GCC currently defaults to -ftrapping-math. This is supposed to generate
code for correct user-visible traps and FP status flags.
However it doesn't work as expected since it doesn't block any floating
point optimizations. For example it continues to perform CSE, moves FP
operations across calls, moves FP operations out of loops, constant folds
and removes dead floating point operations that cause exceptions.
Given the majority of code doesn't contain user trap handlers or inspects
FP status flags, there is no point in enabling it even if it worked as expected.
Simple case that should cause a FP exception:
void f(void)
{
0.0 / 0.0;
}
Compiles to:
f:
ret
OK for commit?
2017-11-16 Wilco Dijkstra <wdijk...@arm.com>
* common.opt (ftrapping-math): Change default to 0.
* doc/invoke.texi (-ftrapping-math): Update documentation.
--
diff --git a/gcc/common.opt b/gcc/common.opt
index
1bb87353f760d7c60c39de8b9de4311c1ec3d892..59940c64356964f8f9b9d842ad3f1a1c02548bab
100644
--- a/gcc/common.opt
+++ b/gcc/common.opt
@@ -2470,7 +2470,7 @@ generate them instead of using descriptors.
; (user-visible) trap. This is the case, for example, in nonstop
; IEEE 754 arithmetic.
ftrapping-math
-Common Report Var(flag_trapping_math) Init(1) Optimization SetByCombined
+Common Report Var(flag_trapping_math) Init(0) Optimization SetByCombined
Assume floating-point operations can trap.
ftrapv
diff --git a/gcc/doc/invoke.texi b/gcc/doc/invoke.texi
index
74c33ea35b9f320b419a3417e6007d2391536f1b..3673b34b3b7f7b57cfa6375b5316f9f282a9e9bb
100644
--- a/gcc/doc/invoke.texi
+++ b/gcc/doc/invoke.texi
@@ -9479,20 +9479,21 @@ This option implies that the sign of a zero result
isn't significant.
The default is @option{-fsigned-zeros}.
-@item -fno-trapping-math
-@opindex fno-trapping-math
-Compile code assuming that floating-point operations cannot generate
+@item -ftrapping-math
+@opindex ftrapping-math
+Compile code assuming that floating-point operations can generate
user-visible traps. These traps include division by zero, overflow,
-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.
+underflow, inexact result and invalid operation.
-This option should never be turned on by any @option{-O} option since
-it can result in incorrect output for programs that depend on
-an exact implementation of IEEE or ISO rules/specifications for
-math functions.
+Note this option has only been partially implemented and does not work
+as expected. For example @option{-ftrapping-math} performs floating
+point optimizations such as loop invariant motion, constant folding
+and scheduling across function calls which have user-visible effects
+on FP exception flags.
+
+This option is turned on when using @option{-fsignaling-nans}.
-The default is @option{-ftrapping-math}.
+The default is @option{-fno-trapping-math}.
@item -frounding-math
@opindex frounding-math
