Re: [PATCH v2 00/10] Rework READ_ONCE() to improve codegen

[Arnd Bergmann]

On Sun, Jan 26, 2020 at 2:10 AM Qais Yousef <qais.yousef@xxxxxxx> wrote:
> On 01/23/20 14:01, Arvind Sankar wrote:
> > On Thu, Jan 23, 2020 at 10:45:08AM -0800, Nick Desaulniers wrote:
> > > On Thu, Jan 23, 2020 at 9:32 AM David Laight <David.Laight@xxxxxxxxxx> wrote:
> >
> > Reposting Arnd's link
> > https://www.spinics.net/lists/linux-kbuild/msg23648.html
>
> This list seems to be x86 centric? I remember when the switch to GCC 4.6
> happened a couple or more archs had to be dropped because they lacked a newer
> compiler.

There are two architectures that already had problems last time:

- unicore32 never had any compiler that shipped with sources, only an ancient
  set of gcc binaries that already had problems building the kernel during the
  move to gcc-4.6. The maintainer said he'd work on providing support for
  modern gcc or clang, but I don't think anything came out of that.

- hexagon had an unmaintained gcc-4.5 port, but internally Qualcomm were
  already using clang to build their kernels, which should now work with the
  upstream version. I don't think there are any plans to have a more modern
  gcc.

Everything else works with mainline gcc now, openrisc and csky were the
last to get added in gcc-9.

Some of the older sub-targets (armv3, s390-g6, powerpcspe) are removed
in gcc-9, but these have a few more years before we need to worry about
them.

     Arnd