Re: [PATCH 1/2] drm/exynos: g2d: Add support for old S5Pv210 type

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Hey Krzysztof,


Krzysztof Kozlowski wrote:
> On 05/24/2016 06:05 PM, Tobias Jakobi wrote:
>> Hello Krzysztof,
>>
>>
>> Krzysztof Kozlowski wrote:
>>> On 05/24/2016 03:49 PM, Tobias Jakobi wrote:
>>>> Hello Krzysztof,
>>>>
>>>> are you sure that these are the only differences. Because AFAIK there
>>>> are quite a few more:
>>>> - DMA submission of commands
>>>> - blend mode / rounding
>>>> - solid fill
>>>> - YCrCb support
>>>> - and probably more
>>>>
>>>> One would need to add least split the command list parser into a v3 and
>>>> v41 version to accomodate for the differences. In fact userspace/libdrm
>>>> would need to know which hw type it currently uses, but you currently
>>>> always return 4.1 in the corresponding ioctl.
>>>
>>> Eh, so probably my patch does not cover fully the support for v3 G2D. I
>>> looked mostly at the differences between v3 and v4 in the s5p-g2d driver
>>> itself. However you are right that this might be not sufficient because
>>> exynos-g2d moved forward and is different than s5p-g2d.
>>>
>>>> Krzysztof Kozlowski wrote:
>>>>> The non-DRM s5p-g2d driver supports two versions of G2D: v3.0 on
>>>>> S5Pv210 and v4.x on Exynos 4x12 (or newer). The driver for 3.0 device
>>>>> version is doing two things differently:
>>>>> 1. Before starting the render process, it invalidates caches (pattern,
>>>>>    source buffer and mask buffer). Cache control is not present on v4.x
>>>>>    device.
>>>>> 2. Scalling is done through StretchEn command (in BITBLT_COMMAND_REG
>>>>>    register) instead of SRC_SCALE_CTRL_REG as in v4.x. However the
>>>>>    exynos_drm_g2d driver does not implement the scalling so this
>>>>>    difference can be eliminated.
>>>> Huh? Where did you get this from? Scaling works with the DRM driver.
>>>
>>> I was looking for the usage of scaling reg (as there is no scaling
>>> command). How the scaling is implemented then?
>> Like you said above the drivers work completly different. The DRM one
>> receives a command list that is constructed by userspace (libdrm
>> mostly), copies it to a contiguous buffer and passes the memory address
>> of that buffer to the engine which then works on it. Of course
>> everything is slightly more complex.
>>
>> You don't see any reference to scaling in the driver because the scaling
>> regs don't need any kind of specific validation.
>>
>> If you want to know how the command list is constructed, the best way is
>> to look into libdrm. The Exynos specific tests actually cover scaling.
> 
> Thanks for explanations. The patch is insufficient then and it requires
> much more effort. Please drop the series as of now.
If you intend to add support for v3 hardware we might want to join
forces. I have an (ongoing) rewrite of the driver to enable additional
functionality and to make it more suitable to provide EXA for a X11 DDX.
However this breaks the existing API and rewrites quite a bit of the driver

Here's what's there so far.
https://github.com/tobiasjakobi/linux-odroid-public/compare/e35ca9aca1214c5e104e6906c1d9affeb80fe5df...3d1ddb86db73b0d664f3e339709e8e8dacdc8e91

And here's the DDX:
https://github.com/tobiasjakobi/xf86-video-armsoc/commits/g2d


With best wishes,
Tobias



> 
> Best regards,
> Krzysztof
> 

--
To unsubscribe from this list: send the line "unsubscribe linux-media" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html



[Index of Archives]     [Linux Input]     [Video for Linux]     [Gstreamer Embedded]     [Mplayer Users]     [Linux USB Devel]     [Linux Audio Users]     [Linux Kernel]     [Linux SCSI]     [Yosemite Backpacking]
  Powered by Linux