[Intel-gfx] [PATCH i-g-t 2/2] lib/igt_kms: Short description between Intel/DRM terminology.
Daniel Vetter
daniel at ffwll.ch
Tue Jan 19 11:45:23 PST 2016
On Tue, Jan 19, 2016 at 05:18:09PM +0200, Ville Syrjälä wrote:
> On Tue, Jan 19, 2016 at 05:04:33PM +0200, Marius Vlad wrote:
> > On Fri, Jan 15, 2016 at 02:46:45PM +0200, Ville Syrjälä wrote:
> > > On Fri, Jan 15, 2016 at 11:06:42AM +0200, Marius Vlad wrote:
> > > > lib/igt_kms: Briefly describe Intel-to-DRM mapping between pipes, encoders and
> > > > connectors.
> > > >
> > > > Signed-off-by: Marius Vlad <marius.c.vlad at intel.com>
> > > > ---
> > > > lib/igt_kms.c | 82 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
> > > > 1 file changed, 82 insertions(+)
> > > >
> > > > diff --git a/lib/igt_kms.c b/lib/igt_kms.c
> > > > index c7a0b77..caa8837 100644
> > > > --- a/lib/igt_kms.c
> > > > +++ b/lib/igt_kms.c
> > > > @@ -68,6 +68,88 @@
> > > > * functions have all igt_ prefixes. This part is still very much work in
> > > > * progress and so also lacks a bit documentation for the individual functions.
> > > > *
> > > > + * Intel/DRM terminology and display connections:
> > > > + *
> > > > + * Intel documentation describes the road from memory to an output connector as
> > > > + * follows:
> > > > + *
> > > > + * |[!<-- language="C" -->
> > > > + * .--------. .-------. .-------------. .-----.
> > > > + * | Memory |---->| Pipes |---->| Transcoders |---->| DDI |
> > > > + * '--------' '-------' '-------------' '-----'
> > > > + * ]|
> > > > + *
> > > > + * Pipes represent the front-end of the display contain the planes, blending,
> > > > + * and color correction, while the transcoders contain timing generators,
> > > > + * encoders, A/V mixers and PSR (Panel-Self Refresh) controllers. Finally the
> > > > + * DDI represent the connectors attached to the display.
> > > > + *
> > > > + *
> > > > + * In DRM we have the following:
> > > > + *
> > > > + * |[!<-- language="C" -->
> > > > + * .-----------. .-------. .-----------. .------------.
> > > > + * | Framebuff |---->| Pipes |---->| Encoders |---->| Connectors |
> > > > + * '-----------' '-------' '-----------' '------------'
> > >
> > > "Framebuff"?
> > >
> > > Also should this say crtc instead of pipe?
> > Right, pipe indeed.
> > >
> > > > + * ]|
> > > > + *
> > > > + *
> > > > + * The frame buffer ties a reference to a memory object and provides a pointer
> > > > + * to the actual data.
> > > > + *
> > > > + * The pipe is used to set the display mode, timings and gamma tables. On some
> > > > + * hardware models this is tied with the transcoder. In DRM-parlance this is
> > > > + * referred as a CRTC.
> > > > + *
> > > > + * Each pipe has multiple planes. On older hardware these planes where known as
> > > > + * primary plane, overlay/sprite plane, and cursor plane. From GEN9 (SKL/BXT)
> > > > + * each pipe has three planes and a cursor plane.
> > >
> > > Not quite true.
> > >
> > > > Each plane can be used as a
> > > > + * primary, as a sprite or as an overlay plane. The planes are the ones that
> > > > + * retrieve the pixels from memory and pushes them to the encoder.
> > > > + *
> > > > + * A pipe prior to GEN9:
> > >
> > > Really more like g4x-bdw. Before g4x it was totally different, and gen4 was
> > > sort of mix of both the old and the new. And vlv/chv have two sprites on
> > > each pipe.
> > >
> > > So given all the variations in the hardware, and the fact that it keeps
> > > changing all the time, I'm not convinced there's any point in
> > > documenting this in igt. It'll get stale real quick, and there are
> > > efforts to decouple igt from i915 as much as possible, so next thing you
> > > know someone else will want to docuemnt their favorite hardware here as
> > > well.
> > >
> > > So if we do want to document this stuff, then I think it should be
> > > somewhere in the kernel modeset code (around crtc/plane init I suppose).
> >
> > Yes, it might make more sense to be in the kernel. What I wanted to achieve
> > with this small intro was to bridge the gap between DRM terminology and
> > the one in Intel 01.org PRM for igt. For a first timer is rather tedious to
> > identify software/hadware abstraction when looking at the PRMs and when
> > looking in igt -- I know for a fact that I'm still having trouble
> > identifying those parts.
> >
> > Daniel has already suggested to add some documention between
> > DRM-to-Intel Mappings in the kernel.
> >
> > Do you suggest that we should document all platforms?
>
> I'd either keep it fairly generic, or document all of them. What you
> had here seemed like something in between.
Yeah I still think bridging the terminology gap in igt/drm would be good,
but indeed a bit more high-level might work better. And I agree with Ville
that the detailed docs are probably best in the kernel, next to the actual
paltform support code.
-Daniel
>
> >
> >
> > >
> > > > + *
> > > > + * |[!<-- language="C" -->
> > > > + * .--------.
> > > > + * | Memory | .--------.
> > > > + * | |------------>| Cursor |------>...
> > > > + * | | '--------'
> > > > + * | |
> > > > + * | | .--------.
> > > > + * | |-------->| Sprite |---------->...
> > > > + * | | '--------'
> > > > + * | |
> > > > + * | | .---------.
> > > > + * | |----->| Primary |------------>...
> > > > + * | | '---------'
> > > > + * '--------'
> > > > + * ]|
> > > > + *
> > > > + * A pipe with universal planes:
> > > > + *
> > > > + * |[!<-- language="C" -->
> > > > + * .--------.
> > > > + * | Memory | .--------.
> > > > + * | |------------>| Cursor |------>...
> > > > + * | | '--------'
> > > > + * | |
> > > > + * | | .-------.
> > > > + * | |-------->| Plane |---------->...
> > > > + * | | '-------'
> > > > + * | |
> > > > + * | | .-------.
> > > > + * | |----->| Plane |------------>...
> > > > + * | | '-------'
> > > > + * '--------'
> > > > + * ]|
> > > > + *
> > > > + * The encoder is used to convert, depending on the output, pixels from pipes
> > > > + * to signals understood by the connector.
> > > > + *
> > > > + * The connector will connect to the output display. This contains information
> > > > + * about the attached display such as EDID, DPMS and information about modes
> > > > + * supported by the display.
> > > > + *
> > > > * Note that this library's header pulls in the [i-g-t
> > > > * framebuffer](intel-gpu-tools-Framebuffer.html) library as a
> > > > * dependency.
> > > > --
> > > > 2.6.2
> > > >
> > > > _______________________________________________
> > > > Intel-gfx mailing list
> > > > Intel-gfx at lists.freedesktop.org
> > > > http://lists.freedesktop.org/mailman/listinfo/intel-gfx
> > >
> > > --
> > > Ville Syrjälä
> > > Intel OTC
>
>
>
> --
> Ville Syrjälä
> Intel OTC
--
Daniel Vetter
Software Engineer, Intel Corporation
http://blog.ffwll.ch
More information about the Intel-gfx
mailing list