https://gcc.gnu.org/g:d317a05ebc91cb1da2a89d9ea9fe5f7dc0747b8a
commit d317a05ebc91cb1da2a89d9ea9fe5f7dc0747b8a
Author: Michael Meissner <meiss...@linux.ibm.com>
Date: Fri Jul 12 20:06:45 2024 -0400
Revert changes
Diff:
---
gcc/ChangeLog.bugs | 56 ++++++++++++------------------------------------------
1 file changed, 12 insertions(+), 44 deletions(-)
diff --git a/gcc/ChangeLog.bugs b/gcc/ChangeLog.bugs
index 61066c019ccd..f2f6f7986da8 100644
--- a/gcc/ChangeLog.bugs
+++ b/gcc/ChangeLog.bugs
@@ -1,49 +1,14 @@
-==================== Branch work171-bugs, patch #304 ====================
+==================== Branch work171-bugs, patch #310 ====================
-Use -mcpu=native instead of -mcpu=power9 when checking for float128 hardware
support.
-
-2024-07-12 Michael Meissner <meiss...@linux.ibm.com>
-
-gcc/testsuite/
-
- PR target/115800
- PR target/113652
- * lib/target-supports.exp (check_ppc_float128_hw_available): Use
- -mcpu=native rather than -mcpu=power9 to check if the machine has
- float128 hardware support.
-
-==================== Branch work171-bugs, patch #303 ====================
-
-Disable adding -mvsx when checking for float128 support.
-
-2024-07-12 Michael Meissner <meiss...@linux.ibm.com>
-
-gcc/testsuite/
-
- PR target/115800
- PR target/113652
- * lib/target-supports.exp (check_ppc_float128_sw_available): Do not add
- -mvsx when checking for float128 support.
- (check_ppc_float128_hw_available): Likewise.
- (check_effective_target___float128): Likewise.
-
-==================== Branch work171-bugs, patch #302 ====================
-
-Do not build IEEE 128-bit libgcc support if VSX is not available.
+Add checks for building the IEEE 128-bit libgcc support if VSX is not
available.
In the past, we would build libgcc and eable the float128 libraries by adding
-mvsx -mabi=altivec to build the support libraries. However, this causes
-problems if the default cpu is a 7450.
-
-With this fix, in order to build the float128 support, the compiler must be
-configured to default to at least power7 to enable using the VSX register set,
-which is required for passing float128 values.
-
-If somebody wanted to enable float128 on big endian systems, they would need to
-use a compiler that defaults to at least power7.
+problems if the default cpu is a 7450. This patch checks to see if can link a
+program using the options, and if we can't link the program, do not build the
+float128 libraries.
-
-2024-07-11 Michael Meissner <meiss...@linux.ibm.com>
+2024-07-12 Michael Meissner <meiss...@linux.ibm.com>
libgcc/
@@ -52,11 +17,14 @@ libgcc/
* config.host (powerpc*-*-linux*): Do not enable the float128 hardware
and float128 power10 hardware support unless the basic float128 support
is added.
- * config/rs6000/t-float128 (FP128_CFLAGS_SW): Do not add -mvsx.
- * configure.ac (powerpc*-*-linux*): Don't enable IEEE 128-bit on PowerPC
- systems without VSX.
+ * configure.ac (powerpc*-*-linux*): Check whether the linker and
+ assembler will be able to build the float128 libraries.
* configure: Regenerate.
+==================== Branch work171-bugs, patch #304 was reverted
====================
+==================== Branch work171-bugs, patch #303 was reverted
====================
+==================== Branch work171-bugs, patch #302 was reverted
====================
+
==================== Branch work171-bugs, patch #301 ====================
Do not build IEEE 128-bit libstdc++ support if VSX is not available.
