__float128 on ppc64le

[iv-m]

I'm trying to compile boost 1.85.0 with gcc 13.2.1 on ppc64le machine that does not have a 'real' long double:

LDBL_MANT_DIG: 106
LDBL_MAX_EXP: 1024

But on this system libquadmath is available and __float128 seems to work.

I'm getting the following error:

[...]
libs/charconv/build/../src/from_chars.cpp: In function 'boost::charconv::from_chars_result boost::charconv::from_chars_erange(const char*, const char*, __ieee128&, chars_format)':
libs/charconv/build/../src/from_chars.cpp:120:48: error: 'compute_float128' is not a member of 'boost::charconv::detail'; did you mean 'compute_float32'?
  120 |     auto return_val = boost::charconv::detail::compute_float128(exponent, significand, sign, success);
      |                                                ^~~~~~~~~~~~~~~~
      |                                                compute_float32
libs/charconv/build/../src/from_chars.cpp: In function 'boost::charconv::from_chars_result boost::charconv::from_chars_erange(const char*, const char*, long double&, chars_format)':
libs/charconv/build/../src/from_chars.cpp:258:48: error: 'compute_float80' is not a member of 'boost::charconv::detail'; did you mean 'compute_float64'?
  258 |     auto return_val = boost::charconv::detail::compute_float80<long double>(exponent, significand, sign, success);
      |                                                ^~~~~~~~~~~~~~~
      |                                                compute_float64
[...]

[iv-m]

It looks like BOOST_CHARCONV_LDBL_BITS is not defined on this system, but BOOST_CHARCONV_HAS_FLOAT128 is. So boost/charconv/detail/compute_float80.hpp is not included by from_chars.cpp, but from_chars_erange overload for __float128 is not preprocessed away.

[mborland]

Are you trying to convert that long double type (__ibm128)? From the docs: Long doubles can be 64, 80, or 128-bit, but must be IEEE 754 compliant. An example of a non-compliant, and therefore unsupported, format is __ibm128.

[mborland]

Actually if you are trying with __float128 can you try latest develop? I reworked __float128 and quadmath support since release freeze: #177

[pdimov]

I don't think he's trying to use long double or __float128, the library just fails to build.

[mborland]

I don't think he's trying to use long double or __float128, the library just fails to build.

I was able to replicate on develop with -mabi=ibmlongdouble -mlong-double-128 on fedora 39. I forgot that it had transitioned to __float128: https://developers.redhat.com/articles/2023/05/16/benefits-fedora-38-long-double-transition-ppc64le#

[iv-m]

I was able to replicate on develop with -mabi=ibmlongdouble -mlong-double-128 on fedora 39. I forgot that it had transitioned to __float128: https://developers.redhat.com/articles/2023/05/16/benefits-fedora-38-long-double-transition-ppc64le#

Yup, Fedora transitioned to ieeelongdouble, but I'm not using Fedora, so we still have ibmlongdouble as the default.

[mborland]

@iv-m can you please give #183 a try? Everything seemed fine on QEMU Fedora 39 with both long double ABIs, but I want to make sure there's not some difference from emulation.

[iv-m]

Sorry for the delay, my ppc64le test environment was broken and it took some time to rebuild it.

Current develop branch (100473b, with #183 merged) works for me when explicitly specifying C++ standard -- for example, the following command run from the top of the boost tree completes successfully:

b2 cxxstd=11 charconv

But when I don't specify cxxstd, the default is used, which for my compiler is gnu++17. This enables GNU extensions, including __float128 and Q literals. Their presence is correctly detected by the build system, and the compilation fails:

In file included from libs/charconv/build/../src/to_chars.cpp:8:
libs/charconv/build/../src/to_chars_float_impl.hpp: In function 'boost::charconv::to_chars_result boost::charconv::detail::to_chars_hex(char*, char*, Real, int)':
libs/charconv/build/../src/to_chars_float_impl.hpp:277:140: error: 'ieee754_binary80' was not declared in this scope; did you mean 'ieee754_binary64'?
  277 |                     typename std::conditional<std::is_same<Real, __float128>::value || BOOST_CHARCONV_LDBL_BITS == 128, ieee754_binary128, ieee754_binary80>::type
      |                                                                                                                                            ^~~~~~~~~~~~~~~~
      |                                                                                                                                            ieee754_binary64
libs/charconv/build/../src/to_chars_float_impl.hpp:277:156: error: template argument 3 is invalid
  277 |                     typename std::conditional<std::is_same<Real, __float128>::value || BOOST_CHARCONV_LDBL_BITS == 128, ieee754_binary128, ieee754_binary80>::type
      |                                                                                                                                                            ^
libs/charconv/build/../src/to_chars_float_impl.hpp:277:157: error: '<declaration error>' is not a template [-fpermissive]
  277 |                     typename std::conditional<std::is_same<Real, __float128>::value || BOOST_CHARCONV_LDBL_BITS == 128, ieee754_binary128, ieee754_binary80>::type
      |                                                                                                                                                             ^~
libs/charconv/build/../src/to_chars_float_impl.hpp:285:13: error: template argument 3 is invalid
  285 |             >::type>::type
      |             ^
libs/charconv/build/../src/to_chars_float_impl.hpp:285:14: error: '<declaration error>' is not a template [-fpermissive]
  285 |             >::type>::type
      |              ^~
libs/charconv/build/../src/to_chars_float_impl.hpp:285:20: error: template argument 3 is invalid
  285 |             >::type>::type
      |                    ^
libs/charconv/build/../src/to_chars_float_impl.hpp:285:21: error: '<declaration error>' is not a template [-fpermissive]
  285 |             >::type>::type
      |                     ^~
libs/charconv/build/../src/to_chars_float_impl.hpp:302:67: error: 'type_layout' has not been declared
  302 |     const Unsigned_Integer denorm_mask = (Unsigned_Integer(1) << (type_layout::significand_bits)) - 1;
      |                                                                   ^~~~~~~~~~~
libs/charconv/build/../src/to_chars_float_impl.hpp:304:61: error: 'type_layout' has not been declared
  304 |     std::uint64_t exponent = extract_exp(value, uint_value, type_layout::significand_bits);
      |                                                             ^~~~~~~~~~~
libs/charconv/build/../src/to_chars_float_impl.hpp:365:33: error: 'type_layout' has not been declared
  365 |         unbiased_exponent = 1 + type_layout::exponent_bias;
      |                                 ^~~~~~~~~~~
libs/charconv/build/../src/to_chars_float_impl.hpp:370:40: error: 'type_layout' has not been declared
  370 |         unbiased_exponent = exponent + type_layout::exponent_bias;
      |                                        ^~~~~~~~~~~

[mborland]

Unfortunately I am unable to replicate that. Running the tests with b2 cxxstd=11 cxxflags="-mabi=ibmlongdouble -mlong-double-128" and with b2 cxxstd=11 cxxstd-dialect=gnu cxxflags="-mabi=ibmlongdouble -mlong-double-128" both pass the entire test suite with the latter correctly detecting support for quadmath.