TPEPlatform/2014-01-09: Difference between revisions

From MozillaWiki
Jump to navigation Jump to search
 
(28 intermediate revisions by 15 users not shown)
Line 1: Line 1:
==Media==
==Media==
'''WiFi Display'''
* {{bug|952359}} - capture video for WFD / vichen
**study render and composition process, start to test by using WebRTC
* {{bug|952361}} - capture audio for WFD / vichen
**not start
'''Media Recording'''
*{{Bug|879668}} - [MediaEncoder] Implement OmxTrackEncoder on B2G / John Lin
** update patches to address review comments
** r+'ed from roc
** push to try server and verifying
** investigate audio playback problem of output MP4 files on Android/FxOS.
*{{Bug|957939}} - Use ContainerWriter's HAS_AUDIO/HAS_VIDEO flags / ayang
** update patche to review
* Check fragment mp4 playback on Android 4.3 and B2G / ayang
** Audio rejects by omx aac encoder, it needs to check further.
** B2G DecodeStateMachine asserts at getDuration().
*{{Bug|950429}} Media Recording - allow NS_DispatchToCurrentThread in Media Encoder thread  /rlin
** landed
*{{Bug|891705}} [MediaEncoder] Implement WebM 1.0 container writer /rlin
** keep reviewing amd fix nits
*{{Bug|957452}} MediaRecorder: use-after-free crash [@mozilla::dom::MediaRecorder::Session::GetEncodedData]
** try to build asan to debug this.
*{{Bug|955981}} - MediaEncoder - Make InterleaveTrackData function can be shared /Shelly
**r+, landed
*{{Bug|936981}} - [MediaEncoder::Gtest] Implement unit test case of OpusTrackEncoder /Shelly
**r+, landed
*{{Bug|957439}} - MediaRecorder: "Assertion failure: NS_IsMainThread()  && mTrackUnionStream" [@mozilla::dom::MediaRecorder::Session::Pause] /Shelly
'''Web Audio'''
*{{Bug|916135}} - Investigate webaudio mochitest timeouts on B2G / JW
** fix patches as suggested by Paul
** try server fail - check failures on other platforms
'''openh264'''
** trace webrtc code / JW
** discuss with China Cisco team about migrating existing tests to gtests
** implement encoder test
** land decoder/encoder to Cisco repo
** refactor code to pipeline decoder output to encoder test in order to remove large yuv files


==Streaming Media==
==Streaming Media==
Line 5: Line 49:
** Modify bbrittain's patch and we can profile the video codec performance on devices.
** Modify bbrittain's patch and we can profile the video codec performance on devices.
** The performance test on unagi, https://bugzilla.mozilla.org/show_bug.cgi?id=890419#c25.
** The performance test on unagi, https://bugzilla.mozilla.org/show_bug.cgi?id=890419#c25.
'''GetUserMedia'''
* {{Bug|947010}} - Create automation test for recording status / SC
** update patch and investigating the root cause of failing on mochitest-b2g-desktop (see bug 956625 and bug 957554)
** r+'ed and ready to land on m-c, waiting for the final call from @fabrice.


* Check for pointerlock permission prompt. / ayang
** It probably needs to add a testcase.
==Graphics ==
==Graphics ==
*Parallel EcmaScript: Trace Intel RiverTrail implementation. /CTai
* Parallel EcmaScript: Trace Intel RiverTrail implementation. /CTai
*GTC 2014 session list check. /CTai
* GTC 2014 session list check. /CTai
*JPEG encoder/decoder: /CTai
* JPEG encoder/decoder: /CTai
**Study JPEG encoder/decoder theory.  
* Study JPEG encoder/decoder theory. /CTai
**Briefly understand what is NEON.
* Briefly understand what is NEON. /CTai
 
* {{Bug|956265}} - [Flatfish][Settings] Sluggish UI when "Enable hardware composer" is set (and is set by default) /vilin
** Hard-coded mRBSwapSupport flag makes some composition unexpectedly go into HWComposer HAL.(which isn't verified yet.)
** HWComposer shouldn't be enabled on FlatFish. Bug 956193 has already disable it temporally.
 
