--- Log opened Втр Июл 25 00:00:07 2006
--- Day changed Втр Июл 25 2006
00:00 < pmdata> this is from the test_clear command on nv28 (dump from sturmflut yesterday)
00:00 < pmdata> Context is 800114f9
00:00 < pmdata> Graphics object
00:00 < pmdata> ENGINE_GRAPHICS[0] = 00001030 = class = 030 (UNKNOWN) | chroma key | patch = SRCCOPY_AND | dither = compat | patch valid = FALSE
00:00 < pmdata> ENGINE_GRAPHICS[1] = ffffffff = mono format = 255 | color format = 255 | notify address = ffff
00:00 < pmdata> ENGINE_GRAPHICS[2] = 00000000 = dma0 address = 0000 | dma1 address = 0000
00:00 < pmdata> ENGINE_GRAPHICS[3] = 00000000 = 0x00000000
00:00 < pmdata> Channel 0
00:00 < pmdata> Object not found in the table, guessing!!!
00:00 < pmdata>                     NvTypeffffffff      [0x0000/4] = object 00000000 type 030 (UNKNOWN)
00:00 < pmdata> 40133   0x00000000   0x00000000                       NvTypeffffffff      [0x0004/4] = 0x00000000
00:01 < pmdata> 40134   0x00000000   0x00000000                       NvTypeffffffff      [0x0008/4] = 0x00000000
00:01 < pmdata> 40135   0x00000000   0x00000000                       NvTypeffffffff      [0x000c/4] = 0x00000000
00:01 < pmdata> 40136   0x00000000   0x00000000                       NvTypeffffffff      [0x0010/4] = 0x00000000
00:01 < Lumag> Hmm... Strange.
00:01 < Lumag> I also got zero-name object handle as a reference, but I didn't expect it to be any real object.
00:01 < pmdata> yep, it seems nv20 as a clear command in the tcl engine
00:03 < Lumag> offsets look really weird. I hadn't seen such small offsets before.
00:05 -!- johill [n=johannes@p5487F022.dip.t-dialin.net] has quit [Remote closed the connection]
00:06 < Lumag> I really don't think it's a real object
00:06 -!- johill [n=johannes@p5487F022.dip.t-dialin.net] has joined #nouveau
00:06 < pmdata> I think so
00:07 < marcheu> pmdata: what's the raw binary dump around that area ?
00:10 < Lumag> marcheu: 261525 NOP's, 0x04000003 and then five zeroes again
00:11 < marcheu> hmm, that appears quite often in there. I wonder if the fifo is going too far... 
00:11 < marcheu> I guess it's some kind of meta command like jump
00:12 < pmdata> nevermind, it was as always the last part of a dump
00:13 < pmdata> so this is a "buggy one"
00:14 < marcheu> pmdata: btw please add a comment when you're uncertain about a command
00:15 < pmdata> ok
00:17 < pmdata> nv20 has 8 texture units, or 4?
00:17 < marcheu> 4
00:18 < marcheu> all nvidia cards have had 4 for a long time
00:18 < marcheu> including geforce FX
00:18 < marcheu> although there is a fx variant that is either 4x1 or 2x2 IIRC
00:19 < marcheu> yep
00:19 < marcheu> http://en.wikipedia.org/wiki/Comparison_of_NVIDIA_Graphics_Processing_Units
00:19 < marcheu> NV31, NV34 and NV36 are 2x2 pipeline designs if running pixel shaders, otherwise they are 4x1
00:23 < pmdata> I see vertex attributes for up to 8 texture in nv_objects.h for nv20, so I was wondering
00:24 < marcheu> well, it's just a semantic for the attribute
00:24 < marcheu> it doesn't have to actually be anything, just think of it as a parameter
00:25 < marcheu> i.e. vertex parameter 12
00:26 < Lumag> pmdata: e.g. there are at least 8 vertex parameters on nv04 :)
00:27 < marcheu> well, for nv04 all parameters match some existing hw-implemented valu
00:27 < marcheu> for nv20 it doesn't any more. I simply used the name from the nvidia extension
00:37 < pmdata> I noticed something about nv20 and register combiners: they are setup when the corresponding texture is enabled
00:38 < pmdata> so there are 4 register combiners on nv20, one for each texture
00:45 < pmdata> arg, the test_tex() miss a texture matrix
00:51 < pmdata> I think the texture matrix enable lies in 420-42c for the 4 texture units
00:53 < pmdata> it is so easy to find stuff, when the tcl stuff is initialized in the same order as nv10
00:53 < pmdata> there are just extra stuff I don't know
00:54  * pmdata should find another word for 'stuff'
00:56 < Lumag> operations?
