[gcc r14-9689] libstdc++: Fix call signature of builtins from masked ++/--

[gcc r14-9689] libstdc++: Fix call signature of builtins from masked ++/--

[Matthias Kretz]

https://gcc.gnu.org/g:0ac2c0f0687b321ab54de271d788b4e0a287b4e2

commit r14-9689-g0ac2c0f0687b321ab54de271d788b4e0a287b4e2
Author: Matthias Kretz <m.kretz@gsi.de>
Date:   Wed Mar 27 08:49:43 2024 +0100

    libstdc++: Fix call signature of builtins from masked ++/--
    
    This resolves failures in the "expensive" where-* test of check-simd
    when targeting AVX-512.
    
    Signed-off-by: Matthias Kretz <m.kretz@gsi.de>
    
    libstdc++-v3/ChangeLog:
    
            * include/experimental/bits/simd_x86.h (_S_masked_unary): Call
            the 4- and 8-byte variants of __builtin_ia32_subp[ds] without
            rounding direction argument.

Diff:
---
 libstdc++-v3/include/experimental/bits/simd_x86.h | 12 +++++++++---
 1 file changed, 9 insertions(+), 3 deletions(-)

diff --git a/libstdc++-v3/include/experimental/bits/simd_x86.h b/libstdc++-v3/include/experimental/bits/simd_x86.h
index 16b207be2a3..6b414486fee 100644
--- a/libstdc++-v3/include/experimental/bits/simd_x86.h
+++ b/libstdc++-v3/include/experimental/bits/simd_x86.h
@@ -3538,17 +3538,23 @@ template <typename _Abi, typename>
 	      }
 	    else
 	      {
-#define _GLIBCXX_SIMD_MASK_SUB(_Sizeof, _Width, _Instr)                        \
+#define _GLIBCXX_SIMD_MASK_SUB_512(_Sizeof, _Width, _Instr)                    \
   if constexpr (sizeof(_Tp) == _Sizeof && sizeof(__v) == _Width)               \
     return __builtin_ia32_##_Instr##_mask(                                     \
 	     __v._M_data, __vector_broadcast<_Np>(_Tp(__pm_one)), __v._M_data, \
 	     __k._M_data, _MM_FROUND_CUR_DIRECTION)
-		_GLIBCXX_SIMD_MASK_SUB(4, 64, subps512);
+#define _GLIBCXX_SIMD_MASK_SUB(_Sizeof, _Width, _Instr)                        \
+  if constexpr (sizeof(_Tp) == _Sizeof && sizeof(__v) == _Width)               \
+    return __builtin_ia32_##_Instr##_mask(                                     \
+	     __v._M_data, __vector_broadcast<_Np>(_Tp(__pm_one)), __v._M_data, \
+	     __k._M_data)
+		_GLIBCXX_SIMD_MASK_SUB_512(4, 64, subps512);
 		_GLIBCXX_SIMD_MASK_SUB(4, 32, subps256);
 		_GLIBCXX_SIMD_MASK_SUB(4, 16, subps128);
-		_GLIBCXX_SIMD_MASK_SUB(8, 64, subpd512);
+		_GLIBCXX_SIMD_MASK_SUB_512(8, 64, subpd512);
 		_GLIBCXX_SIMD_MASK_SUB(8, 32, subpd256);
 		_GLIBCXX_SIMD_MASK_SUB(8, 16, subpd128);
+#undef _GLIBCXX_SIMD_MASK_SUB_512
 #undef _GLIBCXX_SIMD_MASK_SUB
 	      }
 #endif // __clang__