logo

drewdevault.com

[mirror] blog and personal website of Drew DeVault git clone https://hacktivis.me/git/mirror/drewdevault.com.git

Sway-wlroots-at-XDC-2018.md (4691B)


  1. ---
  2. date: 2018-09-30
  3. title: Sway & wlroots at XDC 2018
  4. layout: post
  5. tags: ["sway", "wlroots", "wayland", "roundup"]
  6. ---
  7. Just got my first full night of sleep after the return flight from Spain after
  8. attending [XDC 2018](https://xdc2018.x.org/). It was a lot of fun! I attended
  9. along with four other major wlroots contributors. Joining me were [Simon Ser
  10. (emersion)](https://github.com/emersion) (a volunteer) and [Scott Anderson
  11. (ascent12)](https://github.com/ascent12) of
  12. [Collabora](https://www.collabora.com/), who work on both
  13. [wlroots](https://github.com/swaywm/wlroots) and
  14. [sway](https://github.com/swaywm/sway). [ongy](https://github.com/ongy) works on
  15. wlroots, [hsroots](https://github.com/swaywm/hsroots), and
  16. [waymonad](https://github.com/waymonad/waymonad), and joined us on behalf
  17. of [IGEL](https://www.igel.com/). Finally, we were joined by [Guido Günther
  18. (agx)](https://github.com/agx) of [Purism](https://puri.sm/), who works with us
  19. on wlroots and on the Librem 5. This was my first time meeting most of them
  20. face-to-face!
  21. wlroots was among the highest-level software represented at XDC. Most of the
  22. attendees are hacking on the kernel or mesa drivers, and we had a lot to learn
  23. from each other. The most directly applicable talk was probably VKMS (virtual
  24. kernel mode setting), a work-in-process kernel subsystem which will be useful
  25. for testing the complex wlroots DRM code. We had many chances to catch up with
  26. the presenters after their talk to learn more and establish a good
  27. relationship. We discovered from these chats that some parts of our DRM code
  28. was buggy, and have even started onboarding some of them as contributors to sway
  29. and wlroots.
  30. We also learned a lot from the other talks, in ways that will pay off over time.
  31. One of my favorites was an introduction to the design of Intel GPUs, which went
  32. into a great amount of detail into how the machine code for these GPUs worked,
  33. why these design decisions make them efficient, and their limitations and
  34. inherent challenges. Combined with other talks, we got a lot of insight into the
  35. design and function of mesa, graphics drivers, and GPUs. These folks were very
  36. available to us for further discussion and clarification after their talks, a
  37. recurring theme at XDC and one of the best parts of the conference.
  38. Another recurring theme at XDC was talks about how mesa is tested, with the most
  39. in-depth coverage being on Intel's new CI platform. They provide access to Mesa
  40. developers to test their code on *every* generation of Intel GPU in the course
  41. of about 30 minutes, and brought some concrete data to the table to show that it
  42. really works to make their drivers more stable. I took notes that you can expect
  43. to turn into builds.sr.ht features! And since these folks were often available
  44. for chats afterwards, I think they were taking notes, too.
  45. I also met many of the driver developers from AMD, Intel, and Nvidia; all of
  46. whom had interesting insights and were a pleasure to hang out with. In fact,
  47. Nvidia's representatives were the first people I met! On the night of the
  48. kick-off party, I led the wlroots clan to the bar for beers and introduced
  49. myself to the people who were standing there - who already knew me from my
  50. writings critical of Nvidia. Awkward! A productive conversation ensued
  51. regardless, where I was sad to conclude that we still aren't going to see any
  52. meaningful involvement in open source from Nvidia. Many of their engineers are
  53. open to it, but I think that the engineering culture at Nvidia is unhealthy and
  54. that the engineers have very little influence. We made our case and brought up
  55. points they weren't thinking about, and I can only hope they'll take them home
  56. and work on gradually improving the culture.
  57. Unfortunately, Wayland itself was somewhat poorly represented. Daniel Stone (a
  58. Wayland & Weston maintainer) was there, and Roman Glig (of KDE), but some KDE
  59. folks had to cancel and many people I had hoped to meet were not present. Some
  60. of the discussions I wanted to have about protocol standardization and
  61. cooperation throughout Wayland didn't happen. Regardless, the outcome of XDC was
  62. very positive - we learned a lot and taught a lot. We found new contributors to
  63. our projects, and have been made into new contributors for everyone else's
  64. projects.
  65. Big shoutout to the X Foundation for organizing the event, and to the beautiful
  66. city of A Coruña for hosting us, and to University of A Coruña for sharing their
  67. university - which consequently led to meeting some students there that used
  68. Sway and wanted to contribute! Thanks as well to the generous sponsors, both for
  69. sponsoring the event and for sending representatives to give talks and meet the
  70. community.