[PATCH] libstdc++: Don't fail if math_errhandling is not defined

[PATCH] libstdc++: Don't fail if math_errhandling is not defined

[Matthias Kretz]

[-- Attachment #1: Type: text/plain, Size: 1165 bytes --]

Tested on x86_64-linux and powerpc64le-linux. This fixes the powerpc64le-linux 
failure Jonathan reported. OK for trunk?


Older glibc does not define math_errhandling with -ffast-math, in which
case floating-point exceptions are not used.

Signed-off-by: Matthias Kretz <m.kretz@gsi.de>

libstdc++-v3/ChangeLog:

	* include/experimental/bits/simd.h (__floating_point_flags): Do
	not rely on the presence of the math_errhandling macro.
---
 libstdc++-v3/include/experimental/bits/simd.h | 11 ++++++++++-
 1 file changed, 10 insertions(+), 1 deletion(-)


--
──────────────────────────────────────────────────────────────────────────
 Dr. Matthias Kretz                           https://mattkretz.github.io
 GSI Helmholtz Centre for Heavy Ion Research               https://gsi.de
 stdₓ::simd
──────────────────────────────────────────────────────────────────────────

[-- Attachment #2: 0001-libstdc-Don-t-fail-if-math_errhandling-is-not-define.patch --]
[-- Type: text/x-patch, Size: 790 bytes --]

diff --git a/libstdc++-v3/include/experimental/bits/simd.h b/libstdc++-v3/include/experimental/bits/simd.h
index d837d0efb21..c991e3f223e 100644
--- a/libstdc++-v3/include/experimental/bits/simd.h
+++ b/libstdc++-v3/include/experimental/bits/simd.h
@@ -283,11 +283,20 @@ constexpr inline bool __have_power_vmx = __have_power_vsx;
 
 namespace __detail
 {
+  constexpr bool __handle_fpexcept =
+#ifdef math_errhandling
+    math_errhandling & MATH_ERREXCEPT;
+#elif defined __FAST_MATH__
+    false;
+#else
+    true;
+#endif
+
   constexpr std::uint_least64_t
   __floating_point_flags()
   {
     std::uint_least64_t __flags = 0;
-    if constexpr (math_errhandling & MATH_ERREXCEPT)
+    if constexpr (__handle_fpexcept)
       __flags |= 1;
 #ifdef __FAST_MATH__
     __flags |= 1 << 1;

Re: [PATCH] libstdc++: Don't fail if math_errhandling is not defined

[Jonathan Wakely]

On Mon, 17 Jan 2022 at 10:22, Matthias Kretz <m.kretz@gsi.de> wrote:

> Tested on x86_64-linux and powerpc64le-linux. This fixes the
> powerpc64le-linux
> failure Jonathan reported. OK for trunk?
>

OK, thanks.



>
> Older glibc does not define math_errhandling with -ffast-math, in which
> case floating-point exceptions are not used.
>
> Signed-off-by: Matthias Kretz <m.kretz@gsi.de>
>
> libstdc++-v3/ChangeLog:
>
>         * include/experimental/bits/simd.h (__floating_point_flags): Do
>         not rely on the presence of the math_errhandling macro.
> ---
>  libstdc++-v3/include/experimental/bits/simd.h | 11 ++++++++++-
>  1 file changed, 10 insertions(+), 1 deletion(-)
>
>
> --
> ──────────────────────────────────────────────────────────────────────────
>  Dr. Matthias Kretz                           https://mattkretz.github.io
>  GSI Helmholtz Centre for Heavy Ion Research               https://gsi.de
>  stdₓ::simd
> ──────────────────────────────────────────────────────────────────────────