MostAwesomeDude: So, does anybody know why the weird banding and striping?
MrCooper: MostAwesomeDude: what banding and striping?
MostAwesomeDude: reposts images
MostAwesomeDude: I know it's a fragment program error, I'm just not sure what I'm doing wrong.
MostAwesomeDude: I was about to dump the fragprogs and start wading through those, but I wanted to see if anybody had investigated in the past three hours.
MrCooper: no idea I'm afraid
MostAwesomeDude: No worries. I caused it, I might as well solve it. :3
MostAwesomeDude: would *really* like to be able to play Wine games again
MostAwesomeDude: Hm. At some point, we should rewrite the VP compiler. It's kinda lame.
z3ro: hmm. I woke up at 2am again for some resaon. =/
z3ro: sleep pattern fail.
glisse: z3ro: this is usually a good reason to go back to bed ;)
z3ro: yeah, not sure I'd be able to get to sleep again though.
z3ro: oh well, I'm going to look at some 4870's. I'm *very* tempted to get one of these, even though we don't have docs yet.
chithead: probably it will cost less to buy a cheap x1300 now and a 4870 later when open source 3d support arrives
z3ro: well I have a x1950, but it's still pretty slow on per pixel lighting rendering (eg lots of ALU instructions)
glisse: z3ro: llvm ! ;)
z3ro: yeah, that would probably help... but I'm not convinced it's going to be a really huge speed up.
glisse: z3ro: well on complexe shader i am convinced that any optimization will be huge improvement
glisse: llvm just fit the job for doing so, i am not convinced yet that it will be able to take advantage of all subtilities but i will definitly help
z3ro: I wonder how good fglrx's optimizer is (they use something similar to llvm, right?)
glisse: z3ro: in house compiler
z3ro: it would be interesting to compare the hardware program of a complex shader compiled by our driver, and compiled by fglrx.
z3ro: see what kind of optimizations they are doing that we might be missing.
glisse: but using same compiler technics as llvm from what you can learn from public doc
z3ro: I think we could get something that fits the hardware better by writing something custom. although llvm could still be a good start.
z3ro: I'd rather look at it from the perspective and needs of the newer hardware (r6xx/r7xx) and write it with that in mind, then kind of backport it to the older hardware.
z3ro: because it would suck having a perfect optimizer for r300, then we look at r500, and none of it can be reused.
z3ro: (ok, it wouldn't be that bad, but you know what I mean)
z3ro: I can see this really happening between r500 and r600 though.
glisse: z3ro: compiler technics are the same why redoing them when you can take advantage of existing software
marcheu: z3ro: then you can write 2 llvm backends...
marcheu: z3ro: I'll be writing (mostly) separate backends for nouveau on each card generation
z3ro: I see... I just hope that llvm will fit the hardware good enough to deal with the tricky bits.
z3ro: though from what I understand you can add your own passes to do special things.
z3ro: so that would probably work to deal with the weird things (pre-subtract/mul/etc)
marcheu: that's not weird things
marcheu: FWIW apple is doing the exact same (targetting llvm at ATI hw) ATM
marcheu: so I guess it's something that works...
z3ro: ok, point taken. :)
MostAwesomeDude: We've already got an optimizing compiler for r300 and r500 shaders. Really, LLVM is going to be way more helpful with generic optimizations IMO.
marcheu: you don't imagine how far it is from a real optimizer...
marcheu: either you rewrite it all, or you use something that exists and does a lot more than you'll ever implement yourself...
MostAwesomeDude: marcheu: Yeah, I'm just sayin', we're not completely in the dark over here. :3
MostAwesomeDude: wonders if it might be seriously easier to just start Gallium work now
glisse: MostAwesomeDude: what we got is not at all near what llvm offer ...
MostAwesomeDude: glisse: I know. :3
MostAwesomeDude: Oh, right. For anybody not in #dri-devel...
