Tuesday 29th August 2011
This month's meetings
<< < 2011 / 8 > >>
- 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)
- 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
- 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?
- No comments from the group
- Will keep an eye on it, please everyone do the same
- 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...
- memcpy, memset that are as good as GLIBC?
- memcpy is good in most cases
- 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)