This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
PATCH: Handle TARGET_96_ROUND_53_LONG_DOUBLE
- From: Loren James Rittle <rittle at latour dot labs dot mot dot com>
- To: gcc-patches at gcc dot gnu dot org
- Date: Wed, 30 Sep 2009 11:30:10 -0500 (CDT)
- Subject: PATCH: Handle TARGET_96_ROUND_53_LONG_DOUBLE
- Reply-to: rittle at labs dot mot dot com
This patch kills another 8 gratitious FAILs for
TARGET_96_ROUND_53_LONG_DOUBLE targets (of which, only
i386-unknown-freebsd* is publicly known).
OK, to commit?
On a related note: I think, but am not yet sure, that
gcc.c-torture/execute/conversion.c is failing the 2 sub-tests [as
noted in http://gcc.gnu.org/ml/gcc-bugs/2009-09/msg02378.html] because
sizeof (long long) is greater than LDBL_MANT_DIG (and the test
actually assumes otherwise). Does that sound right?
Regards,
Loren
2009-09-30 Loren J. Rittle <ljrittle@acm.org>
* gcc.c-torture/execute/ieee/copysign1.c: Handle
TARGET_96_ROUND_53_LONG_DOUBLE.
* gcc.c-torture/execute/ieee/copysign2.c: Likewise.
Index: gcc/testsuite/gcc.c-torture/execute/ieee/copysign1.c
===================================================================
--- gcc/testsuite/gcc.c-torture/execute/ieee/copysign1.c (revision 152307)
+++ gcc/testsuite/gcc.c-torture/execute/ieee/copysign1.c (working copy)
@@ -10,7 +10,7 @@
we've got a 10 byte type stuffed into some amount of padding. And
the fact that -ffloat-store is going to stuff this value temporarily
into some bit of stack frame that we've no control over and can't zero. */
-#if LDBL_MANT_DIG == 64
+#if LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 53
# if defined(__i386__) || defined(__x86_64__) || defined (__ia64__)
# undef fpsizeofl
# define fpsizeofl 10
Index: gcc/testsuite/gcc.c-torture/execute/ieee/copysign2.c
===================================================================
--- gcc/testsuite/gcc.c-torture/execute/ieee/copysign2.c (revision 152307)
+++ gcc/testsuite/gcc.c-torture/execute/ieee/copysign2.c (working copy)
@@ -10,7 +10,7 @@
we've got a 10 byte type stuffed into some amount of padding. And
the fact that -ffloat-store is going to stuff this value temporarily
into some bit of stack frame that we've no control over and can't zero. */
-#if LDBL_MANT_DIG == 64
+#if LDBL_MANT_DIG == 64 || LDBL_MANT_DIG == 53
# if defined(__i386__) || defined(__x86_64__) || defined (__ia64__)
# undef fpsizeofl
# define fpsizeofl 10