diff options
author | Paul E. Murphy <murphyp@linux.vnet.ibm.com> | 2016-07-20 11:53:36 -0500 |
---|---|---|
committer | Gabriel F. T. Gomes <gftg@linux.vnet.ibm.com> | 2017-05-15 10:08:56 -0300 |
commit | fcee5905d341fe975f7786e72c831ada1c8fa78b (patch) | |
tree | 20c16f302d00271a3316d33b12dd46c64741022b /ChangeLog | |
parent | e1d6e08dd6f9298ed7f61ef12c17ca83cb761f37 (diff) | |
download | glibc-fcee5905d341fe975f7786e72c831ada1c8fa78b.tar.gz glibc-fcee5905d341fe975f7786e72c831ada1c8fa78b.tar.xz glibc-fcee5905d341fe975f7786e72c831ada1c8fa78b.zip |
float128: Add public _Float128 declarations to libm.
This introduces the machine-dependent bits/floatn.h to control the inclusion of _Float128 ABI. * bits/floatn.h: New file. * bits/huge_val_flt128.h: Likewise. * bits/libm-simd-decl-stubs.h (__DECL_SIMD_cosf128): New macro. (__DECL_SIMD_sinf128): Likewise. (__DECL_SIMD_sincosf128): Likewise. (__DECL_SIMD_logf128): Likewise. (__DECL_SIMD_expf128): Likewise. (__DECL_SIMD_powf128): Likewise. * math/Makefile (headers): Install bits/floatn.h and bits/huge_val_flt128.h. * math/bits/cmathcalls.h (_Mdouble_complex_): Only define if not defined. * math/bits/mathcalls.h (pow10, isinf, finite, drem, significand) (isnan, gamma, nexttoward, scalb): Only define if not __MATH_DECLARING_FLOATN. (nextdown, nextup, llogb, roundeven, fromfp, ufromfp, fromfpx) (ufromfpx, fmaxmag, fminmag, totalorder, totalordermag) (canonicalize, getpayload, setpayload, setpayloadsig): Declare for _FloatN even if __STDC_WANT_IEC_60559_BFP_EXT__ is not defined. * math/complex.h: Include bits/libc-header-start.h and bits/floatn.h. Include bits/cmathcalls.h to get the declarations of _Float128 versions of complex functions. [__HAVE_FLOAT128 && __GLIBC_USE (IEC_60559_TYPES_EXT)] (CMPLXF128): Define macro. * math/math.h: Include bits/floatn.h and bits/huge_val_flt128.h. Include bits/mathcalls-helper-functions.h and bits/mathcalls.h to get the declarations of _Float128 versions of math functions. (__MATH_DECLARING_FLOATN): New macro to control declaration of _FloatN types. [__GNUC_PREREQ (6,0)] (signbit): Since GCC 6.0, __builtin_signbit is type-generic, so use it without __MATH_TG. [__HAVE_DISTINCT_FLOAT128] (isinf): This builtin is broken for _Float128 type on GCC < 7.0. Explicitly call __isinff128 for _Float128 type and GCC < 7.0, otherwise use the builtin. [__HAVE_FLOAT128 && defined __USE_GNU] (M_Ef128, M_LOG2Ef128) (M_LOG10Ef128, M_LN2f128, M_LN10f128, M_PIf128, M_PI_2f128) (M_PI_4f128, M_1_PIf128, M_2_PIf128, M_SQRT2f128, M_SQRT1_2f128): New _GNU_SOURCE enabled macros.
Diffstat (limited to 'ChangeLog')
-rw-r--r-- | ChangeLog | 48 |
1 files changed, 48 insertions, 0 deletions
diff --git a/ChangeLog b/ChangeLog index ece37e9d7e..22352418f4 100644 --- a/ChangeLog +++ b/ChangeLog @@ -1,3 +1,51 @@ +2017-05-14 Paul E. Murphy <murphyp@linux.vnet.ibm.com> + Gabriel F. T. Gomes <gftg@linux.vnet.ibm.com> + + * bits/floatn.h: New file. + * bits/huge_val_flt128.h: Likewise. + + * bits/libm-simd-decl-stubs.h (__DECL_SIMD_cosf128): New macro. + (__DECL_SIMD_sinf128): Likewise. + (__DECL_SIMD_sincosf128): Likewise. + (__DECL_SIMD_logf128): Likewise. + (__DECL_SIMD_expf128): Likewise. + (__DECL_SIMD_powf128): Likewise. + + * math/Makefile (headers): Install bits/floatn.h and + bits/huge_val_flt128.h. + + * math/bits/cmathcalls.h (_Mdouble_complex_): Only define if not + defined. + + * math/bits/mathcalls.h (pow10, isinf, finite, drem, significand) + (isnan, gamma, nexttoward, scalb): Only define if not + __MATH_DECLARING_FLOATN. + (nextdown, nextup, llogb, roundeven, fromfp, ufromfp, fromfpx) + (ufromfpx, fmaxmag, fminmag, totalorder, totalordermag) + (canonicalize, getpayload, setpayload, setpayloadsig): Declare for + _FloatN even if __STDC_WANT_IEC_60559_BFP_EXT__ is not defined. + + * math/complex.h: Include bits/libc-header-start.h and bits/floatn.h. + Include bits/cmathcalls.h to get the declarations of _Float128 + versions of complex functions. + [__HAVE_FLOAT128 && __GLIBC_USE (IEC_60559_TYPES_EXT)] (CMPLXF128): + Define macro. + + * math/math.h: Include bits/floatn.h and bits/huge_val_flt128.h. + Include bits/mathcalls-helper-functions.h and bits/mathcalls.h to + get the declarations of _Float128 versions of math functions. + (__MATH_DECLARING_FLOATN): New macro to control declaration of + _FloatN types. + [__GNUC_PREREQ (6,0)] (signbit): Since GCC 6.0, __builtin_signbit + is type-generic, so use it without __MATH_TG. + [__HAVE_DISTINCT_FLOAT128] (isinf): This builtin is broken for + _Float128 type on GCC < 7.0. Explicitly call __isinff128 for + _Float128 type and GCC < 7.0, otherwise use the builtin. + [__HAVE_FLOAT128 && defined __USE_GNU] (M_Ef128, M_LOG2Ef128) + (M_LOG10Ef128, M_LN2f128, M_LN10f128, M_PIf128, M_PI_2f128) + (M_PI_4f128, M_1_PIf128, M_2_PIf128, M_SQRT2f128, M_SQRT1_2f128): + New _GNU_SOURCE enabled macros. + 2017-05-12 Joseph Myers <joseph@codesourcery.com> * sysdeps/sparc/sparc32/fpu/s_signbit.S (signbit): Remove alias. |