Monday 19th 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

  • Review action items from last meeting
  • Christmas break
  • Bugs
    • 736661: ICE with std::pair<bool,bool> (cdce3.C)

    • 903178: ICE no spill on LO_REGS for TextDiagnosticPrinter.ii (Ramana)

    • 836588: FTBFS due to unresolved label with -fPIC (Ramana)

    • 897583: in FSF 4.6, will get picked up with merge

  • Handover

Action Items from this Meeting

  • TODO: Michael and Richard to send an email to the other ARM branches proposing something re: backports
  • ACTION: Michael to blueprint pipeline bypass, get someone to complete

Action Items from Previous Meeting

  • INPROGRESS: Andrew to try Julian again on 675347

  • ACTION: Andrew to update release notes to match re: key signing
  • DONE: Michael to patch to disable biarch on ubutest
  • DONE: Michael to spin a template blueprint on vld1 alignement

Minutes

Dave, Ken, Asa, Zhenqiang, Revital, Ulrich, Ira, Richard, Richard, Ramana, Matthias

736661: ICE with std::pair<bool,bool>:

  • Was on hold due to the 4.5 failure
  • Just commit, put the cdc failure to the side?
  • Do via the FSF release branch
  • Matthias's deadline is the next Ubuntu release, ~2 months

903178: ICE no spill on LO_REGS for TextDiagnosticPrinter.ii (Ramana)

  • Was looking at last week, need to debug through reload
  • Seen in LLVM 2.9, not trunk, so not as urgent
  • Can call on Ulrich if needed
  • Will follow up offline

836588: FTBFS due to unresolved label with -fPIC (Ramana)

  • More important as upstream and hits a few packages

SMS and the missing MLA forwarding pipeline description:

  • Saw this, Ramana was looking at changing the description
  • ACTION: Michael to blueprint, get someone to complete

Richard:

  • Here apart from the bank holidays
  • sched-pressure
    • Hasn't sent upstream yet
    • Unexpected behaviour in the EEMBC FFT area
    • Will send upstream this week
    • Might not get a response in a reasonable time
    • Note: will be available as an alternative and turned on by default for ARM
    • Happy enough to due the merge during stage 1
    • We'll need to do benchmarking etc
    • Do merge request before or after review is fine
  • SMS memory dependencies
    • Case of handing a patch over to someone
  • Auto inc-dec
    • Has been posted upstream after feedback

General style:

  • Blueprint describing the details of the work
  • Merge request against gcc-linaro-4.6.
  • Can name tests, don't put actual numbers in public

Revital:

  • SMS register pressure
  • 879725: ICE in int_mode_for_mode, at stor-layout.c:490

Dave:

  • Another two upstream in the next few days into GLIBC
  • Might get another few into Newlib
  • GLIBC now has the infrastructure
  • Newlib has the first as an example
  • Bionic? Ken perhaps?
  • memcpy for Newlib
    • Ours is better for A9, other is better for A15
    • No define in general to tell what you're tuning for
    • Want something feature like?
    • GLIBC was CPU vs subarch but is compatibility instead of tuning
    • On GLIBC could do at runtime (but has similar problems?)
    • Has it been asked on the newlib list?
      • was leading towards 'should GCC define something'
    • Richard E would like a coherent proposal for this
  • Update the blueprints to:
    • Links to any upstream discussions
    • Discussion on open problems, and your preferred solution

Rebuild:

  • Matthias is planning to do a 4.7 archive rebuild
  • Michael is pessimistic...
  • The ITM branch is still landing, may cause trouble with sync items
  • Will do the build as a hard float

WorkingGroups/ToolChain/Meetings/Archive/2011-12-19 (last modified 2013-08-30 11:47:49)