What's coming up for PowerPC

[Justin Hibbits]

To all PowerPC-dwellers,

There has been a lot of exiting work going on in PowerPC-space that
we're gearing up to commit for FreeBSD 13 in the next month or two.

First, the news that transcends all PowerPC platforms.  We're gearing
up for a complete switch to Clang/LLVM as our base compiler.  This has
taken effort from many people, both on and off the FreeBSD commit
team.  Special thanks go to:

* Brandon Bergren -- our newest committer.  He's been working on ELFv2
  for nearly a year now, and has been playing with newer GCC and Clang
  the whole time.
* Alfredo Dal'Ava Junior -- He's been leading the effort for completing
  Clang for FreeBSD/powerpc64, and found some interesting things for
  powerpc (32).  Namely, that LLD was missing, but now has, support for
  Secure-PLT (still missing BSS-PLT), which I'll mention more of below.
* Leandro Lupori -- He's been working across the board, high and low
  level debugging, bugfixing, and development.

I'm sure others as well, and apologize to those not mentioned.

So, Clang is pretty close to being ready for powerpc64, and we are
setting a flag day for probably mid-July, once we polish everything
up.  This brings us to powerpc and powerpcspe:
* Support was recently added completing(?) powerpc in LLD, but using
  Secure-PLT, which is used almost everywhere but FreeBSD currently.
  This is about to change.  There is a patch up at
  https://reviews.freebsd.org/D20598 migrating FreeBSD/powerpc and
  FreeBSD/powerpcspe to Secure-PLT for FreeBSD 13.  This is
  inter-operable with BSS-PLT which we currently use, but will require
  a compiler rebuild when it's committed.
* There are patches in-flight adding support for powerpcspe into clang,
  and improving support in llvm.  These may start as local patches for
  us while finishing the work upstream as well, but that is TBD.

The goal is for all powerpc flavors to migrate to Clang/LLD as soon as
possible.  This will bring us roughly 12 years up to date in terms of
compiler technology, giving us new capabilities, including IFUNCs for
CPU_specific optimized functions in base, better overall optimizations
across the board, and other improvements that come with a modern
toolchain.

Thanks to everyone participating in this effort.  We couldn't have made
this level of progress without the drive of everyone involved.

- Justin