.
redeeman: when it's in 320x240
redeeman: scaled up to 1600x1200
onestone: MostAwesomeDude: http://www.tvshows.de/alf/videos/alf20000.avi I've found the float to half conversion code in mesa. I don't think that we can to the calculation in python
MostAwesomeDude: onestone: You need 16-bit floats, right?
MostAwesomeDude: Could you point me where it is in Mesa? I'm pretty good with Python, I could figure it out.
redeeman: MostAwesomeDude: tomorow i will see if i can find an example taht really shows difference between (bi)linear and bicubic
onestone: MostAwesomeDude: src/mesa/main/imports.c:643 we could save 768MB bandwidth on 1600x1200 at 25fps
redeeman: wow that's alot of bandwidth
onestone: MostAwesomeDude: you can see the difference if you have text in the video
redeeman: sharp edges in general i'd say
redeeman: i find that it's particularly easy on cartoons
redeeman: since it's relatively "clean" to begin with
redeeman: but i gotta sleep now
jake_: hi all
jake_: I am getting some pretty good tearing on my second screen - but not the first. It happens in clone mode, and normal, its very weird....
fredlwm: Is there a better channel for that question (DRI related ?) ? I moved from kernel 2.6.25 to 2.6.26, Mesa 7.1-rc1 to 7.1-rc3 and xorg-server 188.8.131.524 to 184.108.40.2065. The problem is that, in an hour or less with DRI (r300 RS690) enabled in the new versions, everything freezes. The mouse pointer still moves, but has a new look and doesn't do anything. Issuing a sysrq+k freezes it. sysrq s+u+b works and reboots. When it freezes, I'm just bro