00:57 < pmdata> too long, another try? :)
00:58 < pmdata> in (maybe not) related news, amd buys ati (and people fears about the ati linux driver state), and renouveau got another quote on nvnews forum
01:01 < Lumag> pmdata: in which forum?
01:03 < Lumag> nevermind, found it.
01:09 < pmdata> as usual, about the legacy driver
01:10 < pmdata> anyone could  try a test_tex() dump, I added some texture matrix change (nv2x or nv3x or higher)
01:19 < pmdata> does nv2x has floating point color buffer?
01:19 < marcheu> I don't think so. but that's some info that should go on the hw info page
01:20 < pmdata> in the dump I have, the line pitch is 0x1400 for a 0x200 pixel wide buffer
01:21 < marcheu> CardFeatures it's called
01:27 < marcheu> hehe it's a post by thunderbird
01:28 < Lumag> Yeah, sounds promicing :)
01:30 < marcheu> I wonder what the nvidia guys think about this. they probably know about us now
01:32 < pmdata> "first they ignore you..."
01:32 < marcheu> "then they hire you..."
01:32 < marcheu> (that's the risk we'll be running very soon)
01:34 < pmdata> seems most of the nv20 tcl engine is same as nv10 (commands have same number, except for texture/vertex stuff)
01:34 -!- stringfellow [n=stringfe@ip56503c9f.direct-adsl.nl] has joined #nouveau
01:37 < marcheu> yes, the nv20 vertex engine seems quite close to the nv10 one
01:38 < pmdata> a simple copy paste would work if you are careful :) but you still must test it
01:39 < marcheu> yep. if you put in stuff I can check it on my nv28 though
01:39 < marcheu> I'm a bit short on time now, but small tests don't take too much time
01:40 < pmdata> I would be interested in test_tex() to see where texture matrix are
01:41 < marcheu> ok, one sec
01:42 < marcheu> http://icps.u-strasbg.fr/~marchesin/nvdri/nv28_test_tex.txt
01:43 < marcheu> we need something like an array of machines with all available nvidia cards, and a test submission system :)
01:44 < pmdata> "imagine a beowulf cluster of these!" :)
01:46 -!- stringfellow_ [n=stringfe@ip56503c9f.direct-adsl.nl] has joined #nouveau
01:46 < pmdata> ok, found the texture matrix
01:57 -!- pmdata [i=patrice@ANantes-154-1-76-177.w86-199.abo.wanadoo.fr] has quit ["using sirc version 2.211+KSIRC/1.3.11"]
01:59 -!- stringfellow [n=stringfe@ip56503c9f.direct-adsl.nl] has quit [Read error: 110 (Connection timed out)]
02:00 -!- stringfellow_ is now known as stringfellow
02:15 -!- Duke` [n=gnu@ANantes-251-1-153-250.w86-203.abo.wanadoo.fr] has quit ["Fatal signal: Segmentation Fault"]
02:32 -!- Lumag [n=Lumag@chimpanzee.school.ioffe.ru] has left #nouveau []
02:35 -!- Lumag [n=Lumag@chimpanzee.school.ioffe.ru] has joined #nouveau
02:55 -!- eLShaman [n=elshaman@pac33-1-82-235-249-217.fbx.proxad.net] has quit ["Au revoir."]
03:16 -!- K [i=hazel@85-57-135-140.gij1.adsl.uni2.es] has joined #nouveau
03:17 -!- stringfellow [n=stringfe@ip56503c9f.direct-adsl.nl] has quit [Remote closed the connection]
03:36 -!- Lumag [n=Lumag@chimpanzee.school.ioffe.ru] has left #nouveau []
03:37 -!- johill [n=johannes@p5487F022.dip.t-dialin.net] has quit [Read error: 110 (Connection timed out)]
03:41 -!- Unavowed [n=silent@13.t6.ds.pwr.wroc.pl] has joined #nouveau
04:00 -!- hiyuh [n=hiyuh@ZL050248.ppp.dion.ne.jp] has joined #nouveau
04:02 -!- K is now known as PutaTV
04:04 -!- PutaTV [i=hazel@85-57-135-140.gij1.adsl.uni2.es] has quit [Remote closed the connection]
04:25 -!- colby [n=colby@adsl-69-211-27-8.dsl.chcgil.ameritech.net] has joined #nouveau
04:25 < colby> Hi
04:26 < colby> Finally fixed my trouble with autoconfig, and built the new nvclock
04:27 < colby> pciid changes but it dosnt change bit 16 of NV_PBUS_DEBUG_1
04:27 < colby> so not stereo buffers yet
04:29 < darktama> do you need to add options to xorg.conf to get stereo visuals? there seem to be a few options documented in the README
04:31 < colby> i dont thinkso. but the cvs of nvclock is only changing the pci id and not the capability bit.
