This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: [RFC] Fix PR tree-optimization/34036
> Probably as well. I guess to get all cases here will be work though :/
It seems that there are only 2 affected testcases. OK for mainline?
2007-11-18 Eric Botcazou <ebotcazou@libertysurf.fr>
PR tree-optimization/34036
* opts.c (set_fast_math_flags): Do not set flag in common
with set_unsafe_math_optimizations_flags, invoke it directly.
(set_unsafe_math_optimizations_flags): Unset flag_trapping_math
and flag_signed_zeros.
* toplev.c (process_options): Force flag_associative_math to 0 and
warn, if it is set and either flag_trapping_math or flag_signed_zeros
are set too.
* doc/invoke.texi (Optimize Options): Document the new behavior
of -funsafe-math-optimizations, that -fassociative-math requires
both -fno-signed-zeros and -fno-trapping-math and make it clear
that -fno-trapping-math requires -fno-signaling-nans.
2007-11-18 Eric Botcazou <ebotcazou@libertysurf.fr>
* gcc.dg/pr33007.c: Expect new warning.
* gcc.dg/tree-ssa/recip-5.c: Likewise.
* g++.dg/opt/pr34036.C: New test.
--
Eric Botcazou
Index: doc/invoke.texi
===================================================================
--- doc/invoke.texi (revision 130131)
+++ doc/invoke.texi (working copy)
@@ -6273,7 +6273,8 @@ it can result in incorrect output for pr
an exact implementation of IEEE or ISO rules/specifications for
math functions. It may, however, yield faster code for programs
that do not require the guarantees of these specifications.
-Enables @option{-freciprocal-math} and @option{-fassociative-math}.
+Enables @option{-fno-signed-zeros}, @option{-fno-trapping-math},
+@option{-fassociative-math} and @option{-freciprocal-math}.
The default is @option{-fno-unsafe-math-optimizations}.
@@ -6287,8 +6288,9 @@ well as ignore NaNs and inhibit or creat
thus cannot be used on a code which relies on rounding behavior like
@code{(x + 2**52) - 2**52)}. May also reorder floating-point comparisons
and thus may not be used when ordered comparisons are required.
-This flag doesn't make much sense without @option{-fno-signed-zeros}
-or @option{-fno-trapping-math} or with @option{-frounding-math}.
+This option requires that both @option{-fno-signed-zeros} and
+@option{-fno-trapping-math} be in effect. Moreover, it doesn't make
+much sense with @option{-frounding-math}.
The default is @option{-fno-associative-math}.
@@ -6330,9 +6332,9 @@ The default is @option{-fsigned-zeros}.
@opindex fno-trapping-math
Compile code assuming that floating-point operations cannot generate
user-visible traps. These traps include division by zero, overflow,
-underflow, inexact result and invalid operation. This option implies
-@option{-fno-signaling-nans}. Setting this option may allow faster
-code if one relies on ``non-stop'' IEEE arithmetic, for example.
+underflow, inexact result and invalid operation. This option requires
+that @option{-fno-signaling-nans} be in effect. Setting this option may
+allow faster code if one relies on ``non-stop'' IEEE arithmetic, for example.
This option should never be turned on by any @option{-O} option since
it can result in incorrect output for programs which depend on
Index: opts.c
===================================================================
--- opts.c (revision 130131)
+++ opts.c (working copy)
@@ -1868,12 +1868,9 @@ set_Wstrict_aliasing (int onoff)
void
set_fast_math_flags (int set)
{
- flag_trapping_math = !set;
flag_unsafe_math_optimizations = set;
- flag_associative_math = set;
- flag_reciprocal_math = set;
+ set_unsafe_math_optimizations_flags (set);
flag_finite_math_only = set;
- flag_signed_zeros = !set;
flag_errno_math = !set;
if (set)
{
@@ -1888,8 +1885,10 @@ set_fast_math_flags (int set)
void
set_unsafe_math_optimizations_flags (int set)
{
- flag_reciprocal_math = set;
+ flag_trapping_math = !set;
+ flag_signed_zeros = !set;
flag_associative_math = set;
+ flag_reciprocal_math = set;
}
/* Return true iff flags are set as if -ffast-math. */
Index: toplev.c
===================================================================
--- toplev.c (revision 130131)
+++ toplev.c (working copy)
@@ -1988,6 +1988,13 @@ process_options (void)
if (flag_signaling_nans)
flag_trapping_math = 1;
+ /* We cannot reassociate if we want traps or signed zeros. ?*/
+ if (flag_associative_math && (flag_trapping_math || flag_signed_zeros))
+ {
+ warning (0, "-fassociative-math disabled; other options take precedence");
+ flag_associative_math = 0;
+ }
+
/* With -fcx-limited-range, we do cheap and quick complex arithmetic. */
if (flag_cx_limited_range)
flag_complex_method = 0;
Index: testsuite/gcc.dg/pr33007.c
===================================================================
--- testsuite/gcc.dg/pr33007.c (revision 130131)
+++ testsuite/gcc.dg/pr33007.c (working copy)
@@ -1,5 +1,6 @@
/* { dg-do compile } */
/* { dg-options "-O -ffast-math -ftrapping-math" } */
+/* { dg-warning "-fassociative-math disabled" "" { target *-*-* } 1 } */
long
foo (int i)
Index: testsuite/gcc.dg/tree-ssa/recip-5.c
===================================================================
--- testsuite/gcc.dg/tree-ssa/recip-5.c (revision 130131)
+++ testsuite/gcc.dg/tree-ssa/recip-5.c (working copy)
@@ -1,5 +1,6 @@
/* { dg-options "-O1 -funsafe-math-optimizations -ftrapping-math -fdump-tree-recip -fdump-tree-optimized" } */
/* { dg-do compile } */
+/* { dg-warning "-fassociative-math disabled" "" { target *-*-* } 1 } */
/* Test the reciprocal optimizations together with trapping math. */