I came across the following kernel-2.6.27 commit in ALSA:
http://git.kernel.org/?p=linux/kernel/git/torvalds/linux-2.6.git;a=commit;h=...
I'm curious if this overlaps/how this combines with your work?
txs and LieGrue,
strub
--- Michael Cree
Von: Michael Cree
Betreff: Re: [radeonhd] HDMI support including Audio An: "Christian König" CC: "radeonhd" Datum: Freitag, 10. Oktober 2008, 1:44 On 10/10/2008, at 11:31 AM, Christian König wrote: Am Freitag, den 10.10.2008, 09:46 +1300 schrieb Michael Cree:
Right, I have also tried this patch (plus patch number 0002). System is Alpha architecture, with HD 2400 Pro card, RV610, which admittedly doesn't match the list of compatible hardware given for the patch, but thought I'd try anyway. In theory this path should work on RV610 just fine, it was tested successfully on RV630 wich is AFAIK not so much different.
Good.
I verified that the sound files (signed 16bit, 44.1kHz, pcm) would play through my other sound devices (on board sound chip, and a cheap surround sound card), but aplay would hang and not return when playing via the ATI HDMI sound device (was breakable with ctrl-c). Examination of /proc/interrupts revealed that interrupts were not clocking up for the ATI HDMI interrupt. This is evidence that things are not working as they should. That's odd, even without this patch interrupts are initialized at bootup by atombios. But since your system is alpha based i don't even know how video gets initialized, does it use some kind of x86 emulation or does this rely on radeonhd to fully setup video stuff?
I've got the video card plugged in a miata variant alpha and the SRM (the alpha equivalent of a BIOS) does have x86 emulation for the purposes of posting video cards, but it is far too old and can't post the RV610 video card. I bring the system up on serial port console. My understanding is it is the X server that contains the x86 emulation code that POSTs the video card; at least the display powers up and the Gnome login window appears when X is started. This is a system that would definitely benefit from having the video card initialisation code in the kernel rather than in X! The initialisation of the card is not perfect; the display appears with all the colours all mixed up. Running 'xgamma -gamma 1.0' repeatedly, slowly fixes up the colours until the display is quite useable and seems to be working in all other aspects.
I do have another slightly newer Alpha on which the SRM can POST the RV610 video card, but other issues arise with using the card, and its my main production system, so require better stability and features than radeonhd can provide at the moment.
Could you make of register dump with rhd_dump of the 7300-73ff range? Could you also send me a log with X started with the "-logverbose 7" option? Maybe i can do something about this.
Not at the moment as I am not at the machine. I'll try and have a go at this later today when I get home.
I also noted a number of "unaligned trap" warnings appearing in the kernel log due to the Xorg process. I don't recall seeing them prior to this date, but will need to investigate further to verify whether they are due to the latest radeonhd plus audio. The unaligned traps, on Alpha, are not a show stopper, it just means the kernel emulated a misaligned memory access the CPU couldn't perform. The main detriment is a large performance hit. But I understand that on the Sparc platform misaligned accesses result in shutting down the process that generated them so ultimately, one should avoid them, if you want generally portable code. I will investigate further to verify whether the radeonhd driver is causing them, or whether it is something else in Xorg.
Thanks for the clarification of DVI versus HDMI.
Michael.
-- To unsubscribe, e-mail: radeonhd+unsubscribe@opensuse.org For additional commands, e-mail: radeonhd+help@opensuse.org
-- To unsubscribe, e-mail: radeonhd+unsubscribe@opensuse.org For additional commands, e-mail: radeonhd+help@opensuse.org