04:32 -!- colby is now known as Colbert
04:33 < Unavowed> hi! i have a nv15 card -- are you interested in renouveau's output?
04:41 < Colbert> iss it a 2 pro?
04:48  * Colbert is away: Colbert
04:54 < Unavowed> yep
04:54 < Unavowed> i'll give you the link
04:54 < Unavowed> http://furio.vexillium.org/~silent/tmp/2006-07-25_renouveau_nv15_log
04:54 < Unavowed> it's the output from all the tests
04:55 < Unavowed> for every one the program was restarted
04:55 < Unavowed> uh, wait
04:55 < Unavowed> it's gts, not pro ;-)
04:56 < Unavowed> at least, that's what i remember buying
04:56 < Unavowed> lspci says it's gts/pro
05:08 < ddl> hey
05:12 < darktama> hey ddl
05:13 < ddl> hi there
05:13 < ddl> hows things going?
05:14 < darktama> hmm, not badly.. slowly working on getting my card to draw a triangle :)
05:14 < ddl> cool!
05:14 < ddl> so you kind of got it working?
05:15 < darktama> well, yes.. but only in a test app that creates the context with nvidia's libGL, then steals the FIFO off it
05:15 < ddl> ah
05:16 < ddl> so the next step is to create the context with the test app
05:17 < darktama> I have code in the ddx to try and draw, but it doesn't work.. the engine processes the commands, but nothing is drawn.. and the FIFO locks after sending a few quads
05:17 < darktama> I just scrapped it, starting fresh :)
05:17 < darktama> we can't really use a test app outside of the X server yet, because we can't switch contexts :(
05:37  * Colbert is back
05:38 < Colbert> i'm back. had to eat
05:39 < Colbert> Unavowed, thanks for the data. too bad thats its a geforce, i was hoping it was a quadro 2 pro ;)
05:44 < Colbert> it seems that i can run nvclock -Q and change the pciid without rebooting provided i modprobe -r nvidia
05:44 < Colbert> not real saf probably
05:44 < Colbert> brb
05:45 -!- Colbert [n=colby@adsl-69-211-27-8.dsl.chcgil.ameritech.net] has quit [Remote closed the connection]
05:48 < darktama> hmm, NVRM messages are quite helpful :)
05:49 -!- colby [n=colby@adsl-69-211-27-8.dsl.chcgil.ameritech.net] has joined #nouveau
05:49 -!- colby is now known as Colbert
06:09 < Colbert> darktama, you might be right about having to set up stereo visuals, looking at glxinfo dumps on nvnws from real quadros
06:09 < Colbert> gtg, brb
06:09 < darktama> :)
06:10  * Colbert is back
06:10  * Colbert is away: colby
06:12  * Colbert is back
06:23 < Colbert> NV_PBUS_DEBUG_1 (0x1084): 00205749 shoould be 00215749, on quadros bit 16 counting from 0 (i think) 
06:23 < Colbert> or am I getting confused from endiness
06:24 < Colbert> anyone have a real nvclock -D dump from a quadro?
06:26 < Colbert>               xxxx                    ^ real
06:26 < Colbert> meant: anyone have a nvclock -D dump from a real quadro?
07:13 < Colbert> brb
07:13 -!- Colbert [n=colby@adsl-69-211-27-8.dsl.chcgil.ameritech.net] has quit [Remote closed the connection]
07:36 -!- colby [n=colby@adsl-69-211-27-8.dsl.chcgil.ameritech.net] has joined #nouveau
07:37 -!- colby is now known as Colbert
07:39 < Colbert> ;-(> nvidia must be checking more than the pciid :-(
07:39 < Colbert> (II) NVIDIA(0): NVIDIA GPU Quadro FX 500/FX 600 at PCI:1:0:0
07:39 < Colbert> (WW) NVIDIA(0): Stereo is only available on Quadro cards
07:39 < Colbert> (II) NVIDIA(0): Disabling stereo.
