"Ask ATI" dev thread

[nanonyme]

*shrug* Of course one solution would be to have the "glue" be fully GPL-compliant and integrated into kernel (and thus kernel developers would be forced to maintain it unless they wanted to prove themselves liers), then have the binary blobs distributed separately, like firmware nodadays is.
I personally believe one of the major reasons why Linux is still so small in desktop market is the fact that develoepers fail to understand that most end-users will not care if everything they use is Free - as long as it works. It is the developers' battle for their own Intellectual Property that we're currently seeing, really. It is their fair right to do so though, of course. Every company does so too.

[mirv]

Users may not care much about the Free aspect - but the kernel devs sure do.
That aside, the ability to change an interface to clean up drivers / do things a (hopefully!) better way is very useful from a programming perspective.

[energyman]

firmware and drivers are two completly different subjects.

[Melcar]

Any plans on implementing profiles to CCCLE like in the Windows driver? Specifically for 3D settings (AA, AF, etc.). This would particularly useful for games (or for any 3D app. really), since changing settings on the fly is far better than having to load up CCCLE. Something that the user can manually load would be enough (no reason to map individual applications).

[Mazur]

"Ask ATI" dev thread. Ok, if this is thread where can I ask something then, Are you going to support us better or still mess around?

[mirv]

"Ask ATI" dev thread. Ok, if this is thread where can I ask something then, Are you going to support us better or still mess around?

In what fashion do you think they're not? They're releasing docs to support open source drivers, actively helping write them (more than they said they would) and working on a more "feature complete" binary driver. So a bit more info of what you mean would be good.

[Mazur]

In what fashion do you think they're not? They're releasing docs to support open source drivers, actively helping write them (more than they said they would) and working on a more "feature complete" binary driver. So a bit more info of what you mean would be good.

That kernel 2.6.29 (stable) is not even supported whereas 2.6.30 is on its way and nothing was even said that it will be supported. This month's driver (9.5) was released and it was not even better than 9.4. It looks like nothing's changed from two years. I wanted to buy full AMD computer with my AMD Processor and ATI graphics but if nvidia is much better in this then ATI will lost their customers very quickly. True is that until ATI will release good driver all customers will be gone. Sorry for my english.

[Kano]

I would like to know if ATI has plans to test new kernels earlier, because even with all possible tricks you need an extra export in the 2.6.30 tree, would be much more clever for an ati dev to submit that as patch, because some distros are not willig to add extra exports for fglrx - but you can not change old drivers, like 9-3 when there will be no update for legacy cards. the correct way would be of course updating the legacy driver with full 26.30 support.

diff --git a/arch/x86/mm/tlb.c b/arch/x86/mm/tlb.c
index 61ddfa0..3d26c0b 100644
--- a/arch/x86/mm/tlb.c
+++ b/arch/x86/mm/tlb.c
@@ -279,6 +279,7 @@ void flush_tlb_page(struct vm_area_struct *vma, unsigned long va)

        preempt_enable();
 }
+EXPORT_SYMBOL(flush_tlb_page);

 static void do_flush_tlb_all(void *info)
 {

[mirv]

That kernel 2.6.29 (stable) is not even supported whereas 2.6.30 is on its way and nothing was even said that it will be supported. This month's driver (9.5) was released and it was not even better than 9.4. It looks like nothing's changed from two years. I wanted to buy full AMD computer with my AMD Processor and ATI graphics but if nvidia is much better in this then ATI will lost their customers very quickly. True is that until ATI will release good driver all customers will be gone. Sorry for my english.

There are patches to allow for the latest drivers to work with 2.6.29 - as for 2.6.30, well that I can't say for myself, but I imagine it would work too. AMD nor Nvidia are in the habit of saying much about a driver before it is released, btw. And there is a good article on AMD's driver development cycle - chances are there weren't many changes between 9.4 and 9.5 internally. However, I must say that if you're basing all experience off those two drivers, it doesn't do much for your arguement. And much has changed in the past two years for the average end consumer - even though AMD have also always stated that their primary support belongs to the workstation people (that's where the money is after all).
And no, nvidia is not much better. I hear a lot of people say "nvidia is better" but it's usually just because they have compiz running nicer for them. And that's it. The nvidia drivers have their fair share of other bugs (go to the nvidia forums and poke around a bit for examples) - and their drivers really haven't moved much in the past couple of years. All I'm saying is: you can expect to have problems with an nvidia card as well.
But, as you appear to have had problems with the drivers (I'm assuming fglrx) - what problems exactly did you have (other than not supporting kernel 2.6.29)?

[Kano]

And you have direct comparison for running games with wine with nvidia and ati cards, do you?