Re: [PATCH 2/7] libstdc++: Annotate most lambdas with always_inline

[Jonathan Wakely via Gcc-patches]

On Wed, 15 Feb 2023 at 20:50, Matthias Kretz via Libstdc++
 wrote:
>
>
>
> All of the annotated lambdas are simply a necessary means for
> implementing these functions and should never result in an actual
> function call. Many of these lambdas would go away if C++ had better
> language support for packs.

(I hope we'll get p1061 for C++26.)

OK for trunk.

N.B. I noticed some pre-existing cases of a non-reserved name "Parts"
there, which should be fixed (separately from this patch though).

[PATCH 2/7] libstdc++: Annotate most lambdas with always_inline

[Matthias Kretz via Gcc-patches]

All of the annotated lambdas are simply a necessary means for
implementing these functions and should never result in an actual
function call. Many of these lambdas would go away if C++ had better
language support for packs.

Signed-off-by: Matthias Kretz 

libstdc++-v3/ChangeLog:

PR libstdc++/108030
* include/experimental/bits/simd_detail.h: Define
_GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA.
* include/experimental/bits/simd.h: Annotate lambdas with
_GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA.
* include/experimental/bits/simd_builtin.h: Ditto.
* include/experimental/bits/simd_converter.h: Ditto.
* include/experimental/bits/simd_fixed_size.h: Ditto.
* include/experimental/bits/simd_math.h: Ditto.
* include/experimental/bits/simd_neon.h: Ditto.
* include/experimental/bits/simd_x86.h: Ditto.
---
 libstdc++-v3/include/experimental/bits/simd.h | 239 ++--
 .../include/experimental/bits/simd_builtin.h  | 351 ++
 .../experimental/bits/simd_converter.h|  22 +-
 .../include/experimental/bits/simd_detail.h   |   3 +
 .../experimental/bits/simd_fixed_size.h   | 265 ++---
 .../include/experimental/bits/simd_math.h |  52 +--
 .../include/experimental/bits/simd_neon.h |  14 +-
 .../include/experimental/bits/simd_x86.h  | 122 +++---
 8 files changed, 575 insertions(+), 493 deletions(-)


--
──
 Dr. Matthias Kretz   https://mattkretz.github.io
 GSI Helmholtz Centre for Heavy Ion Research   https://gsi.de
 stdₓ::simd
──diff --git a/libstdc++-v3/include/experimental/bits/simd.h b/libstdc++-v3/include/experimental/bits/simd.h
index 3de966bbf22..ffe72fa6ccf 100644
--- a/libstdc++-v3/include/experimental/bits/simd.h
+++ b/libstdc++-v3/include/experimental/bits/simd.h
@@ -609,28 +609,34 @@ template 
 	  operator&(_Ip __rhs) const
 	  {
 	return __generate_from_n_evaluations<_Np, _Ip>(
-	  [&](auto __i) { return __rhs._M_data[__i] & _M_data[__i]; });
+	  [&](auto __i) _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA {
+		return __rhs._M_data[__i] & _M_data[__i];
+	  });
 	  }
 
 	  _GLIBCXX_SIMD_INTRINSIC constexpr _Ip
 	  operator|(_Ip __rhs) const
 	  {
 	return __generate_from_n_evaluations<_Np, _Ip>(
-	  [&](auto __i) { return __rhs._M_data[__i] | _M_data[__i]; });
+	  [&](auto __i) _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA {
+		return __rhs._M_data[__i] | _M_data[__i];
+	  });
 	  }
 
 	  _GLIBCXX_SIMD_INTRINSIC constexpr _Ip
 	  operator^(_Ip __rhs) const
 	  {
 	return __generate_from_n_evaluations<_Np, _Ip>(
-	  [&](auto __i) { return __rhs._M_data[__i] ^ _M_data[__i]; });
+	  [&](auto __i) _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA {
+		return __rhs._M_data[__i] ^ _M_data[__i];
+	  });
 	  }
 
 	  _GLIBCXX_SIMD_INTRINSIC constexpr _Ip
 	  operator~() const
 	  {
 	return __generate_from_n_evaluations<_Np, _Ip>(
-	  [&](auto __i) { return ~_M_data[__i]; });
+	  [&](auto __i) _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA { return ~_M_data[__i]; });
 	  }
 	};
 	return _Ip{};
@@ -1391,7 +1397,7 @@ template 
 operator^=(const _BitMask& __b) & noexcept
 {
   __execute_n_times<_S_array_size>(
-	[&](auto __i) { _M_bits[__i] ^= __b._M_bits[__i]; });
+	[&](auto __i) _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA { _M_bits[__i] ^= __b._M_bits[__i]; });
   return *this;
 }
 
@@ -1399,7 +1405,7 @@ template 
 operator|=(const _BitMask& __b) & noexcept
 {
   __execute_n_times<_S_array_size>(
-	[&](auto __i) { _M_bits[__i] |= __b._M_bits[__i]; });
+	[&](auto __i) _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA { _M_bits[__i] |= __b._M_bits[__i]; });
   return *this;
 }
 
@@ -1407,7 +1413,7 @@ template 
 operator&=(const _BitMask& __b) & noexcept
 {
   __execute_n_times<_S_array_size>(
-	[&](auto __i) { _M_bits[__i] &= __b._M_bits[__i]; });
+	[&](auto __i) _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA { _M_bits[__i] &= __b._M_bits[__i]; });
   return *this;
 }
 
@@ -1797,8 +1803,9 @@ template 
   __vector_broadcast(_Tp __x)
   {
 return __call_with_n_evaluations<_Np>(
-  [](auto... __xx) { return __vector_type_t<_Tp, _Np>{__xx...}; },
-  [&__x](int) { return __x; });
+  [](auto... __xx) _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA {
+	return __vector_type_t<_Tp, _Np>{__xx...};
+  }, [&__x](int) _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA { return __x; });
   }
 
 // }}}
@@ -2205,7 +2212,7 @@ template (
-	  __x, [](auto... __entries) {
+	  __x, [](auto... __entries) _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA {
 	return reinterpret_cast<_R>(_Up{__entries...});
 	  });
   }
@@ -2607,7 +2614,7 @@ template 
 
 _GLIBCXX_SIMD_INTRINSIC constexpr _SimdWrapper(initializer_list<_Tp> __init)
   : _Base(__generate_from_n_evaluations<_Width, _BuiltinType>(
-	[&](auto __i) { return