[builtins] Avoid using long double in generic sources

[arichardson]

Use of long double can be error-prone since it could be one of 80-bit
extended precision float, IEEE 128-bit float, or IBM 128-bit float.
Instead use an explicit xf_float typedef for the remaining cases where
long double is being used in the implementation. This patch does not touch
the PPC specializations which still use long double.

Depends on #68132 (to be committed after that PR).

[github-actions]

✅ With the latest revision this PR passed the C/C++ code formatter.

[nico]

This seems to not build on Windows, at least with clang-cl as host compiler: http://45.33.8.238/win/85105/step_4.txt

Could you take a look?

[arichardson]

This seems to not build on Windows, at least with clang-cl as host compiler: http://45.33.8.238/win/85105/step_4.txt

Could you take a look?

My guess is that HAS_80_BIT_LONG_DOUBLE is not being defined for some reason?

#if (defined(__i386__) || defined(__x86_64__)) &&                              \
    !(defined(_MSC_VER) || defined(__ANDROID__))
#define HAS_80_BIT_LONG_DOUBLE 1
#elif defined(__m68k__) || defined(__ia64__)
#define HAS_80_BIT_LONG_DOUBLE 1
#else
#define HAS_80_BIT_LONG_DOUBLE 0
#endif

#if HAS_80_BIT_LONG_DOUBLE

[arichardson]

So either clang-cl does not use 80 bit long double and we shouldn't be compiling those files or the guard needs to be adjusted. Not familiar with Windows ABIs so I'd have to defer to someone with experience there.

[pranavk]

Yes, Windows long double is 64 bits, not 80 bits.

[arichardson]

Yes, Windows long double is 64 bits, not 80 bits.

Does this mean the *xf files should not be compiled for Windows ABIs?

[pranavk]

Yes, that would make sense to me. *xf pertains to 80 bit float as far as I know.

Microsoft's long double

[nico]

Apologies, https://reviews.llvm.org/D82153 already took care of that a while ago. Config problem on that bot.