[12/12] ldbl-128ibm-compat: Provide ieee128 symbols to narrow functions

Message ID 20180620020426.20372-13-tuliom@linux.ibm.com
State New
Headers show
Series
  • Introduce ieee128 symbols and redirections
Related show

Commit Message

Tulio Magno Quites Machado Filho June 20, 2018, 2:04 a.m.
Add ieee128 aliases for faddl, fdivl, fmull, fsubl, daddl, ddivl, dmull, dsubl.

2018-06-19  Tulio Magno Quites Machado Filho  <tuliom@linux.ibm.com>

	* math/math-narrow.h [__HAVE_DISTINCT_FLOAT128 &&
	!__HAVE_FLOAT128_UNLIKE_LDBL] (libm_alias_float32_float128,
	libm_alias_float64_float128): Provide *ieee128 aliases.
	* sysdeps/ieee754/ldbl-128ibm-compat/Versions: Add
	__f32addieee128, __f32divieee128, __f32mulieee128,
	__f32subieee128, __f64addieee128, __f64divieee128,
	__f64mulieee128 and __f64subieee128.

Signed-off-by: Tulio Magno Quites Machado Filho <tuliom@linux.ibm.com>

---
 math/math-narrow.h                          | 10 ++++++++++
 sysdeps/ieee754/ldbl-128ibm-compat/Versions |  8 ++++++++
 2 files changed, 18 insertions(+)

-- 
2.14.4

Comments

Joseph Myers June 20, 2018, 10:55 a.m. | #1
On Wed, 20 Jun 2018, Tulio Magno Quites Machado Filho wrote:

> Add ieee128 aliases for faddl, fdivl, fmull, fsubl, daddl, ddivl, dmull, dsubl.

> 

> 2018-06-19  Tulio Magno Quites Machado Filho  <tuliom@linux.ibm.com>

> 

> 	* math/math-narrow.h [__HAVE_DISTINCT_FLOAT128 &&

> 	!__HAVE_FLOAT128_UNLIKE_LDBL] (libm_alias_float32_float128,


I don't think __HAVE_FLOAT128_UNLIKE_LDBL is logically right here.  That's 
about the current compilation, i.e. about whether these particular files 
end up getting built with IEEE long double.  I don't know what files in 
the source tree might end up getting built with what long double setting, 
but it shouldn't matter for these files whether they are built with IEEE 
long double or not.

Your property is something different, "glibc has *ieee128 aliases for 
_Float128 functions" (a property of the glibc configuration, not of the 
current compilation, but a different property from 
__HAVE_DISTINCT_FLOAT128).  That suggests to me that either you need some 
new macro tested by this file, or you need e.g. 
math-narrow-alias-float128.h (included by math-narrow.h) which contains 
these float128 alias macros and for which ldbl-128ibm-compat provides a 
different version.

-- 
Joseph S. Myers
joseph@codesourcery.com

Patch

diff --git a/math/math-narrow.h b/math/math-narrow.h
index c4065e88f7..bfc1c8dced 100644
--- a/math/math-narrow.h
+++ b/math/math-narrow.h
@@ -356,6 +356,16 @@ 
   libm_alias_float32_float128_main (func)
 # define libm_alias_float64_float128(func)	\
   libm_alias_float64_float128_main (func)
+#elif __HAVE_DISTINCT_FLOAT128 && !__HAVE_FLOAT128_UNLIKE_LDBL
+# define libm_alias_float32_float128(func)			\
+  libm_alias_float32_float128_main (func)			\
+  weak_alias (__f32 ## func ## f128, f32 ## func ## f64x)	\
+  weak_alias (__f32 ## func ## f128, __ ## f32 ## func ## ieee128)
+# define libm_alias_float64_float128(func)			\
+  libm_alias_float64_float128_main (func)			\
+  weak_alias (__f64 ## func ## f128, f64 ## func ## f64x)	\
+  weak_alias (__f64 ## func ## f128, f32x ## func ## f64x)	\
+  weak_alias (__f64 ## func ## f128, __ ## f64 ## func ## ieee128)
 #else
 # define libm_alias_float32_float128(func)			\
   libm_alias_float32_float128_main (func)			\
diff --git a/sysdeps/ieee754/ldbl-128ibm-compat/Versions b/sysdeps/ieee754/ldbl-128ibm-compat/Versions
index 01e67aec1a..8fc6fb05f3 100644
--- a/sysdeps/ieee754/ldbl-128ibm-compat/Versions
+++ b/sysdeps/ieee754/ldbl-128ibm-compat/Versions
@@ -47,6 +47,14 @@  libm {
     __exp2ieee128;
     __expieee128;
     __expm1ieee128;
+    __f32addieee128;
+    __f32divieee128;
+    __f32mulieee128;
+    __f32subieee128;
+    __f64addieee128;
+    __f64divieee128;
+    __f64mulieee128;
+    __f64subieee128;
     __fabsieee128;
     __fdimieee128;
     __floorieee128;