[Ksummit-2012-discuss] [ATTEND] ARM architecture, profiling with asymmetric event sources

Will Deacon will.deacon at arm.com
Fri Jun 22 17:27:05 UTC 2012


The majority of work and discussion around the ARM port is currently
focussed on platform code and soc support, as shown by the git traffic
generated from the new arm-soc tree.

However, there are still a few of us working on the core architecture
which now has support for large physical addresses, virtualisation and
other exciting bits and pieces which require work in the core port.
There's also a 64-bit architecture on the way.

Alongside these developments, the addition and improvement of core code
with architecture-specific parts (e.g. audit, word-at-a-time, perf, etc)
means we can end up in a couple of situations:

  1. Playing `catch-up' with other architectures where a feature is
     simply not yet enabled for ARM.

  2. Introduction of broken arch code via a subsystem tree where
     somebody tried to add ARM support but got it wrong. This sometimes
     happens because a previously correct assumption no longer holds
     true on ARM (e.g. 7-argument syscalls and tracing).

With this in mind, my selfish reasons for wanting to attend this year's
kernel summit are to see what proposals others have and ensure that we
can support them on current and future ARM platforms. On the less
selfish side, I can provide input on the architecture and am happy to
help with hacking (especially if there's another ARM subarchitecture
maintainer mini-summit). I also maintain the ARM perf and debugging
support, so I can bore you with my thoughts on profiling for big.LITTLE
systems and across SoCs. I might also count as a `younger contributor',
but that's subjective :)

Finally, I probably owe tglx a beer.


More information about the Ksummit-2012-discuss mailing list