Differences between revisions 1 and 2
Revision 1 as of 2011-12-15 09:55:41
Size: 2825
Editor: 204
Comment:
Revision 2 as of 2013-08-30 11:49:03
Size: 2895
Editor: 213
Comment: Create Archive for Old Meeting Minutes
Deletions are marked like this. Additions are marked like this.
Line 1: Line 1:
## page was renamed from WorkingGroups/ToolChain/Meetings/2011-12-15

Thursday 15th December 2011

This month's meetings

WorkingGroups/ToolChain/Meetings
<< <  2011 / 12 >  >>
Mon Tue Wed Thu Fri Sat Sun
      1 2 3 4
5 6 7 8 9 10 11
12 13 14 15 16 17 18
19 20 21 22 23 24 25
26 27 28 29 30 31  

Attendees

Agenda

  • Stand up call

Minutes

Dave:

  • Moving the code buffer to a fixed address is upstream
  • Looking at context routines, hopefully today
  • Note that libgo is switching to use this as well
  • Do have to understand all before implementing
  • Hopefully some tests in libc, have found some others

Ramana:

  • Trying addressing modes for vld1, running into issues in the backend
  • Looked at ICE when building LLVM, needs more debugging
  • Bug doesn't seem to show on trunk, reload behaviour is different

Ken:

  • Disabled NLS, got SATO image to build!
  • Also built the Qt demo
  • Both boot in QEMU seem to work just fine - ~250 packages in SATO, ~100 in Qt
  • QEMU/OpenOCD configuration is ARMv5
  • Need ARMv7 Thumb-2 config somehow
  • Has approval to contribute to OpenEmbedded

Revital:

  • Resubmitting the patch to estimate register pressure
  • Testing, will send
  • Then reopen the merge request

Ulrich:

  • Continuing discussion on remote core file
  • Response has become be even more generalised...
  • ...to read any file, instead of just /proc

Ira:

  • Figured out how to control MEM_ALIGN from the vectoriser
  • 16 % improvement on one benchmark...

Richard:

  • Finishing off sched pressure
  • Running SPEC and similar this week
  • Had a look at bugzilla on poor code for intrinsics that load one lane, work on lane, store out
  • Much of the vecotriser was uninitialised and wasn't generating good code
  • Prefer using a load and duplicate instead of tracking values in GCC

Asa:

  • Finished cross compiler benchmark report
  • Having a look at EEMBC and regressions between gcc-linaro 4.4 and 4.6
  • and one with 4.4 loop unrolling turned off
  • Will continue analysing the numbers

Michael:

  • Finishing off binary toolchains
  • OpenOCD, Renderscript cards
  • Drop LLVM
  • Making blueprints to go against the existing cards

WorkingGroups/ToolChain/Meetings/Archive/2011-12-15 (last modified 2013-08-30 11:49:03)