Tuesday 29th August 2011

This month's meetings

WorkingGroups/ToolChain/Meetings
<< <  2011 / 8 >  >>
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
  • Unaligned access patch
  • Shrink wrap patch
  • 823711 libvirt due to alias visibility (backport)

  • 836401 ICE on a | (b << negative-constant)

  • 823708 / 809761 ICE in get_arm_condition_code()

  • Sync primitives
  • 1.0 release of cortex-strings

Action Items from this Meeting

  • See below

Action Items from Previous Meeting

  • DONE: Andrew to send Martin an email re: things being OK
  • DONE: Michael will run the stable merge tomorrow
  • DONE: Michael to postpone the meeting by a day

Minutes

Shrink wrap:

  • Has been divided into three patches
  • Structure ones have gone in
  • Viable to backport to 4.6? Could be quite invasive
  • Benchmark off trunk beforehand to see if its worthwhile?

Unaligned accesses:

  • No comments from the group
  • Will keep an eye on it, please everyone do the same

Sync primitives:

  • Status?
  • Ramana/ARM has not reviewed the two GCC patches
  • Blocked upstream on GCC
  • Kernel is in the 3.1 tree, already in Linaro
  • libgcc is part of the same series in GCC review
  • QEMU patch looks OK, send to qemu-devel
    • Reasonable to get it in by 2011.09
  • Can we make GCCs job easier?
    • Currently split into four chunks
    • Two are fixes found along the way
    • ARMv6 version (libgcc change)
    • ...and the one that uses the instructions
    • New tests, C changes, and .md changes
  • Backport should be trivial
  • Revisit this next week once Ramana is back.
  • Michael is keen to get it in...

cortex-strings:

  • memcpy, memset that are as good as GLIBC?
    • memcpy is good in most cases
    • memset
    • ACTION: Dave to see where they're up to. Check them in if decent enough
  • How do we benchmark?
    • Has memchr and examples where it is better
    • No justification for replacing GLIBCs memcpy
    • Have EEMBC and SPEC results, don't show anything huge
    • Do we have a workload profile?
    • Yip, his earlier latrace run on part of SPEC
  • How's memchr going up to GLIBC?
    • Joseph had sent back feedback
    • Michael's eglibc build is better but not best
    • Dave has some input on that
    • What classes did the feedback fit into?
      • None on the routine
      • Feedback on how it figures out what ARM it will compile for
      • Some on boilerplate
      • Most is on machine/submachine
      • Will sort out / 'pipe clean' for the next one

WorkingGroups/ToolChain/Meetings/Archive/2011-08-30 (last modified 2013-08-30 11:49:25)