Attendees

Agenda

  • Action items from last meeting.
  • Time to change conference dial-in?
  • Status on git.linaro.org accounts.
  • Status on access to IRC and git.linaro.org, etc. from offices.
  • Blueprint status.

    Blueprint

    Assignee

    DRI/Mali

    Chunsang/JesseBarker

    GL Proxy

    jazh

    Compositors

    jazh

    Compiz

    Amaranth

    Cairo GLES

    alf

    Benchmark

    alf

    GLEW GLES

    Shariq

  • AOB (Call for travel arrangements)?

Action Items

  • Everyone to report access status for git.linaro.org (cc Rajeev).
  • Jesse to ensure that we have valid conference call numbers for all members of the gfx WG (under way).
  • Jesse to keep track of nux licensing issues (under way).
  • Everyone to send Jesse a list of the licenses of the projects they are working on (both new and existing).

Action Items from Previous Meeting

  • Update summary below [everyone].
  • Jammy to inform Tom of additional packages needed to support kwin in our images.
  • Jesse to sort out nux licensing issues with Linaro TSC and/or Board.
    • - INPROGRESS
  • Jesse to ensure that we have valid conference call numbers for all members of the gfx WG.
    • - INPROGRESS
  • Everyone to send to Jesse a list of the licenses of the projects they are working on (both new and existing).
    • - PENDING

Minutes

  • China holiday: Jammy and Chunsang on vacation
  • nux licensing
    • A lot of people involved from Canonical and Linaro
    • Could take some time, but we don't expect this to be an issue.
  • git.linaro.org access
    • Is now included in starter tasks.
    • ACTION: Everyone to report access status to git.linaro.org (cc Rajeev).
  • DRI/Mali (Jesse)
    • Tom Cooksey (ARM) will be following the calls when possible
      • Responsible for EGL (among other things) for ARM Mali
      • GLES support for QT
    • The consensus is that we need an upstream MM solution
      • Of course not all solutions will make it upstream
      • We need to identify common needs by all vendors
      • Jesse has started a wiki page containing the requirement for such a solution, will be made public soon
    • Binary redistribution
      • No update yet
  • Cairo GLES (Alexandros)
    • Patch for fixed-function code path removal landed upstream
    • Patch for shader ftransform() removal submitted for review
      • Discussion about the best approach for this is ongoing upstream (vertex transforms on CPU vs GPU)
      • Both patches are solid, although there is some concern that transforms on CPU will not favor embedded setups
  • Benchmarks (Alexandros)
    • glmark2 11.01 in natty
    • Investigation for qtwebkit tracing
      • Seems that we can trace normally if:
        • Tile Mode is off (its off by default)
        • The page is not using HTML <canvas> (canvas is rendered always rendered to pixmap)

      • Still need to ensure the traces we get are actually drawing commands and not blits.
  • GLEW GLES (Shariq)
    • In contact with GLEW upstream
    • Jammy leveraged perl scripts from GLEW for his GL proxy work, perhaps he could help
  • Connectivity issues (Shariq)
    • IT is trying to set up access to IRC, git etc for all Linaro assignees in office using a "clean-room" approach
  • Vertex processing in CPU vs GPU discussion (Jesse)
    • How to address contention over tuning vertex processing for platforms with poor performance?
    • Performing vertex processing in the vertex shader is the right answer for OpenGL ES, but some application profile/platform combinations may show degradation.
    • Need to profile on a per-application and use case basis.
    • For embedded systems we must take power consumption into consideration.
  • Travel arrangements for LDS/UDS
    • Make arrangements as soon as possible
    • 1st March is deadline for participation
    • If there are any issues contact Jesse!

Team Work - Summary

Current Weekly Report

Complete History

Team Work - Detailed

Alexandros Frantzis (alf)

  • Cairo GLES
  • Benchmarks
    • glmark2 11.01 landed in natty universe
    • In contact with qt/qtwebkit devs about qt trace issue
      • The QT trace mechanism only tracks drawing operations to screen
      • qtwebkit when in tile mode draws to offscreen pixmap and blits to screen, so we end up tracing only blits
      • According to devs, for tile mode off (which is the default) qtwebkit renders directly to screen
      • Still need to ensure 100% that the produced traces contain normal drawing operations not just blits
        • Idea: dissect a sample qtwebkit trace to see what is actually in it

Jammy Zhou (jazh)

Chunsang Jeong

Shariq Hasnain

  • Adding OpenGL ES support to GLEW library:
    • Checking if other groups are planning on (or have already started) working on a similar task.
    • Announced my intentions to the upstream mailing list to get feedback.
    • Studying Perl language.

Travis Watkins (Amaranth)

  • compiz
    • working on using VBOs and shaders, pile of half working code so far
    • got approval from upstream to remove GLFragment (ARB fragment program), will simplify desktop/mobile differences

WorkingGroups/Middleware/Graphics/Notes/2011-02-02 (last modified 2011-02-04 11:18:17)