[Intel-gfx] frame buffer compression and 915GM

[SD]

--- On Sun, 5/29/11, Ben Widawsky <ben at bwidawsk.net> wrote:

> From: Ben Widawsky <ben at bwidawsk.net>
> Subject: Re: [Intel-gfx] frame buffer compression and 915GM
> To: "SD" <sd.domrep at yahoo.com>
> Cc: intel-gfx at lists.freedesktop.org
> Date: Sunday, May 29, 2011, 2:47 AM
> On Sat, May 28, 2011 at 01:50:12AM
> -0700, SD wrote:
> > 
> > 
> > --- On Fri, 5/27/11, Ben Widawsky <ben at bwidawsk.net>
> wrote:
> > 
> > > From: Ben Widawsky <ben at bwidawsk.net>
> > > Subject: Re: [Intel-gfx] frame buffer compression
> and 915GM
> > > To: "SD" <sd.domrep at yahoo.com>
> > > Cc: intel-gfx at lists.freedesktop.org
> > > Date: Friday, May 27, 2011, 11:48 PM
> > > On Fri, May 27, 2011 at 01:51:53AM
> > > -0700, SD wrote:
> > > > OK, thank you. I checked it, and I got
> unexpected
> > > answer -
> > > > 
> > > > FBC unsupported on this chipset.
> > > 
> > > I'd need to know exactly which chipset you have.
> For
> > > example. Ironlake
> > > mobile, we disabled FBC because it was actually
> consuming
> > > more power.
> > > 
> > > You can check your specific chipset, and look at
> the
> > > has_fbc field here:
> > > drivers/gpu/drm/i915/i915_drv.c
> > > 
> > > Ben
> > > 
> > 
> > Thanks.
> > I checked it out.
> > You threw FBC from i915_drv since 2.6.33, because in
> 2.6.32 there was not any sign of pci_id detection. And since
> then I got awful screen picture quality, because I start to
> test with Fedora 12 and then jumped to Fedora 13 with 33
> kernel, and since then never had normal picture - always
> flickers.
> > 
> > I know you are all very clever here and are thinking
> about some Watts, but I am, like a user, thinking about my
> eyes first and want to tell you that you did mistake.
> > May be FBC uses more power (I am not sure about it, as
> long 3D performance on i915gm is "nothing" anyway), but it
> looks like FBC helps to produce nice and smooth picture on
> laptop LSD.
> 
> I don't think FBC should have any positive impact on the
> display, but I
> suppose that is a subjective thing.
> 
> You should see Jesse's commit in keithp's drm-intel-next
> (c1a9f047638b27e481d097910604316b8a0d132b). If you're
> capable of writing
> code, I'd suggest you do the same thing for
> intel_i915gm_info where fbc
> can be enabled, but is off by default. If not, hopefully I,
> or someone
> else will get around to it eventually.
> 
> Ben
> 



Actually, I prefer if some of you can add module parameter to control FBC from the kernel command line, as it was popped up somewhere about 2.6.40 and, which is more important, was previously in xorg.conf as - Option "FramebufferCompression" "boolean"
Module parameter for grub would be the best way, because as long as you can't test video output thoughtfully by yourself, let allow users to decide.

But, to be able to enable FBC, I suppose 

My card id:
[    0.100352] pci 0000:00:02.0: [8086:2592] type 0 class 0x000300
[    0.100431] pci 0000:00:02.1: [8086:2792] type 0 class 0x000380


those:

static const struct intel_device_info intel_i945g_info = {
         .gen = 3, .has_hotplug = 1, .cursor_needs_physical = 1,
         .has_overlay = 1, .overlay_needs_physical = 1,
 };
 static const struct intel_device_info intel_i945gm_info = {
         .gen = 3, .is_i945gm = 1, .is_mobile = 1,
         .has_hotplug = 1, .cursor_needs_physical = 1,
         .has_overlay = 1, .overlay_needs_physical = 1,
         .supports_tv = 1,
 };

should have at least: ".has_fbc = 1/0" as initial state, otherwise it always will be unsupported.


And please, do it yourself, as it is still yours responsibility. I am just unhappy user.


Sincerely.