Monday 14th Feb 2011
This month's meetings
<<MonthCalendar: execution failed [Argument "anniversary" must be a boolean value, not "workinggroups/toolchain/meetingtemplate"] (see also the log)>>
- Review action items from last meeting
- IBM contribution status
- FSF copyright assignment status
- Profiling tools: perf vs gprof
- Number of iterations in DENBench
- What's the next step?
- Who needs hardware?
- GDB watchpoints
- 4.6 status
- Realview PBX status
IBM contribution status:
- Looking into the issue from an IBM approvals status
- Decision: as work done in the Linaro context is for Linaro, follow Linaro rules
- Question comes up: what are Linaro rules?
- One includes having appropriate, approved licenses
- Do need explicit approval from TSC for GPLv3, is in progress
FSF copyright assignment:
- Has been 'put in the mail'
- Michael will check on status
- Ken's: run Ubuntu Natty alpha 2, not Linaro. Will track down.
- Ulrich: OK
Ira: has a BeagleBoard, but it's broken
- Revital: OK on Michael's
- Peter: has XM, otherwise OK
- perf or gprof?
- Ken: gprof is based on instrumenting, perf doesn't
- Revital: DENBench can easily use gprof
- Michael: how much does gprof change the results?
- Michael will add to the builds
- Regresses in ARMv7 vs ARMv5
- Regresses in NEON vs just core
- Ramana: use perf to see how the instruction selection changes
- Chung-Lin: have a look at the disassembly
- Michael said that Realview PBX is there, but a bit useless from an end-user point of view
- Discussion going on with ARM re: what the well supported A9 should be
- Andrew is ticking on
- Stuck a bit as Richard is on sabbatical until the end of the month
- Wants to benchmark on A9, waiting on hardware
- Patches done for the A8, might be appropriate for A9, at least need to be conditional
- Michael will make ursa2 available and idle
- Ramana will review patches
- When should we start landing in the 4.6 branch?
- Andrew: upstream seems a bit stuck, may start merging into the consolidation
- For everyone else: start porting to the Linaro 4.6 and 4.5 branch
GDB watchpoint status:
- Ulrich: got things working as expected on the A9
- A8 is harder as it uses the memory mapped access to the watchpoint registers
- Things could be improved, such as:
- Use more than one watchpoint (has more)
- Seems to be some fundamental SMP related bugs in software breakpoints
- Works fine with CPU affinity
- Michael: higher priority than hardware watchpoints
- Ramana: prod Will about this?
WorkingGroups/ToolChain/Meetings/Archive/2011-02-14 (last modified 2013-08-30 11:49:49)