In <http://gcc.gnu.org/ml/gcc-patches/2013-10/msg01131.html>, I noted
that, as I'd previously noted in
<http://gcc.gnu.org/ml/gcc-patches/2010-10/msg01951.html>, ISO C
conformance modes should set -ffp-contract=on (= -ffp-contract=off)
rather than the default -ffp-contract=fast, given the ISO C
restriction to contraction taking place within source language
expressions. No-one indicated any reason not to do this, so I've now
applied this patch to do so (and to make an explicit
-ffp-contract=fast in ISO C mode imply __GCC_IEC_559=0). Bootstrapped
with no regressions on x86_64-unknown-linux-gnu.
c-family:
2013-11-06 Joseph Myers <jos...@codesourcery.com>
* c-opts.c (c_common_post_options): Set -ffp-contract=off in C
standards modes.
* c-cppbuiltin.c (cpp_iec_559_value): Consider -ffp-contract=fast
to mean lack of IEEE 754 support.
testsuite:
2013-11-06 Joseph Myers <jos...@codesourcery.com>
* gcc.dg/torture/c99-contract-1.c: New test.
Index: gcc/c-family/c-cppbuiltin.c
===================================================================
--- gcc/c-family/c-cppbuiltin.c (revision 204453)
+++ gcc/c-family/c-cppbuiltin.c (working copy)
@@ -726,16 +726,19 @@ cpp_iec_559_value (void)
ret = 0;
/* In strict C standards conformance mode, consider unpredictable
- excess precision to mean lack of IEEE 754 support. ??? The same
- should apply to unpredictable contraction, but at present
- standards conformance options do not enable conforming
- contraction. For C++, and outside strict conformance mode, do
- not consider these options to mean lack of IEEE 754 support. */
+ excess precision to mean lack of IEEE 754 support. The same
+ applies to unpredictable contraction. For C++, and outside
+ strict conformance mode, do not consider these options to mean
+ lack of IEEE 754 support. */
if (flag_iso
&& !c_dialect_cxx ()
&& TARGET_FLT_EVAL_METHOD != 0
&& flag_excess_precision_cmdline != EXCESS_PRECISION_STANDARD)
ret = 0;
+ if (flag_iso
+ && !c_dialect_cxx ()
+ && flag_fp_contract_mode == FP_CONTRACT_FAST)
+ ret = 0;
/* Various options are contrary to IEEE 754 semantics. */
if (flag_unsafe_math_optimizations
Index: gcc/c-family/c-opts.c
===================================================================
--- gcc/c-family/c-opts.c (revision 204453)
+++ gcc/c-family/c-opts.c (working copy)
@@ -827,6 +827,15 @@ c_common_post_options (const char **pfilename)
? EXCESS_PRECISION_STANDARD
: EXCESS_PRECISION_FAST);
+ /* ISO C restricts floating-point expression contraction to within
+ source-language expressions (-ffp-contract=on, currently an alias
+ for -ffp-contract=off). */
+ if (flag_iso
+ && !c_dialect_cxx ()
+ && (global_options_set.x_flag_fp_contract_mode
+ == (enum fp_contract_mode) 0))
+ flag_fp_contract_mode = FP_CONTRACT_OFF;
+
/* By default we use C99 inline semantics in GNU99 or C99 mode. C99
inline semantics are not supported in GNU89 or C89 mode. */
if (flag_gnu89_inline == -1)
Index: gcc/testsuite/gcc.dg/torture/c99-contract-1.c
===================================================================
--- gcc/testsuite/gcc.dg/torture/c99-contract-1.c (revision 0)
+++ gcc/testsuite/gcc.dg/torture/c99-contract-1.c (revision 0)
@@ -0,0 +1,21 @@
+/* Test floating-point contraction occurs only within source language
+ expressions. */
+/* { dg-do run } */
+/* { dg-options "-std=c99 -pedantic-errors" } */
+
+extern void abort (void);
+extern void exit (int);
+
+volatile float a = 1 + 0x1p-23f, b = 1 - 0x1p-23f, c = -1;
+
+int
+main (void)
+{
+ float av = a, bv = b, cv = c;
+ float p = av * bv;
+ float r = p + cv;
+ if (r == 0)
+ exit (0);
+ else
+ abort ();
+}
--
Joseph S. Myers
jos...@codesourcery.com