* {{Bug|951846}} - [FlatFish] ClientThebesLayer::PaintThebes() costs a lot on RotatedBuffer::DrawBufferWithRotation /vilin
** Performance issue on copying GraphicBuffer.
** Test on recommendation from IMG, but still not improved.
*** Only set READ_OFTEN flag to src buffer and WRITE_OFTEN flag to dst buffer.
 
* {{Bug|955987}} - [Text Selection] Change selection status in FrameScript /vilin
** Splitted from Bug 924692.
** Study how Input events are sent from InputDispatch to Selection and how current gesture implementation was.
 
*Profiler and LayerScope tool /Jerry
**Design wireframe
 
* {{Bug|945082}} - lockscreen fade-out animation is still < 60fps /Jerry
** Compositor costs a lot for opacity anim.
** With both opacity and transform anim, b2g costs almost 2x time for app's first frame.
 
* {{Bug|945082}} - [Display][gonk-kk] Porting GonkDisplay, nativewindow and libui / solomon
** Refine the patch according Micheal and Peter's comment
** Check and remove all pixel format handling of PIXEL_FORMAT_A_8 in GrallocTexture and TextureHostOGL
** GonkNativeWindow modification according to KK's change
 
* {{Bug|893304}} - Convert CanvasClientWebGL (and Host) to the new textures /Morris
** Rebase patch to latest m-c
** r+ but try push failed
 
* {{Bug|926128}} - reduce layer programs in use to 6, add colormatrix and blur support /Morris
** Found that some uniform not bind correctly.
** Still encounter some try failed
 
* {{Bug|951531}} - framebuffer to use RGB565 16bit /Morris
** Help to check if we actually use 16bit FB
 
* {{Bug|946596}} - Hi-res phone does not take hi-res screenshots /peter
** wait for reviewer feedback
 
* {{Bug|951241}} - Gallery edit mode artifact /peter
** Debugging
 
* {{Bug|956432}} - cut the rope display abnormal /peter
** happened when set canvas backend as skia, debugging drawimage call
 
* Parallel EcmaScript: Trace Intel RiverTrail implementation. /CTai
* GTC 2014 session list check. /CTai


==Input==
==Input==
* {{Bug|924692}} - Add support for selecting using touch / Phoebe
**add textarea scrolling boundery check, will attach touch caret at the right position on textarea boundary.
**fixing cross document touch caret position issue, refactor and working for another WIP.


==Stability==
==Stability==
* {{Bug|950485}} - permission broken / vichen
** study permission manager in gecko
** need to roll back bug 816956


==Free Agent==
==Platofrm==
* {{Bug|956625}} - [b2g-desktop][mochitest] ReferenceError: Cc is not defined while using SpecialPowers.loadChromeScript / SC
* {{Bug|956625}} - [b2g-desktop][mochitest] ReferenceError: Cc is not defined while using SpecialPowers.loadChromeScript / SC
** simply add |var Cc = Components.classes;| while Cc is undefined.
** simply add |var Cc = Components.classes;| while Cc is undefined.
** r+'ed
** r+'ed and landed on m-c
* {{Bug|957554}} - [b2g-desktop][mochitest] shell.js doesn't receive mozbrowserloadstart event / SC
** unable to test mozChromeEvent because all the event are pending by shell.js before receiving mozbrowserloadstart
** root cause is that mochitest-b2g-desktop turn off the pref "dom.mozBrowserFramesEnabled" due to not supporting OOP.
** ask @fabrice if anything we can fixed
* {{Bug|854795}} - decode and downsample images at the same time for memory savings / SC
** discussed architecture and implementation plan with @thinker, @kanru
** preparing the WIP for ImageDownsampler
* enable miniperf for unagi and peak / slee
* enable miniperf for unagi and peak / slee
** Have a simple introduction.
** Have a simple introduction.
*{{Bug|908995}} - [B2G] Task tracer / Shelly
*{{Bug|916409}} - Data retriever for task tracer /Shelly
** Experiment on different ways to log logfiles and start up task tracer.
*** 1. Hook up with profile.sh -> Not very feasible and too many steps to start up.
*** 2. Use PR_LOG -> Can't just open once file for multiple processes.
*** 3. Use adb logcat, and write the result to a file -> Might increase the overhead
*** 4. Use PR_LOG, but open log file per process -> Somehow fail to open the log file with "Homescreen" process.
** Coordinate with Gina and alive about the log format.


==Study==
==Study==
* Keyboard api study / slee
* Keyboard api study / slee
* IPDL,IPC study / alin
** realize how top level protocols were constructed when launching a content process.
** how many types of message loops connected to to specified threads
** what is message channel? it's related to IO thread.
* JPEG encoder/decoder. /CTai
* Study JPEG encoder/decoder theory. /CTai
*Understand what is NEON /CTai
* study Android RenderScript / JW

Latest revision as of 09:31, 10 January 2014

Media

WiFi Display

  • bug 952359 - capture video for WFD / vichen
    • study render and composition process, start to test by using WebRTC
  • bug 952361 - capture audio for WFD / vichen
    • not start

Media Recording

  • bug 879668 - [MediaEncoder] Implement OmxTrackEncoder on B2G / John Lin
    • update patches to address review comments
    • r+'ed from roc
    • push to try server and verifying
    • investigate audio playback problem of output MP4 files on Android/FxOS.
  • bug 957939 - Use ContainerWriter's HAS_AUDIO/HAS_VIDEO flags / ayang
    • update patche to review
  • Check fragment mp4 playback on Android 4.3 and B2G / ayang
    • Audio rejects by omx aac encoder, it needs to check further.
    • B2G DecodeStateMachine asserts at getDuration().
  • bug 950429 Media Recording - allow NS_DispatchToCurrentThread in Media Encoder thread /rlin
    • landed
  • bug 891705 [MediaEncoder] Implement WebM 1.0 container writer /rlin
    • keep reviewing amd fix nits
  • bug 957452 MediaRecorder: use-after-free crash [@mozilla::dom::MediaRecorder::Session::GetEncodedData]
    • try to build asan to debug this.
  • bug 955981 - MediaEncoder - Make InterleaveTrackData function can be shared /Shelly
    • r+, landed
  • bug 936981 - [MediaEncoder::Gtest] Implement unit test case of OpusTrackEncoder /Shelly
    • r+, landed
  • bug 957439 - MediaRecorder: "Assertion failure: NS_IsMainThread() && mTrackUnionStream" [@mozilla::dom::MediaRecorder::Session::Pause] /Shelly

Web Audio

  • bug 916135 - Investigate webaudio mochitest timeouts on B2G / JW
    • fix patches as suggested by Paul
    • try server fail - check failures on other platforms

openh264

    • trace webrtc code / JW
    • discuss with China Cisco team about migrating existing tests to gtests
    • implement encoder test
    • land decoder/encoder to Cisco repo
    • refactor code to pipeline decoder output to encoder test in order to remove large yuv files

Streaming Media

GetUserMedia

  • bug 947010 - Create automation test for recording status / SC
    • update patch and investigating the root cause of failing on mochitest-b2g-desktop (see bug 956625 and bug 957554)
    • r+'ed and ready to land on m-c, waiting for the final call from @fabrice.
  • Check for pointerlock permission prompt. / ayang
    • It probably needs to add a testcase.

