This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
[PATCH] PR libgcc/59714 complex division is surprising on aarch64
- From: vladimir dot mezentsev at oracle dot com
- To: gcc-patches at gcc dot gnu dot org
- Cc: vladimir dot mezentsev at oracle dot com
- Date: Wed, 25 Oct 2017 10:23:31 -0700
- Subject: [PATCH] PR libgcc/59714 complex division is surprising on aarch64
- Authentication-results: sourceware.org; auth=none
From: Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
FMA (floating-point multiply-add) instructions are supported on aarch64.
These instructions can produce different result if two operations executed separately.
-ffp-contract=off doesn't allow the FMA instructions.
Tested on aarch64-unknown-linux-gnu. No regression. Two failed tests now passed.
ChangeLog:
2017-10-25 Vladimir Mezentsev <vladimir.mezentsev@oracle.com>
PR libgcc/59714
* libgcc/Makefile.in: Add -ffp-contract=off
---
libgcc/Makefile.in | 8 ++++++++
1 file changed, 8 insertions(+)
diff --git a/libgcc/Makefile.in b/libgcc/Makefile.in
index a1a392d..f313556 100644
--- a/libgcc/Makefile.in
+++ b/libgcc/Makefile.in
@@ -455,6 +455,14 @@ ifeq ($(LIB2_SIDITI_CONV_FUNCS),)
lib2funcs += $(subst XX,di,$(dwfloatfuncs))
endif
+# Disable FMA (floating-point multiply-add) instructions for complex division.
+# These instructions can produce different result if two operations executed separately.
+LIBGCC2_FFP_CONTRAST_CFLAGS = -ffp-contract=off
+LIB2_DIV3_FUNCS = _divdc3 _divhc3 _divsc3
+lib2-div3-o = $(patsubst %,%$(objext),$(LIB2_DIV3_FUNCS))
+lib2-div3-s-o = $(patsubst %,%_s$(objext),$(LIB2_DIV3_FUNCS))
+$(lib2-div3-o) $(lib2-div3-s-o): LIBGCC2_CFLAGS += $(LIBGCC2_FFP_CONTRAST_CFLAGS)
+
# These might cause a divide overflow trap and so are compiled with
# unwinder info.
LIB2_DIVMOD_FUNCS = _divdi3 _moddi3 _divmoddi4 \
--
1.8.3.1