[PATCH] rs6000: Fix 2 for PR92661, Do not define builtins that overload disabled builtins

[Segher Boessenkool]

On Mon, Dec 09, 2019 at 02:16:51PM -0600, Peter Bergner wrote:
> I'm not sure that's necessary.  DFP enablement isn't triggered by
> assembler support.  Just the gcc/configure fragment (ignoring manually
> using --enable-decimal-float):
> 
>   case $target in
>     powerpc*-*-linux* | i?86*-*-linux* | x86_64*-*-linux* | s390*-*-linux* | \
>     i?86*-*-elfiamcu | i?86*-*-gnu* | x86_64*-*-gnu* | \
>     i?86*-*-mingw* | x86_64*-*-mingw* | \
>     i?86*-*-cygwin* | x86_64*-*-cygwin*)
>       enable_decimal_float=yes

Hmm.  Do we still only support it on Linux?  Is that a bit too
conservative?  The only thing required from the OS is treating the
FPSCR as a 64-bit register (instead of 32-bit), which is required by
the architecture since, what, ISA 2.05?  12 years ago?

OTOH, if there is no demand for this elsewhere, the status quo will
work fine of course, and it's simple as can be.

This configure thing does not gate whether the backend code supports
DFP though?  Or is that handled somewhere?  I.e. can we ever end up
with TARGET_DFP set while enable_decimal_float is not set?


Segher