Nightwulf|work: hi all
Obscene_CNN: goody!! some fresh new patches
udovdh: any info on progress, planning, features, etc?
udovdh: recent updates that is?
udovdh: stuff that we could test, help with?
Obscene_CNN: you could test a patch of mine
udovdh: will have a look
Obscene_CNN: I show a slight improvement in speed with it. Please bench mark be fore and after to see if you notice anything
udovdh: ehm.. where is the patch itself?
udovdh: and how much does it help on rv630?
udovdh: would be interesting
Obscene_CNN: maybe thats why I haven't got much feed back on it
udovdh: I am looking for details on progerss for rv630
udovdh: could be, yes
Obscene_CNN: hmmm.... I attached a patch when I mailed it to the list
udovdh: maybe it was scrubbed?
Obscene_CNN: could be...
udovdh: just put it online somewhere and post a link?
udovdh: maybe pastebin it when it's not big?
Obscene_CNN: its at http://www.pastebin.ca/1244219
udovdh: how can I benchmar before and after?
Obscene_CNN: I have used glxgears and gtkbench
udovdh: i what rpm could gtkbench be?
udovdh: in what...
udovdh: and how does the patch speed up the driver?
udovdh: I see you define a function of sorts
udovdh: and have routines use that function
Obscene_CNN: I changed a function that doesn't get inlined into a macro
Obscene_CNN: well functions that don't get inlined often anyway
Obscene_CNN: I left the original functions intact
Obscene_CNN: hmmm... I don't know what rpm gtkbench would be in. I compile mine from source
udovdh: thanks, will have a look into it!
udovdh: maybe post the url to that mailinglist
Obscene_CNN: I'm going to try and find a more perminant place
adirex: Hi everyone, I'm trying to setup HDMI audio output for Mobility Radeon HD2400. I have the latest radeonhd driver installed from source, alsa compiled with all codecs and all cards, but when I type "aplay -l" the HDMI output is not listed. Any ideas on how to solve this?
purplep: Hello, I had a question about rhdAtomInit() in rhd_atombios.c. I'm trying to trace the initialization path of the driver, and the various ways the driver looks for the GPU BIOS. Is it all in rhd_atombios()?
purplep: I mean rhd_atombios() and RHDReadPCIBios(), since atombios() sometimes calls RHDReadPCIBios().
purplep: egbert: You were very helpful with my problem on Saturday. I've actually gotten the dual card setup working now, with some trouble.
purplep: egbert: I have to play a trick on rhd_atombios() to get it to use the right initialization path, so I'm wondering if rhd_atombios() should be tweaked a bit.
purplep: egbert: I'm going to go ahead and explained what worked, if you have a chance to read this and respond that would be great.
purplep: As I mentioned on Saturday, I have a x86_64 MB with a built-in RadeonHD 3200 and a PCIE HD 2600 XT. I have the MB BIOS setup to make the onboard GPU primary, because I understand it can otherwise be difficult to access the onboard GPU BIOS, since it is part of the system BIOS.
purplep: The machine boots with the onboard GPU active, and the PCIE GPU disabled but present on "lspci". Xorg and radeonhd from Fedora updates-testing/rawhide (xorg-x11-drv-radeonhd-1.2.3-1.3.20081029git.fc9.x86_64) don't get anywhere with the PCIE GPU since it is disabled.
purplep: This is apparently due to libpciaccess breakage in Xorg server. So as a workaround for this, I do "echo 1 > /sys/bus/pci/devices/.../rom", followed by "setpci -s ... ROM_ADDRESS=xxxx0001" and "setpci -s ... COMMAND=2".
purplep: This allows both cards to be accessed by radeonhd, but radeonhd doesn't get the correct initialization path usually.
purplep: Prior to doing the setpci commands, Xserver would see two cards and consider the onboard GPU primary. radeonhd, in rhdAtomInit() uses xf86IsEntityPrimary() to determine whether to try getting the BIOS copy from the legacy VBIOS location. This works for my onboard GPU.
purplep: After the setpci commands, Xorg no longer sees the onboard GPU as primary, it says "More than one possible primary device found". This causes rhdAtomInit() to not check the legacy VBIOS location, and it no longer finds the onboard GPU BIOS.
purplep: Now if there is only one card in the system, Xorg seems to always consider it primary. So I can work around this problem by running Xorg with IsolateDevice for the onboard GPU.
purplep: On the secondary GPU, I have the opposite problem. I can get it working after the setpci commands, but only if rhdAtomInit() calls RHDReadPCIBios().
Zayec: adirex: you still here?
purplep: So I have to use a xorg.conf without any IsolateDevice, and with the Screen section referencing the PCIE GPU occuring before the Screen section referencing the onboard GPU. This allows X to work on the PCIE GPU, and it then fails and ignore on the onboard GPU.
