On 2 December 2015 at 16:33, Alex Deucher <alexdeucher@xxxxxxxxx> wrote: > On Wed, Dec 2, 2015 at 6:15 AM, Emil Velikov <emil.l.velikov@xxxxxxxxx> wrote: >> On 30 November 2015 at 19:47, Alex Deucher <alexdeucher@xxxxxxxxx> wrote: >>> On Sat, Nov 28, 2015 at 4:01 PM, Pavel Machek <pavel@xxxxxx> wrote: >>>> >>>> Thinkpad T40p needs agpmode 1. >>>> >>>> Signed-off-by: Pavel Machek <pavel@xxxxxx> >>> >>> Seems odd that this wouldn't have been found earlier given how popular >>> thinkpads are. Applied. Thanks, >>> >> Wondering if it wouldn't be better to apply these restrictions within >> the AGP driver. I have a vague recollection that (at least?) some of >> these are chipset limitations/bugs ? > > I think it's probably better in the GPU drivers. AGP was always > problematic. There are a number of quirks that were necessary for > some users but not others on the same hardware. Additionally, the > combination of vendors or even driver versions/behaviors were > problematic. E.g., with UMS, the drivers didn't use much gart since > there was just a single static allocation. Once we switched to KMS, > lots of new bugs surfaced. Chipset A plus GPU vendor B worked fined, > but chipset A with GPU vendor C was problematic. > Looking at how many cases with identical vendor/device (yet differing sub vendor/device) are there, plus some of these also available in nouveau I'm wondering if it's truly a case of things "working" for some as opposed to different test(ing procedure) being applied. That assumption comes along nicely with how 'delayed' this report is. That said I'm not pushing on anything, just pointing out the pattern I see. Emil _______________________________________________ dri-devel mailing list dri-devel@xxxxxxxxxxxxxxxxxxxxx http://lists.freedesktop.org/mailman/listinfo/dri-devel
