Thursday 19th May 2011

This month's meetings

WorkingGroups/ToolChain/Meetings
<< <  2011 / 5 >  >>
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

  • Standup call

Minutes

Revtial:

  • Sent patches for SMS upstream as follow up to the Linaro GCC merged one
  • Patches improve code size
  • Looking into oppertunities in SPEC and others
  • Committed the approved patches to the branches

Ken:

  • Has installed the Linaro Android image on his PandaBoard

  • Installed SDK, NDK, etc to try some of the examples
  • To look into the trace utility that Jim mentioned
  • Trying to measure -funwind-table overhead. Want to measure C++ and C applications
  • Perhaps try Python and Firefox? xdeb?

Chung-Lin:

  • Just returned!
  • NEON DImode constant patch - two different versions, one in CSL, one in trunk
  • Should we take all of the tree of patches?

Peter:

  • Closing off previous cycle's blueprints, planning for this cycle
  • Remaining is floating point exception flags. Easy are fixed, rest are hard
  • Will send out patch set for the easy and postpone the less important/hard ones
  • Fleshing out the QEMU blueprint page

Ira:

  • Committed reduction support in SOP upstream
  • Continued looking into different benchmarks such as FFMEG, EEMBC, pixman
  • Looking into implementation of reverse access
  • ...and potential engineering blueprints

Ulrich:

  • Various reviews as he's on the review roster
  • ...and Yao's patchset for GDB displaced stepping
  • Still has issues so unfortunately not in this release. Last issue remaining...
  • Ran the release process. Has a clean testsuite except for displaced stepping

Dave:

  • Sent latrace some patches for ARM
  • Looked at ltrace as it was hanging on a particular ptrace call
  • Added the stub for ARM and it seems to work for a single process
  • Optimising unaligned memcpy(). Some nasty cases to investigate.
  • Such as if source misaligned by 12, dest by 8: hurts
  • Running the SPEC latrace through analysis tools
  • Some such as tonto have many libm calls that were hurting the trace
  • Checked that the binutils errata fix fixes apt/GTK error

Ramana:

  • Looking into the vmov.i64 truncation problem
  • Looks like Natty is clear as it's not built in the 'has 64 bit extra target' option
  • Looked at the NEON vs non-NEON GCC testsuite results
  • With NEON on there are some extra ICEs - will look into them
  • Michael: doesn't think they're regressions so OK for now

Michael:

  • Stubbed out blueprints and delegated
  • Set up to use Ubuntu as a testsuite
  • Ran the release process
  • Shifted the meeting
  • Re-ran CoreMark

Andrew:

  • Trying to sort out 16x16->64 bit multiply patch

  • Complicated as there's a bootstrap failure
  • Bernd re: shrink-wrap: no change from Richard's patch. Not working there at the moment.

WorkingGroups/ToolChain/Meetings/Archive/2011-05-19 (last modified 2013-08-30 11:48:27)