Zayec: adirex: i think HD2400 simply can't play sound over HDMI
Zayec: doesn't have needed piece of hardware
adirex: Zayec, thanks for the info. I thought it can... I should try to find that out in windows
adirex: Zayec, it does support: http://ati.amd.com/products/mobilityradeonhd2400/specs.html
adirex: "Integrated HD audio controller with multi-channel (5.1) AC3 support, enabling a plug-and-play cable-less audio solution "
adirex: isn't this what I am trying to enable?
purplep: So the upshot is that it seems like rhdAtomInit() is using xf86IsEntityPrimary() to test whether the GPU is the boot video device, and at the moment that test isn't reliable.
purplep: So is this a side effect of the libpciaccess breakage, or is there a different test that should be used to determine if the GPU is the boot video device?
Zayec: adirex: interesting...
adirex: Zayec, yeah, I double checked now that the drivers loaded are the latest ones and still, the device is not listed
Zayec: in case aplay -l doesn't list this audio device on GPU, it's alsa problem I think
Zayec: do you use up to date alsa?
Zayec: that may mean lack of support for this hadrware in alsa... however wait for someone to confirm this
Zayec: probably would be nice to report this to alsa in case of confirming
adirex: how can I know if radeonhd driver correctly exposes the audio device?
Zayec: radeonhd is the last step in playing sound
yangman: adirex: it doesn't. exposure as sound device is entirely ALSA
Zayec: radeonhd just put sound signal between video signal refreshes... saying that easy way :)
Zayec: alsa first has to support it
adirex: I c
adirex: I have one more question
egbert: purplep: have you tried the fix i put in the bugzilla?
adirex: yangman, I compiled and installed the alsa-lib, alsa-utils, alsa-driver. Is is all that I need from the pack of available software on alsa website?
purplep: egbert: No, I didn't see it, I guess because the bug is still listed as RESOLVED, so I didn't get emailed.
Zayec: adirex: alsa is splitted in two parts
Zayec: userspace and kernel space
Zayec: you need quite new kernel as well
yangman: adirex: afaik, yes. I'm not the best person to help you with this. I don't actually own any HDMI-capable radeonHD hardware
egbert: purplep: strange, it should get emailed regardless of the resolution state.
adirex: Zayec, I have the kernel that came with Ubuntu 8.10, so it's quite new... How do I know which kernel exactly I should have?
adirex: yangman, thanks for all your help anyway :)
egbert: i've changed the detection: now it looks for the vram size.
purplep: egbert: I see that.
Zayec: 8.10 (means 2.6.27) should be absolutely fine
purplep: egbert: Perhaps it is my mistake with bugzilla, I made a comment on someone else's bug and assumed I would get added to the CC: list.
Zayec: adirex: i think you'll have to fill bug on alsa bugzilla
purplep: Perhaps I need to add myself manually.
egbert: purplep: maybe i should check for !isprimary || !vramsize.
egbert: purplep: that
egbert: s possible
purplep: egbert: With your patch I'll still need to do the setpci commands, as that is a separate problem?
egbert: you should add yourself to the cc.
adirex: Zayec, will do. Do you have any guidelines how I should do that and what info I should provide?
egbert: purplep: yes.
purplep: egbert: Done, added to CC: Sorry about the oversite, I wish I had seend your patch earlier. :-)
egbert: purplep: no problem!
egbert: purplep: i'm not 100% happy with such a solution. it looks like more and more workarounds are required inside the driver.
egbert: it makes the driver messy.
Zayec: adirex: i don't ... don't know much abous alsa
purplep: egbert: BTW, this problem that requires the need to do the setpci commands, what component should that be reported as an error against?
adirex: np, thanks anyway :)
adirex: will dig that out
Zayec: adirex: for sure, attach output of lspci and kernel version
Zayec: adirex: hope they will able to solve that problem :)
purplep: Leaving for a bit to test that patch.
ndim: Regarding that setpci magic... do we have a prominent link on the wiki page to an appropriate bugzilla or something simiar, so that people can find that?
purplep: ndim: Not that I'm aware of. I've spent several days trying to figure out this dual card problem. It's been quite vexing.
purplep: egbert: I tried that patch you referred me to. The short answer is that it worked the first time I run X, but after that if I try to run X again, it crashes.
purplep: egbert: The long answer is on
purplep: egbert: Just FYI, I do have both cards working separately now, by working around rhdAtomInit() as I described earlier. If you want to get this working cleanly, I'm happy to do any testing.
purplep: egbert: Thanks for your help with this.
egbert: purplep: i went ahead and added something to the QA section on the wiki.
egbert: purplep: the issue is: the chip is POSTed after the first run.
egbert: so the vram address is correct. still the bios at 0xc0000 isn't the correct one.
egbert: purplep: there is nothing what i can do from the driver side.
purplep: egbert: I took a look at the new wiki section, that will help people to get a secondary card visible to Xserver.
purplep: egbert: But radeonhd will still have trouble finding the right BIOS image, sometimes finding none and sometimes finding the image of the wrong card.
purplep: egbert: It seems like there must be a way to improve the BIOS image detection in rhdAtomInit(). Did you have a chance to read the comments I logged on bug 16264 on bugs.freedesktop.org?
Obscene_CNN: heh, yet another 780 chip but then again maybe not. http://www.theinquirer.net/gb/inquirer/news/2008/11/01/amd-introduces-780g-sorry-780e
Determined1: whattup niggas