[PATCH] libstdc++: Fix SFINAE for __is_intrinsic_type on ARM

[PATCH] libstdc++: Fix SFINAE for __is_intrinsic_type on ARM

[Matthias Kretz]

[-- Attachment #1: Type: text/plain, Size: 1267 bytes --]

OK for master and all branches? (this issue only surfaced because of the new 
test)

------------ 8< -----------------

On ARM NEON doesn't support double, so __is_intrinsic_type_v<double,
whatever> should say false (instead of being ill-formed).

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

libstdc++-v3/ChangeLog:

	PR libstdc++/109261
	* include/experimental/bits/simd.h (__intrinsic_type):
	Specialize __intrinsic_type<double, 8> and
	__intrinsic_type<double, 16> in any case, but provide the member
	type only with __aarch64__.
---
 libstdc++-v3/include/experimental/bits/simd.h | 12 +++++++++---
 1 file changed, 9 insertions(+), 3 deletions(-)


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

[-- Attachment #2: 0001-libstdc-Fix-SFINAE-for-__is_intrinsic_type-on-ARM.patch --]
[-- Type: text/x-patch, Size: 942 bytes --]

diff --git a/libstdc++-v3/include/experimental/bits/simd.h b/libstdc++-v3/include/experimental/bits/simd.h
index b0571ca26c4..d1f388310f9 100644
--- a/libstdc++-v3/include/experimental/bits/simd.h
+++ b/libstdc++-v3/include/experimental/bits/simd.h
@@ -2369,15 +2369,21 @@ struct __intrinsic_type<float, 8, void>
   struct __intrinsic_type<float, 16, void>
   { using type = float32x4_t; };
 
-#if _GLIBCXX_SIMD_HAVE_NEON_A64
 template <>
   struct __intrinsic_type<double, 8, void>
-  { using type = float64x1_t; };
+  {
+#if _GLIBCXX_SIMD_HAVE_NEON_A64
+   using type = float64x1_t;
+#endif
+  };
 
 template <>
   struct __intrinsic_type<double, 16, void>
-  { using type = float64x2_t; };
+  {
+#if _GLIBCXX_SIMD_HAVE_NEON_A64
+    using type = float64x2_t;
 #endif
+  };
 
 #define _GLIBCXX_SIMD_ARM_INTRIN(_Bits, _Np)                                   \
 template <>                                                                    \

Re: [PATCH] libstdc++: Fix SFINAE for __is_intrinsic_type on ARM

[Jonathan Wakely]

[-- Attachment #1: Type: text/plain, Size: 1493 bytes --]

On Wed, 24 May 2023 at 11:59, Matthias Kretz via Libstdc++ <
libstdc++@gcc.gnu.org> wrote:

> OK for master and all branches? (this issue only surfaced because of the
> new
> test)
>

OK.



>
> ------------ 8< -----------------
>
> On ARM NEON doesn't support double, so __is_intrinsic_type_v<double,
> whatever> should say false (instead of being ill-formed).
>
> Signed-off-by: Matthias Kretz <m.kretz@gsi.de>
>
> libstdc++-v3/ChangeLog:
>
>         PR libstdc++/109261
>         * include/experimental/bits/simd.h (__intrinsic_type):
>         Specialize __intrinsic_type<double, 8> and
>         __intrinsic_type<double, 16> in any case, but provide the member
>         type only with __aarch64__.
> ---
>  libstdc++-v3/include/experimental/bits/simd.h | 12 +++++++++---
>  1 file changed, 9 insertions(+), 3 deletions(-)
>
>
> --
> ──────────────────────────────────────────────────────────────────────────
>  Dr. Matthias Kretz                           https://mattkretz.github.io
>  GSI Helmholtz Centre for Heavy Ion Research               https://gsi.de
>  stdₓ::simd
> ──────────────────────────────────────────────────────────────────────────