* [PATCH 0/8] std::experimental::simd patchset
@ 2023-02-23 8:48 Matthias Kretz
2023-02-23 8:49 ` [PATCH 1/8] libstdc++: Simplify three helper functions into one Matthias Kretz
` (7 more replies)
0 siblings, 8 replies; 19+ messages in thread
From: Matthias Kretz @ 2023-02-23 8:48 UTC (permalink / raw)
To: gcc-patches, libstdc++
Tested on x86_64-pc-linux.
This patchset provides the final changes for PR108030 and resolves
PR108856. The latter is a pure optimization and could wait for Stage 1 (I'm
submitting the patch because simd is experimental/TS)
Matthias Kretz (8):
libstdc++: Simplify three helper functions into one
libstdc++: Fix simd build failure on clang
libstdc++: More efficient masked inc-/decrement implementation
libstdc++: Add missing constexpr on simd shift implementation
libstdc++: Always-inline most of non-cmath fixed_size implementation
libstdc++: Fix formatting
libstdc++: Fix -Wsign-compare issue
libstdc++: Test that integral simd reductions are precise
libstdc++-v3/include/experimental/bits/simd.h | 485 ++++++------
.../include/experimental/bits/simd_builtin.h | 721 +++++++++---------
.../include/experimental/bits/simd_detail.h | 3 +-
.../experimental/bits/simd_fixed_size.h | 286 ++++---
.../include/experimental/bits/simd_neon.h | 24 +-
.../include/experimental/bits/simd_ppc.h | 3 +-
.../include/experimental/bits/simd_scalar.h | 362 +++++----
.../include/experimental/bits/simd_x86.h | 158 ++--
.../experimental/simd/tests/reductions.cc | 3 +-
9 files changed, 1075 insertions(+), 970 deletions(-)
--
──────────────────────────────────────────────────────────────────────────
Dr. Matthias Kretz https://mattkretz.github.io
GSI Helmholtz Centre for Heavy Ion Research https://gsi.de
stdₓ::simd
──────────────────────────────────────────────────────────────────────────
^ permalink raw reply [flat|nested] 19+ messages in thread
* [PATCH 1/8] libstdc++: Simplify three helper functions into one
2023-02-23 8:48 [PATCH 0/8] std::experimental::simd patchset Matthias Kretz
@ 2023-02-23 8:49 ` Matthias Kretz
2023-02-23 11:05 ` Jonathan Wakely
2023-02-23 8:49 ` [PATCH 2/8] libstdc++: Fix simd build failure on clang Matthias Kretz
` (6 subsequent siblings)
7 siblings, 1 reply; 19+ messages in thread
From: Matthias Kretz @ 2023-02-23 8:49 UTC (permalink / raw)
To: gcc-patches, libstdc++
[-- Attachment #1: Type: text/plain, Size: 1085 bytes --]
Broadcast is a very common function. This should reduce compile-time
effort.
Signed-off-by: Matthias Kretz <m.kretz@gsi.de>
libstdc++-v3/ChangeLog:
PR libstdc++/108030
* include/experimental/bits/simd.h (__vector_broadcast):
Implement via __vector_broadcast_impl instead of
__call_with_n_evaluations + 2 lambdas.
(__vector_broadcast_impl): New.
