Patch/RFC: -mlongcall and #pragma longcall for PowerPC

[Stan Shebs]

David Edelsohn wrote:
> 
>         Any reason that you list the symbolic call patterns before the
> indirect call patterns?
> 
>         Also, when you document the longcall pragma, would you please make
> it clear to which GCC configurations it applies.  SVR4?  eABI?  It
> definitely does not apply to AIX and Darwin.

Although it could make sense for Darwin - in fact we have a local
patch that uses a different code sequence (jbsr) to build drivers,
which are loaded dynamically at fixed addresses, possibly far
from the kernel.  (We've studied switching to longcall, although
the kernel folks are reluctant unless there are demonstrable
performance improvements.)

Stan