Tuesday 29 Nov

This month's meetings

WorkingGroups/ToolChain/Meetings
<< <  2011 / 11 >  >>
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        

Attendees

Agenda

  • Review action items from last meeting
  • AOB.

Action Items from this Meeting

  • TBA

Action Items from Previous Meeting

  • Revital to check on A8 perf numbers - DONE
  • Ramana to check on vmla cases. Still Investigating.

Minutes

  • Revital spoke about the numbers with SMS and register pressure estimation -
    • libav (numbers with register pressure estimation and without) - sms with and without doloop optimization and without register pressure estimation. Discussion on fixed registers and register pressure to be continued over email . Richard says that it really isn't a problem given that IRA only considers the allocatable registers. What about the patch with SMS and without the doloop_end pattern ? Revital had some fails with this patch but will investigate further ?
    • Dynamic stack realignment with the ARM backend similar to the x86 backend for stating the STACK_BOUNDARY. Dynamic stack realignment - Richard to look into the implementation of the dynamic stack realignment and propose what could be done in the ARM backend. What is the benefit of going from 64 to 128 bit alignment ? It might be worth doing the peeling to align up to 128 bits - cost benefit analysis.
    • Discussion on general thoughts on what to avoid on the A15
      • can think of ldm/stm vs ldrd / strd.
    • 4.5 vs 4.6 regression in one case through constant pool vs movw/movt https://bugs.launchpad.net/gcc-linaro/+bug/886124 . Ramana to take a look .

    • 64 bit neon stuff

WorkingGroups/ToolChain/Meetings/Archive/2011-11-29 (last modified 2013-08-30 11:49:09)