Re: [PATCH] drm/fourcc: add LINEAR modifiers with an exact pitch alignment

[Simona Vetter <simona.vetter@ffwll.ch>]

On Tue, Jan 21, 2025 at 02:21:57PM -0500, Marek Olšák wrote:
> On Mon, Jan 20, 2025 at 1:41 PM Simona Vetter <simona.vetter@ffwll.ch>
> wrote:
> 
> > On Mon, Jan 20, 2025 at 08:58:20AM +0100, Thomas Zimmermann wrote:
> > > Hi
> > >
> > >
> > > Am 18.01.25 um 03:37 schrieb Marek Olšák:
> > > [...]
> > > >
> > > > 3) Implementing DRM_FORMAT_MOD_LINEAR as having 256B pitch and offset
> > > > alignment. This is what we do today. Even if Intel and some AMD chips
> > > > can do 64B or 128B alignment, they overalign to 256B. With so many
> > > > AMD+NV laptops out there, NV is probably next, unless they already do
> > > > this in the closed source driver.
> >
> > I don't think this works, or at least not any better than the current
> > linear modifier. There's way too many users of that thing out there that I
> > think you can realistically redefine it.
> >
> 
> DRM_FORMAT_MOD_LINEAR was redefined on PC a long time ago to mean 256B
> pitch alignment because of laptops with AMD+Intel. Drivers redefined it
> because that's what happens when it's under-defined. As you say,
> DRM_FORMAT_MOD_LINEAR can't be removed, but then it can't work with any
> other pitch alignment on all PC hw either, so there is no other choice.
> 
> The options for PC are either a new parameterized linear modifier (with
> properly defined addressing and size equations) or DRM_FORMAT_MOD_LINEAR
> with 256B pitch alignment. There is no 3rd option. Even if you totally
> disregard AMD, you won't get it below 128B or 64B on the rest of PC hw
> anyway, and that's the same problem.

Ah I missed that, but just checked in mesa, happened in 2021 apparently.
Would be really good to document this in the kernel's drm_fourcc.h
comments as the defacto rule. It's better if the docs reflect actual
reality, whatever that is.
-Sima
-- 
Simona Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch