Platform/GFX/2008-July-22: Difference between revisions

From MozillaWiki
< Platform‎ | GFX
Jump to navigation Jump to search
 
(3 intermediate revisions by one other user not shown)
Line 24: Line 24:
** SSE2 blur code is a little faster (42135ms vs 45007ms), but no rush to clean it up for merge
** SSE2 blur code is a little faster (42135ms vs 45007ms), but no rush to clean it up for merge
** building a little demo to see what type of graphic performance the n810 is capable of
** building a little demo to see what type of graphic performance the n810 is capable of
** look into running existing graphics benchmarks on n810


* Bobby
* Bobby
-Precaching patch in review process with joe. Takes us from 20% to 10% regression on talos tp
**Precaching patch in review process with joe. Takes us from 20% to 10% regression on talos tp
-Wrote sse2 implementation of matrixmul. Currently debugging. Perf testing the broken version shows about a 30% reduction in overall perf hit of CM (artificial numbers go from 20%->13.5%)
**Wrote sse2 implementation of matrixmul. Currently debugging. Perf testing the broken version shows about a 30% reduction in overall perf hit of CM (artificial numbers go from 20%->13.5%)
**intent patch reviewed and ready for checkin
**glenn just posted something about how libpng might leak with cm using the current version. Mentioned something about apng? He put it as a blocker for CM.
**Chris Murphy has some concerns that a lot of mac profiles might be from bogus data. Not sure how much we care.


* Rob
* Rob

Latest revision as of 21:10, 22 July 2008

index

  • Conference ID 217

Status Updates

  • Vlad
    • border fun: ready to land
    • cairo upgrade: ready to land
    • looking at DirectFB merging this week
  • Joe
    • image caching patch (review waiting on Stuart)
    • removing photon/beos/etc. in preparation for removing nsIImage
  • Jeff
    • landed fix to device offset bug
    • landed fbOver speedup patch
    • measured performance of pixman image scaling vs. old imgScaler.cpp.
      • pixman is about 4.5x slower.
      • have hacky patch to integrate imgScaler.cpp into pixman (will need lots of cleanup)
    • SSE2 blur code is a little faster (42135ms vs 45007ms), but no rush to clean it up for merge
    • building a little demo to see what type of graphic performance the n810 is capable of
    • look into running existing graphics benchmarks on n810
  • Bobby
    • Precaching patch in review process with joe. Takes us from 20% to 10% regression on talos tp
    • Wrote sse2 implementation of matrixmul. Currently debugging. Perf testing the broken version shows about a 30% reduction in overall perf hit of CM (artificial numbers go from 20%->13.5%)
    • intent patch reviewed and ready for checkin
    • glenn just posted something about how libpng might leak with cm using the current version. Mentioned something about apng? He put it as a blocker for CM.
    • Chris Murphy has some concerns that a lot of mac profiles might be from bogus data. Not sure how much we care.
  • Rob
    • Double buffer patch for vista caused flickering, now needs compositor for a proper fix :(
    • Super-secret intern project has been reclaimed by Vlad
    • Looked into reverse engineering WPF glyph rasterization API
    • Aero glass
      • I unbitrotted the patch for the layout bug
      • Fallback code is better handled in CSS. But that requires that I be...
      • Working on a new bug to expose metric changes (DWM/default theme in particular) to CSS
  • Eric
    • Landed Canvas text bidi
    • Canvas Cairo->Thebes conversion nearly ready to land
    • Updating Canvas mochitests

Triage

Other

Retrieved from "https://wiki.allizom.org/index.php?title=Platform/GFX/2008-July-22&oldid=101598"