Re: Revisiting More Complete long double Support

On Aug 18 17:16, Joel Sherrill wrote:
> I've dug in a bit and tried to bring in one function. Found
> a few things that are going to shape the solution:
> 
> (1)  Newlib and FreeBSD do not agree on filenames. truncl.c
>       gets replaced by s_truncl.c
>      (a) Great work on the new build system. :)

We can easily change this to the FreeBSD naming convention.  We're
using it mostly anyway.

> (2) FreeBSD long double methods rely on _fpmath.h to define
>      the long double format. It doesn't have to be 128 bits for the
>      code to work. But only a handful of architectures have this file:
> 
>       aarch64 amd64 arm i386 mips powerpc powerpc64 riscv sparc64
> 
>       This file would go in libc/machine/... I think based on the FreeBSD
>       source layout.

_fpmath.h defines one union and 5 macros.  It's very basic.

We have already multiple files defining structs and unions based on the
value of LDBL_MANT_DIG, libc/include/ieeefp.h,
libc/include/machine/ieee.h, and libc/stdio/vfieeefp.h.

The most useful way of action would probably be to define a common
_fpmath.h with all the required structs, unions and macros and use
that throughout, i. e., we include it from the above files.

And ideally we only have one *ieee*.h file, not three, in future...

> (3) I think my suggestion to have the default implementation and
>       then this implementation in parallel and then select the appropriate
>       one at configure time. For the architectures with _fpmath.h, the
>       new FreeBSD implementation can be used. For others, architectures
>       the old implementation could be used.

Dunno about others, but I'd rather not see two parallel implementations
of math in newlib.  It might be a bit more work to port the code, but
it's certainly not unfeasible.


Corinna