MostAwesomeDude: *Should* do fogcoords, but doesn't. Fixes the current issues, though. Also causes banding on some Wine games.
mcgreg: I though fglrx is supposed to be working with radeon 4670?
jcristau: mcgreg: /topic is your friend. you should read it.
adamk: I would expect a regular like you to know that fglrx questions should be asked on #ati :-)
mcgreg: hmm well... most of the experienced users are in here anyway ;)
mcgreg: at least VESA works
TobiasIsACommie: i'm going to buy a new amd card today
TobiasIsACommie: as anyone in here that can tell me "no, don't buy that card, buy this one instead"
TobiasIsACommie: i'm thinking a Radeon HD 4830
z3ro: was going to do the same, until the major fuck up of the day. =/
z3ro: I ordered a t500, they charged my card, then refunded it about 10 days later...
mcgreg: ok, fixed my problem. it needed the kernel module to load. yeah I know, off topic ;)
z3ro: and I couldn't contact lenovo at all...
z3ro: so I thought, well, they must have canceled my order
z3ro: and ordered a laptop from another place...
z3ro: so now they just changed my card again.
z3ro: so apparently I've got two laptops on the way, when I only wanted one. =/
z3ro: so hopefully doing a return won't be too hard... or I can cancel the second one.
z3ro: major fail. :(
TobiasIsACommie: ok, i know this is a really stupid and annoying question
TobiasIsACommie: anyone know if the saphire radeon cards work with the driver from amd(my current asus card doesn't)
TobiasIsACommie: and i'm asking windows driver here
TobiasIsACommie: jsut to make it doubly stupid to ask in here
glisse: TobiasIsACommie: i would be astonish if it doesn't
z3ro: TobiasIsACommie: no idea about the win32 driver, but the sapphire cards I've have all work with fglrx.
TobiasIsACommie: glisse: honestly, my asus doesn't, it is horrible
TobiasIsACommie: so, i assume when buying a card now(going for r770), that 4870 is better than 4850, and that 4830 is rought out of the picture
TobiasIsACommie: would any 4870 card be aok? or do you guys know of any specific brand that is good/better?
rx__: they marketed 4850/4870 as performance parts
rx__: 4830 is supposed to be mainstream
TobiasIsACommie: what i read was that 4830 was a castrated version of 4850
TobiasIsACommie: well, the card i have right now was just some cheap crap.. now i want a good card that can last me more than one year
TobiasIsACommie: it isn't because i want the bestest of the bestest and intend to upgrade every year, i just want something that can actually play stuff like red alert 3..
rx__: via linux or via dualboot
TobiasIsACommie: so basically, unless i want to run the very latest games at the very highest resolution, the 4830 will do?
rx__: before i go making assumptions.. what resolution?
TobiasIsACommie: a good card, that will be supported properly by this driver at some point, is what i want, and i would rather buy a bit better than i really need
TobiasIsACommie: and, afaik, one(or more) developer in here has the r770, so i'm going for that one specifically
rx__: i think the 4830 would suffice but that's just my opinion
TobiasIsACommie: the price difference here is so low between those cards that i would rather feel safe
TobiasIsACommie: thanks for your help :)
rx__: don't mention it
TobiasIsACommie: i guess i should go for 1gb ram right?
rx__: if a game requires it
rx__: doesn't look like there's 1gb variants
TobiasIsACommie: The Sapphire HD 4850 1GB is a version of the HD 4850 designed by Sapphire for the user who demands more memory for specific applications like rendering. It offers similar overall performance to the standard model, with clock speeds of 625MHz (core) and 993MHz (memory).
TobiasIsACommie: well, i found it
TobiasIsACommie: again, thanks
TobiasIsACommie: k, ordered the card
glisse: airlied: i changed relocation format in modesetting gem
glisse: so be aware that ddx needs update
glisse: i just put all fields i believe we might need to optimize relocation in the future :)
logari81: could someone tell me please if the following result of xvinfo on a X1950pro is normal , someone has problems with video playback with it, http://pastebin.ubuntu.com/70987/
agd5f: logari81: looks fine. XAA + composite + textured video doesn't work. use EXA instead
agd5f: Option "AccelMethod" "EXA"
logari81: already has EXA on
agd5f: logari81: what sort of problems?
logari81: he sees half transparent lines across his screen, in fullscreen video playback
logari81: it is strange because he gets the same problem (or even heavier) also with fglrx
agd5f: weird. never seen that one.
oga: sounds almost like dicked hardware to me
agd5f: yeah, I was just going to say that
logari81: it is a GeCube, even if I dont know what this means
chithead: gecube is a video card manufacturer
logari81: thnx a lot anyway