[Nouveau] Self introduction Hans de Goede

[Ilia Mirkin]

On Tue, May 26, 2015 at 3:43 AM, Hans de Goede <hdegoede at redhat.com> wrote:
> Hi,
>
>
> On 26-05-15 09:41, Martin Peres wrote:
>>
>> On 26/05/15 10:29, Hans de Goede wrote:
>>>
>>> Hi All,
>>>
>>> Since I will be working on nouveau pretty much starting today I thought
>>> it would be good to write a quick self introduction.
>>>
>>> I'm an FOSS enthusiast / developer since 1996. I've written (and still
>>> maintain) various hwmon drivers, various USB webcam drivers, libv4l and
>>> libgphoto camlibs for photoframes which use a custom usb protocol and
>>> usb redirection over the network for qemu.
>>>
>>> And although not the original author I'm now a days a maintainer of
>>> libusb, usb-3 bulk streams support in the xhci driver, the uas driver,
>>> and libahci_platform.
>>>
>>> In my spare time I work on u-boot and Linux support for Allwinner ARM
>>> SoCs, I'm a u-boot maintainer for these, and maintain the sata, mmc
>>> and usb kernel drivers.
>>>
>>> About a year ago I joined the Red Hat graphics team, where my first
>>> task was to help to get libinput up to a level where it could be
>>> used as the input stack for Wayland. The work on this is winding
>>> down, so my next project is to help out with nouveau and here I am.
>>>
>>> My knowledge of GPU-s and basically anything 3d is quite limited
>>> atm, so I will likely be asking a lot of questions to get up to
>>> speed. I'll also join the #nouveau channel on irc, so I'll see
>>> you all there.
>>>
>>> Regards,
>>>
>>> Hans
>>
>>
>> Hey Hans.
>>
>> That's very good news! Welcome to the project!
>>
>> Do you know if you are going to help mostly on the kernel or mesa side?
>
>
> Where-ever help is needed is the best way I can describe the plan
> I guess. Step 1 is getting up to speed, I'll likely focus on Mesa
> first for that.

On the mesa end of things, I was able to get up to speed (starting at
roughly the same place you are in terms of graphics knowledge and
NVIDIA hw familiarity) by fixing piglit tests. This forces you to (a)
understand what the test is testing and teaches you bits of GL and (b)
figure out what the driver does to implement the feature. There's a
trello where I've kept some of the various TODO items:
https://trello.com/b/ZudRDiTL/nouveau but there are more failures than
that (and note that some of the items are "catalog the failures"). I
haven't done a full piglit run on nv4x in quite a while, and even back
when I did, I didn't really go into most of the failures. However on
nv50/nvc0, I have a pretty good idea of why each piglit fails.

Most of this isn't "feature" work though... just fixing/improving
things. If you're looking for feature work, buffers and images are the
next big items to tackle (and will require some st/mesa plumbing
although there's a GSoC student working on it for buffers in the
context of atomics).

Cheers,

  -ilia