07:42 < Colbert> they must have put in a "bug fix" to disable softquadro, going to go read unwinder's paper on softquadro
07:43  * Colbert is away: colbert
08:05 -!- gnarlin [n=gnarlin@194-144-217-11.du.xdsl.is] has joined #nouveau
08:20 < airlied> anyone working on PCI-E cards?
08:25 < Colbert> only have pci and agp
08:37 < Colbert> i think marcheu might have access to a pci-e g70 at work
09:06 < airlied> I have a PCI-E 6600 in a g5.. have to investigate renouveau under macosx :-)
09:16 < Colbert> I miss NeXT
09:33 -!- gnarlin [n=gnarlin@194-144-217-11.du.xdsl.is] has quit [Read error: 110 (Connection timed out)]
09:41 < Colbert> did apple change their kernel to bsd from mach? used to be mach with a bsd userspace on nextstep
09:44 < airlied> Colbert: pretty similiar... mach + bsd on top
09:45 < Colbert> liked playing with hardware too much to make the change to apple
09:47 < airlied> Colbert: well free makes a big difference to me :-)
09:49 < Colbert> built this machine i'm using now for nexstep, couldnt get the scsi to work, moved to linux, never looked back
09:50 < Colbert> thought of trying to cram this mb in an old Cube case though
09:50 < airlied> hehe... 
10:08 -!- Colbert_ [n=colby@adsl-69-211-3-251.dsl.chcgil.ameritech.net] has joined #nouveau
10:21 -!- Colbert [n=colby@adsl-69-211-27-8.dsl.chcgil.ameritech.net] has quit [Read error: 110 (Connection timed out)]
10:23 -!- Colbert_ is now known as Colbert
10:57 -!- Colbert [n=colby@adsl-69-211-3-251.dsl.chcgil.ameritech.net] has quit [Read error: 104 (Connection reset by peer)]
10:59 -!- colby [n=colby@adsl-69-211-3-251.dsl.chcgil.ameritech.net] has joined #nouveau
10:59 -!- colby is now known as Colbert
11:01 < Colbert> Heh, nvidia crashed X. I wonder if they added some dirty tricks for softmods like in windows
11:01 < Colbert> (WW) NVIDIA(0): WAIT (2, 1, 0x8000, 0x000050bc, 0x00004720, 0)
11:01 < Colbert> over and over
11:03 < Colbert> anyone know a way to add timestamps to xorg.log?
11:04 < darktama> I don't.. but is there a corresponding NVRM message in the kernel log to go with that WAIT?
11:05 < Colbert> checking
11:10 < Colbert> NVRM: Xid (0001:00): 6, PE0002 0c58 f2150040 01000100 00000000 001428ec
11:11 < Colbert> this was first:NVRM: Xid (0001:00): 6, PE0002 0000 3f000000 000c0aac 00000000 00000000
11:11 < Colbert> repeted 3 times
11:11 < Colbert> NVRM: Xid (0001:00): 28,  L1 -> L0
11:11 < Colbert> once
11:12 < darktama> hmm, not sure what Xid 6 is.. 13 contains useful info (well, 6 probably does too).. no idea what 28 is, I get them after each NVRM
11:12 < darktama> 13 is (channel) (object name) (object type) (command) (value) (???)
11:14 < Colbert> my card is 10de:0326 and was -Q 11 to 10de:032b
11:15 < Colbert> it might have checked the real bootstrap. ran gl-info from flightgear right before the crash
11:15 < darktama> well, NVRM usually happens when the driver does something bad.. so I guess you either found a bug that's only triggered when the card is :032b, or the driver deliberately does it when it detects a softmod
11:16 < Colbert> thats what unwinder said hthey did in windows
11:16 < darktama> then it's likely the linux driver does the same since they share most of the code
11:18 < Colbert> might have to write a kernel module to get around this
11:19 < darktama> how's that going to help?
