Make common fdim implementation generic.
commit7b7c39450b3c4ab35b4960346e61d7b177ee728e
authorPaul E. Murphy <murphyp@linux.vnet.ibm.com>
Thu, 25 Aug 2016 16:25:33 +0000 (25 11:25 -0500)
committerPaul E. Murphy <murphyp@linux.vnet.ibm.com>
Thu, 1 Sep 2016 14:28:05 +0000 (1 09:28 -0500)
tree251612e6e9a4192b1e5ca8010589ec64a692b9fa
parent4d728087ef8cc826b05bd21d0c74d4eca9b1a27d
Make common fdim implementation generic.

The only difference is the usage of math_narrow_eval when
building s_fdiml.c.  This should be harmless for long double,
but I did observe some code generation changes on m68k, but
lack the resources to test it.

Likewise, to more easily support overriding symbol generation,
the aliasing macros are always conditionally defined on their
absence to reduce boilerplate.

I also ran builds for i486, ppc64, sparcv9, aarch64,
s390x and observed no changes to s_fdim* objects.
14 files changed:
ChangeLog
math/Makefile
math/s_fdim.c [deleted file]
math/s_fdim_template.c [moved from math/s_fdimf.c with 84% similarity]
math/s_fdiml.c [deleted file]
sysdeps/generic/math-type-macros-double.h
sysdeps/ieee754/ldbl-opt/math-type-macros-double.h
sysdeps/ieee754/ldbl-opt/math-type-macros-ldouble.h
sysdeps/ieee754/ldbl-opt/s_fdim.c [deleted file]
sysdeps/ieee754/ldbl-opt/s_fdiml.c [deleted file]
sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_fdim-vis3.c
sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_fdim.c
sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_fdimf-vis3.c
sysdeps/sparc/sparc32/sparcv9/fpu/multiarch/s_fdimf.c