diff options
| author | Joseph Myers <joseph@codesourcery.com> | 2015-10-28 21:57:10 +0000 |
|---|---|---|
| committer | Joseph Myers <joseph@codesourcery.com> | 2015-10-28 21:57:10 +0000 |
| commit | 5878dd9d53a00fa28800236d79e38e55f85fc98d (patch) | |
| tree | 85ce332ade1b23eb4ece3c34f6382133a8d5734b /ChangeLog | |
| parent | 1f4dafa3ea24df63a6550dbdbbe8769caa7a9fc3 (diff) | |
| download | glibc-5878dd9d53a00fa28800236d79e38e55f85fc98d.zip glibc-5878dd9d53a00fa28800236d79e38e55f85fc98d.tar.gz glibc-5878dd9d53a00fa28800236d79e38e55f85fc98d.tar.bz2 | |
Use C11 CMPLX* macros in libm tests.
libm-test.inc has a macro BUILD_COMPLEX to construct a complex number
with given real and imaginary parts while allowing properly for signed
zeroes, infinities and NaNs (which don't work properly with a simple
real + I * imag, in the absence of compiler support for imaginary
types), using assignment to __real__ and __imag__ parts of the number.
C11 defines CMPLX* macros for this purpose, which GCC 4.7 and above
provide suitable built-in functions for. This patch redefines
BUILD_COMPLEX in terms of the standard macros.
Tested for x86_64 and x86.
* math/libm-test.inc (BUILD_COMPLEX): Remove macro.
* math/test-double.h (BUILD_COMPLEX): New macro.
* math/test-float.h (BUILD_COMPLEX): Likewise.
* math/test-ldouble.h (BUILD_COMPLEX): Likewise.
Diffstat (limited to 'ChangeLog')
| -rw-r--r-- | ChangeLog | 5 |
1 files changed, 5 insertions, 0 deletions
@@ -1,5 +1,10 @@ 2015-10-28 Joseph Myers <joseph@codesourcery.com> + * math/libm-test.inc (BUILD_COMPLEX): Remove macro. + * math/test-double.h (BUILD_COMPLEX): New macro. + * math/test-float.h (BUILD_COMPLEX): Likewise. + * math/test-ldouble.h (BUILD_COMPLEX): Likewise. + * math/libm-test.inc (min_subnorm_value): Use LDBL_TRUE_MIN, DBL_TRUE_MIN and FLT_TRUE_MIN instead of __LDBL_DENORM_MIN__, __DBL_DENORM_MIN__ and __FLT_DENORM_MIN__. |