11:20 < Colbert> like nvstrap does, have to block what the driver  really looks for, then nvidia changes it in the next version. and you change it again
11:20 < darktama> ah
11:22 < Colbert> the first softquadro just change the pci id, nvidia wrote the driver around it and made things crash to make it look like softmoding dosnt work and theres a real chiplevel difference
11:23 < Colbert> was waiting around for it to crash on me, meant to check the time i made the cheange and forgot to
11:34 < Colbert> think I'll put in my 171 and mod it to a Quadro4 550, wait for the crash and see if its a 1
11:36 < marcheu> yeah, softquadro will only work on some cards
11:36 < marcheu> I think only nv10 & nv20
11:37 < marcheu> and not even all of them
11:37 < marcheu> it's easy to know. if you run softquadro before starting X, it will change the output of lspci. otherwise you have a locked card
11:37 < Colbert> the new cvs isnt change the NV_PBUS_DEBUG_1 register though, just the id
11:40 < Colbert> i did a nvclock -D before the change , after it, and also after I started X
11:46 < Colbert> from mt Xorg.log (posted earlier) (II) NVIDIA(0): NVIDIA GPU Quadro FX 500/FX 600 at PCI:1:0:0
11:46 < Colbert> WW) NVIDIA(0): Stereo is only available on Quadro cards
11:46 < Colbert> (II) NVIDIA(0): Disabling stereo.
11:46 < marcheu> what does lspci say ?
11:47 < Colbert> so it knows its not a quadro, at X startup
11:47 < marcheu> but it's true that depending on the cards there are locked differently
11:47 < Colbert> didnt think to do that before, and I'm not modded right now
11:47 < Colbert> I will do it nexttime
11:48 < marcheu> the only card softquadro works on is my nv15. and pmdata has another almost similar nv15 on which it doesn't work...
11:50 < Colbert> the code you had changed bit 16 on NV_PBUS_DEBUG_1, but i dont see that in the nvclock cvs (not ours)
11:54 < Colbert> Could move to an old 4X.XX driver, LOL
11:56 < Colbert> brb
11:56 -!- Colbert [n=colby@adsl-69-211-3-251.dsl.chcgil.ameritech.net] has quit [Remote closed the connection]
12:01 -!- colby [n=colby@adsl-69-211-3-251.dsl.chcgil.ameritech.net] has joined #nouveau
12:01 -!- colby is now known as Colbert
12:01 < Colbert> lscpi calls it a quadro
12:02 < marcheu> hmm, things should work then, at least during primitive submission
12:03 < Colbert> not if nvidia has other checks in there
12:05 < Colbert> and deliberate crashes to discourage their coporate customers from softmoding
12:07 < Colbert> I dont know that its true, but thats what unwinder thought
12:08 < Colbert> http://www.nvworld.ru/docs/sq4e.html
12:10 -!- johill [n=johannes@p5487EFD2.dip.t-dialin.net] has joined #nouveau
12:17 < Colbert> I wish unwinder wrote more about the technical aspects instead of all the benchmarks ;-)
12:28 -!- Duke` [n=gnu@ANantes-251-1-93-215.w86-203.abo.wanadoo.fr] has joined #nouveau
12:34  * Colbert is away: Colbert
13:04 -!- marcheu [n=marcheu@lattice.u-strasbg.fr] has quit ["leaving"]
13:04 -!- marcheu [n=marcheu@lattice.u-strasbg.fr] has joined #nouveau
13:21  * Colbert is back
13:26 -!- EdB [n=EdB@ARennes-251-1-3-236.w83-195.abo.wanadoo.fr] has joined #nouveau
13:30 < Colbert> going to put in my 440mx and see how nvclock works with it. brb
13:30 -!- Colbert [n=colby@adsl-69-211-3-251.dsl.chcgil.ameritech.net] has quit [Remote closed the connection]
13:41 -!- johill [n=johannes@p5487EFD2.dip.t-dialin.net] has quit [Read error: 113 (No route to host)]
13:44 -!- colby [n=colby@adsl-69-211-3-251.dsl.chcgil.ameritech.net] has joined #nouveau
13:45 -!- colby is now known as Colbert
13:45 < Colbert> no luck, driver still knows its not really a quadro
13:52 -!- johill [n=johannes@p5487EFD2.dip.t-dialin.net] has joined #nouveau
14:13 < Colbert> i wonder if nvstrap would run in ndiswrapper? anyone try?
14:16 < marcheu> no, I didn't try :)
14:16 < marcheu> but if we knew what nvstrap does, it'd be a 5 minute job to port it using our own code
14:18 < Colbert> true. ;-)
14:19 < marcheu> hmm maybe you could mmtrace nvstrap under ndiswrapper and see what it tries to do
14:19 < marcheu> but that's getting complicated
14:21 < Colbert> haven't looked at that yet
14:24 < Colbert> dont know why I'm so damned obsessed with the buffers, lol
14:28 < Colbert> marcheu do you still have a copy of the old nvclock Q code still, I foolishy erased mine before I CO from cvs
14:29 < marcheu> http://icps.u-strasbg.fr/~marchesin/nvdri/nvclock-softquadro.tgz ?