Graphics

  • Parallel EcmaScript: Trace Intel RiverTrail implementation. /CTai
  • GTC 2014 session list check. /CTai
  • JPEG encoder/decoder: /CTai
  • Study JPEG encoder/decoder theory. /CTai
  • Briefly understand what is NEON. /CTai
  • bug 956265 - [Flatfish][Settings] Sluggish UI when "Enable hardware composer" is set (and is set by default) /vilin
    • Hard-coded mRBSwapSupport flag makes some composition unexpectedly go into HWComposer HAL.(which isn't verified yet.)
    • HWComposer shouldn't be enabled on FlatFish. Bug 956193 has already disable it temporally.
  • bug 951846 - [FlatFish] ClientThebesLayer::PaintThebes() costs a lot on RotatedBuffer::DrawBufferWithRotation /vilin
    • Performance issue on copying GraphicBuffer.
    • Test on recommendation from IMG, but still not improved.
      • Only set READ_OFTEN flag to src buffer and WRITE_OFTEN flag to dst buffer.
  • bug 955987 - [Text Selection] Change selection status in FrameScript /vilin
    • Splitted from Bug 924692.
    • Study how Input events are sent from InputDispatch to Selection and how current gesture implementation was.
  • Profiler and LayerScope tool /Jerry
    • Design wireframe
  • bug 945082 - lockscreen fade-out animation is still < 60fps /Jerry
    • Compositor costs a lot for opacity anim.
    • With both opacity and transform anim, b2g costs almost 2x time for app's first frame.
  • bug 945082 - [Display][gonk-kk] Porting GonkDisplay, nativewindow and libui / solomon
    • Refine the patch according Micheal and Peter's comment
    • Check and remove all pixel format handling of PIXEL_FORMAT_A_8 in GrallocTexture and TextureHostOGL
    • GonkNativeWindow modification according to KK's change
  • bug 893304 - Convert CanvasClientWebGL (and Host) to the new textures /Morris
    • Rebase patch to latest m-c
    • r+ but try push failed
  • bug 926128 - reduce layer programs in use to 6, add colormatrix and blur support /Morris
    • Found that some uniform not bind correctly.
    • Still encounter some try failed
  • bug 951531 - framebuffer to use RGB565 16bit /Morris
    • Help to check if we actually use 16bit FB
  • bug 946596 - Hi-res phone does not take hi-res screenshots /peter
    • wait for reviewer feedback
  • bug 951241 - Gallery edit mode artifact /peter
    • Debugging
  • bug 956432 - cut the rope display abnormal /peter
    • happened when set canvas backend as skia, debugging drawimage call
  • Parallel EcmaScript: Trace Intel RiverTrail implementation. /CTai
  • GTC 2014 session list check. /CTai

Input

  • bug 924692 - Add support for selecting using touch / Phoebe
    • add textarea scrolling boundery check, will attach touch caret at the right position on textarea boundary.
    • fixing cross document touch caret position issue, refactor and working for another WIP.

Stability

  • bug 950485 - permission broken / vichen
    • study permission manager in gecko
    • need to roll back bug 816956

Platofrm

  • bug 956625 - [b2g-desktop][mochitest] ReferenceError: Cc is not defined while using SpecialPowers.loadChromeScript / SC
    • simply add |var Cc = Components.classes;| while Cc is undefined.
    • r+'ed and landed on m-c
  • bug 957554 - [b2g-desktop][mochitest] shell.js doesn't receive mozbrowserloadstart event / SC
    • unable to test mozChromeEvent because all the event are pending by shell.js before receiving mozbrowserloadstart
    • root cause is that mochitest-b2g-desktop turn off the pref "dom.mozBrowserFramesEnabled" due to not supporting OOP.
    • ask @fabrice if anything we can fixed
  • bug 854795 - decode and downsample images at the same time for memory savings / SC
    • discussed architecture and implementation plan with @thinker, @kanru
    • preparing the WIP for ImageDownsampler
  • enable miniperf for unagi and peak / slee
    • Have a simple introduction.
  • bug 908995 - [B2G] Task tracer / Shelly
  • bug 916409 - Data retriever for task tracer /Shelly
    • Experiment on different ways to log logfiles and start up task tracer.
      • 1. Hook up with profile.sh -> Not very feasible and too many steps to start up.
      • 2. Use PR_LOG -> Can't just open once file for multiple processes.
      • 3. Use adb logcat, and write the result to a file -> Might increase the overhead
      • 4. Use PR_LOG, but open log file per process -> Somehow fail to open the log file with "Homescreen" process.
    • Coordinate with Gina and alive about the log format.

Study

  • Keyboard api study / slee
  • IPDL,IPC study / alin
    • realize how top level protocols were constructed when launching a content process.
    • how many types of message loops connected to to specified threads
    • what is message channel? it's related to IO thread.
  • JPEG encoder/decoder. /CTai
  • Study JPEG encoder/decoder theory. /CTai
  • Understand what is NEON /CTai
  • study Android RenderScript / JW
Retrieved from "https://wiki.mozilla.org/index.php?title=TPEPlatform/2014-01-09&oldid=878136"