X driver for omap DSS

Hi,

Kalle Vahlman has been working on a X driver for omapfb[2] and there
is a package available for testing[1]. It currently has working YUV2
and UYVY modes, provided your video has a width with a multiple of 16.
Please help improving this driver :slight_smile:

regards,

Koen

[1] http://dominion.thruhere.net/koen/OE/xf86-video-omapfb_0.0.1+r4+gitfc5958ba70ad76aadcf2993d7713ee13fe262b33-r4_armv7a.ipk
[2] http://gitweb.pingu.fi/?p=xf86-video-omapfb.git;a=summary

Yes, please do :slight_smile:

The code currently works on at least N800 and beagleboard, but I'm not
using any board-specific stuff (at least knowingly) so it should work
with almost any omapfb. Naturally if there is no overlay support,
there will not be XV support either.

In short, the currently working things are:

- Basic framebuffer drawing
- DPMS
- XVideo to certain extent (YUV2, UYVY, (up)scaling, on-screen
location...)

Things that need work:

- XV: I420 and YV12 support (needs colorspace conversion?)
- XV: resizing the video window (mostly works, but capping to 16-
divisible sizes produces drawing errors in the edges)
- Any kind of acceleration for the regular drawing operations
- Video clipping (when partially obscured or partially off-screen)
- Proper display mode setting

/me goes back to coding

- Kalle

Kalle Vahlman has been working on a X driver for omapfb[2] and there
is a package available for testing[1]. It currently has working YUV2
and UYVY modes, provided your video has a width with a multiple of 16.
Please help improving this driver :slight_smile:

Looking through the omap-driver.c, it looks like there's no 2d hw
acceleration. Has it not been written yet, have I mis-read the source
or does the hardware only have video overlays?

Koen Kooi wrote:

Hi,

Kalle Vahlman has been working on a X driver for omapfb[2] and there
is a package available for testing[1]. It currently has working YUV2
and UYVY modes, provided your video has a width with a multiple of 16.
Please help improving this driver :slight_smile:

regards,

Koen

You may want to look at two other places for OMAP video stuff.

First, DirectFB has full accelerated drivers for this chip.

Second, there has been traffic as early as yesterday on the linux-fbdev-devel mailing list regarding OMAP video drivers. This includes brand new video patches.

Keith

You may want to look at two other places for OMAP video stuff.

First, DirectFB has full accelerated drivers for this chip.

What kind of acceleration is there? To my knowledge the SGX drivers are
not public, at least not yet, and thus there can't be any acceleration. If
it just supports video overlays, well, I wouldn't call that acceleration
=).

Second, there has been traffic as early as yesterday on the
linux-fbdev-devel mailing list regarding OMAP video drivers. This
includes brand new video patches.

Third place to check out is my version of DSS rewrite at
http://www.bat.org/~tomba/git/linux-omap-dss.git/

Just yesterday I got the DSI PLL to produce pixel clock, and I've a nice
1280x1024@57 display coming from beagle board =). Although for some reason
I cannot allocate enough memory for 1280x1024x24bpp framebuffer, even if
there should be enough free continuous memory. So only 16 bit at that
resolution.

Keith

  Tomi

"the hardware" actually includes many versions with varying support,
but you are right, there is no 2D acceleration at all yet. TI is
working on an accelerated X.Org driver, and I'm hoping the work I'm
doing helps them to achieve that. There's a limited amount of
acceleration you actually *can* do for eg. N800 and I want this driver
to support those as well.

- Kalle

Koen Kooi wrote:
> Hi,

> Kalle Vahlman has been working on a X driver for omapfb[2] and there
> is a package available for testing[1]. It currently has working YUV2
> and UYVY modes, provided your video has a width with a multiple of 16.
> Please help improving this driver :slight_smile:

> regards,

> Koen

You may want to look at two other places for OMAP video stuff.

Thanks for the pointers!

First, DirectFB has full accelerated drivers for this chip.

I peeked at the directfb git and saw the driver but not a lot of code,
so I'm a bit unsure what the "full acceleration" includes... but I
know one of the developers so I know where to ask for more info :slight_smile:

Second, there has been traffic as early as yesterday on the
linux-fbdev-devel mailing list regarding OMAP video drivers. This
includes brand new video patches.

I'm actually following the linux-omap list traffic to keep up to date,
and I see those patches are discussed there. I'll be sure to check it
out.

- Kalle

Tomi Valkeinen wrote:

You may want to look at two other places for OMAP video stuff.

First, DirectFB has full accelerated drivers for this chip.

What kind of acceleration is there? To my knowledge the SGX drivers are not public, at least not yet, and thus there can't be any acceleration. If it just supports video overlays, well, I wouldn't call that acceleration =).

The best I can tell you here is to download the DirectFB sources and look at the driver. It's been a few months since looking, but I'm pretty sure that they're doing more than just overlay/alpha blending. However, at the time I was not working with OMAP, so I could be wrong with respect to how 'accelerated' the drive actually is.

Keith

http://git.directfb.org/?p=core/DirectFB.git;a=tree;f=gfxdrivers/omap;hb=HEAD
looks pretty incomplete to me...

regards,

Koen

Koen Kooi wrote:

Tomi Valkeinen <tomba@bat.org> writes:

You may want to look at two other places for OMAP video stuff.

First, DirectFB has full accelerated drivers for this chip.

What kind of acceleration is there? To my knowledge the SGX drivers are
not public, at least not yet, and thus there can't be any acceleration. If
it just supports video overlays, well, I wouldn't call that acceleration
=).

It should be possible to use the SDMA copying rectangles with optional
colour keying. You could also use the second overlay for the cursor,
though I doubt software cursor is very expensive.

If you want to get crazy, you could use the DSP to accelerate some
operations.

Second, there has been traffic as early as yesterday on the
linux-fbdev-devel mailing list regarding OMAP video drivers. This
includes brand new video patches.

Third place to check out is my version of DSS rewrite at
http://www.bat.org/~tomba/git/linux-omap-dss.git/

Just yesterday I got the DSI PLL to produce pixel clock, and I've a nice
1280x1024@57 display coming from beagle board =). Although for some reason
I cannot allocate enough memory for 1280x1024x24bpp framebuffer, even if
there should be enough free continuous memory. So only 16 bit at that
resolution.

That's great news. This should give us all the HDMI modes up to
1080p30 with correct timings.