14:29 < Colbert> thanks
14:40 -!- johill [n=johannes@p5487EFD2.dip.t-dialin.net] has quit [Read error: 113 (No route to host)]
15:23 -!- Colbert [n=colby@adsl-69-211-3-251.dsl.chcgil.ameritech.net] has quit [Remote closed the connection]
17:29 -!- hiyuh [n=hiyuh@ZL050248.ppp.dion.ne.jp] has quit ["Leaving"]
17:37 -!- jkolb [n=jkolb@8.8.128.16] has joined #nouveau
19:28 -!- EdB [n=EdB@ARennes-251-1-3-236.w83-195.abo.wanadoo.fr] has quit [Remote closed the connection]
19:29 -!- EdB [n=EdB@ARennes-251-1-3-236.w83-195.abo.wanadoo.fr] has joined #nouveau
19:42 -!- stringfellow [n=stringfe@ip56503c9f.direct-adsl.nl] has joined #nouveau
20:10 -!- jkolb [n=jkolb@8.8.128.16] has quit ["Leaving"]
20:39 -!- EdB_ [n=EdB@ARennes-251-1-63-32.w81-53.abo.wanadoo.fr] has joined #nouveau
20:50 -!- Anne_ [n=EdB@ARennes-251-1-33-96.w81-250.abo.wanadoo.fr] has joined #nouveau
20:53 -!- EdB [n=EdB@ARennes-251-1-3-236.w83-195.abo.wanadoo.fr] has quit [Read error: 110 (Connection timed out)]
21:07 -!- EdB_ [n=EdB@ARennes-251-1-63-32.w81-53.abo.wanadoo.fr] has quit [Read error: 110 (Connection timed out)]
21:13 -!- pmdata [i=patrice@ANantes-154-1-21-66.w81-53.abo.wanadoo.fr] has joined #nouveau
21:14 < pmdata> hello
21:14 < pmdata> marcheu> I got nvclock -Q to work, I did not know there was an extra param to add
21:14 < pmdata> however, no change in glxinfo output
21:14 < pmdata> the pci id has not been changed to a quadro2 card, only the register
21:15 < marcheu> it depends what version you used. my hack doesn't require it, the official version does
21:15 < marcheu> it seems your card is locked
21:15 < pmdata> rev a4
21:16 < marcheu> anyway, one of these days I'll get my hands on a real quadro at the university, I'll run benchmarks and we'll be done with that
21:16 < marcheu> I know there's at least a quadro FX somewhere
21:17 < marcheu> s/benchmarks/renouveau/
21:21 < pmdata> I read the article on nvworld.ru about softquadro
21:22 < pmdata> at least for quadro2, there are not much more, besides drawing lines
21:22 < pmdata> maybe also the gl[draw|read]pixels that would use dma on it instead of cpu
21:24 < marcheu> nope, gldraw/readpixels is already using dma on geforces
21:24 < marcheu> but ATI does that with radeon/firegl
21:24 < marcheu> only firegls have fast read/drawpixels
21:46 -!- jkolb [n=jkolb@wsi-128-16.wsi.com] has joined #nouveau
21:50 < pmdata> just found more npot texture stuff while reading nv_texture_rectangle dumps
22:09 -!- zaphands [n=zaphands@87.69.23.196.cable.012.net.il] has quit [Remote closed the connection]
23:21 < pmdata> ah, glreadpixels uses nv_memory_to_memory_format, and some dma_to and dma_from object
23:28 < pmdata> hum, pitch_in is negative, and 0x324 seems to be the format
23:28 < pmdata> time to iterate through the formats
23:53 < marcheu> negative pitch usually means vertical flip
23:53 < pmdata> ah
23:53 < marcheu> just like the old amiga blittes :)
23:53 < marcheu> blitter*
23:53 < marcheu> yeah, it shows you're an atari guy :p
23:54 < pmdata> yeah, I was thinking myself that having disassembled m68k code in my youngness was useful for renouveau
23:54 < pmdata> finding some pattern, guess what it could be, etc...
23:59 < pmdata> http://www.nvworld.ru/docs/sq2e.html interesting to read
--- Log closed Срд Июл 26 00:00:58 2006
