about summary refs log tree commit diff
path: root/bits/floatn.h
diff options
context:
space:
mode:
authorJoseph Myers <joseph@codesourcery.com>2017-11-17 22:01:43 +0000
committerJoseph Myers <joseph@codesourcery.com>2017-11-17 22:01:43 +0000
commit015c6dc288003cd6b192f5901a8df2ea16205ee4 (patch)
treefe864d1d425b88fe43d51127590599f64c45a7ea /bits/floatn.h
parentce003e5d4cd94c5380699b0dadeaaf825813afbe (diff)
downloadglibc-015c6dc288003cd6b192f5901a8df2ea16205ee4.tar.gz
glibc-015c6dc288003cd6b192f5901a8df2ea16205ee4.tar.xz
glibc-015c6dc288003cd6b192f5901a8df2ea16205ee4.zip
Support bits/floatn.h inclusion from .S files.
Further _FloatN / _FloatNx type alias support will involve making
architecture-specific .S files use the common macros for libm function
aliases.  Making them use those macros will also serve to simplify
existing code for aliases / symbol versions in various cases, similar
to such simplifications for ldbl-opt code.

The libm-alias-*.h files sometimes need to include <bits/floatn.h> to
determine which aliases they should define.  At present, this does not
work for inclusion from .S files because <bits/floatn.h> can define
typedefs for old compilers.  This patch changes all the
<bits/floatn.h> and <bits/floatn-common.h> headers to include
__ASSEMBLER__ conditionals.  Those conditionals disable everything
related to C syntax in the __ASSEMBLER__ case, not just the problem
typedefs, as that seemed cleanest.  The __HAVE_* definitions remain in
the __ASSEMBLER__ case, as those provide information that is required
to define the correct set of aliases.

Tested with build-many-glibcs.py for a representative set of
configurations (x86_64-linux-gnu i686-linux-gnu ia64-linux-gnu
powerpc64le-linux-gnu mips64-linux-gnu-n64 sparc64-linux-gnu) with GCC
6.  Also tested with GCC 6 for i686-linux-gnu in conjunction with
changes to use alias macros in .S files.

	* bits/floatn-common.h [!__ASSEMBLER]: Disable everything related
	to C syntax instead of availability and properties of types.
	* bits/floatn.h [!__ASSEMBLER]: Likewise.
	* sysdeps/ia64/bits/floatn.h [!__ASSEMBLER]: Likewise.
	* sysdeps/ieee754/ldbl-128/bits/floatn.h [!__ASSEMBLER]: Likewise.
	* sysdeps/mips/ieee754/bits/floatn.h [!__ASSEMBLER]: Likewise.
	* sysdeps/powerpc/bits/floatn.h [!__ASSEMBLER]: Likewise.
	* sysdeps/x86/bits/floatn.h [!__ASSEMBLER]: Likewise.
Diffstat (limited to 'bits/floatn.h')
-rw-r--r--bits/floatn.h8
1 files changed, 6 insertions, 2 deletions
diff --git a/bits/floatn.h b/bits/floatn.h
index 78eb0f47ab..dc0cf93cbf 100644
--- a/bits/floatn.h
+++ b/bits/floatn.h
@@ -25,13 +25,17 @@
    from the default float, double and long double types in this glibc.  */
 #define __HAVE_DISTINCT_FLOAT128 0
 
+#ifndef __ASSEMBLER__
+
 /* Defined to concatenate the literal suffix to be used with _Float128
    types, if __HAVE_FLOAT128 is 1.
    E.g.: #define __f128(x) x##f128.  */
-#undef __f128
+# undef __f128
 
 /* Defined to a complex binary128 type if __HAVE_FLOAT128 is 1.
    E.g.: #define __CFLOAT128 _Complex _Float128.  */
-#undef __CFLOAT128
+# undef __CFLOAT128
+
+#endif /* !__ASSEMBLER__.  */
 
 #include <bits/floatn-common.h>