---
libstdc++-v3/include/experimental/bits/simd.h | 12 ++++++------
1 file changed, 6 insertions(+), 6 deletions(-)
--
──────────────────────────────────────────────────────────────────────────
Dr. Matthias Kretz https://mattkretz.github.io
GSI Helmholtz Centre for Heavy Ion Research https://gsi.de
stdₓ::simd
──────────────────────────────────────────────────────────────────────────
[-- Attachment #2: 0001-libstdc-Simplify-three-helper-functions-into-one.patch --]
[-- Type: text/x-patch, Size: 1049 bytes --]
diff --git a/libstdc++-v3/include/experimental/bits/simd.h b/libstdc++-v3/include/experimental/bits/simd.h
index 2f615d13b73..7482d109291 100644
--- a/libstdc++-v3/include/experimental/bits/simd.h
+++ b/libstdc++-v3/include/experimental/bits/simd.h
@@ -1798,15 +1798,15 @@ __to_intrin(_Tp __x)
// }}}
// __vector_broadcast{{{
+template <size_t _Np, typename _Tp, size_t... _I>
+ _GLIBCXX_SIMD_INTRINSIC constexpr __vector_type_t<_Tp, _Np>
+ __vector_broadcast_impl(_Tp __x, index_sequence<_I...>)
+ { return __vector_type_t<_Tp, _Np>{((void)_I, __x)...}; }
+
template <size_t _Np, typename _Tp>
_GLIBCXX_SIMD_INTRINSIC constexpr __vector_type_t<_Tp, _Np>
__vector_broadcast(_Tp __x)
- {
- return __call_with_n_evaluations<_Np>(
- [](auto... __xx) _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA {
- return __vector_type_t<_Tp, _Np>{__xx...};
- }, [&__x](int) _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA { return __x; });
- }
+ { return __vector_broadcast_impl<_Np, _Tp>(__x, make_index_sequence<_Np>()); }
// }}}
// __generate_vector{{{
^ permalink raw reply [flat|nested] 19+ messages in thread
* [PATCH 2/8] libstdc++: Fix simd build failure on clang
2023-02-23 8:48 [PATCH 0/8] std::experimental::simd patchset Matthias Kretz
2023-02-23 8:49 ` [PATCH 1/8] libstdc++: Simplify three helper functions into one Matthias Kretz
@ 2023-02-23 8:49 ` Matthias Kretz
2023-02-23 11:06 ` Jonathan Wakely
2023-02-23 8:49 ` [PATCH 3/8] libstdc++: More efficient masked inc-/decrement implementation Matthias Kretz
` (5 subsequent siblings)
7 siblings, 1 reply; 19+ messages in thread
From: Matthias Kretz @ 2023-02-23 8:49 UTC (permalink / raw)
To: gcc-patches, libstdc++
[-- Attachment #1: Type: text/plain, Size: 1070 bytes --]
Clang does not support __attribute__ on lambdas. Therefore, only set
_GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA if __clang__ is not defined.
Signed-off-by: Matthias Kretz <m.kretz@gsi.de>
libstdc++-v3/ChangeLog:
PR libstdc++/108030
* include/experimental/bits/simd_detail.h
(_GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA): Define as empty for
__clang__.
---
libstdc++-v3/include/experimental/bits/simd_detail.h | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
--
──────────────────────────────────────────────────────────────────────────
Dr. Matthias Kretz https://mattkretz.github.io
GSI Helmholtz Centre for Heavy Ion Research https://gsi.de
stdₓ::simd
──────────────────────────────────────────────────────────────────────────
[-- Attachment #2: 0002-libstdc-Fix-simd-build-failure-on-clang.patch --]
[-- Type: text/x-patch, Size: 1156 bytes --]
diff --git a/libstdc++-v3/include/experimental/bits/simd_detail.h b/libstdc++-v3/include/experimental/bits/simd_detail.h
index a0ad10efe0f..30cc1ef0eef 100644
--- a/libstdc++-v3/include/experimental/bits/simd_detail.h
+++ b/libstdc++-v3/include/experimental/bits/simd_detail.h
@@ -254,15 +254,16 @@ namespace experimental
#ifdef __clang__
#define _GLIBCXX_SIMD_NORMAL_MATH
+#define _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA
#else
#define _GLIBCXX_SIMD_NORMAL_MATH \
[[__gnu__::__optimize__("finite-math-only,no-signed-zeros")]]
+#define _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA __attribute__((__always_inline__))
#endif
#define _GLIBCXX_SIMD_NEVER_INLINE [[__gnu__::__noinline__]]
#define _GLIBCXX_SIMD_INTRINSIC \
[[__gnu__::__always_inline__, __gnu__::__artificial__]] inline
#define _GLIBCXX_SIMD_ALWAYS_INLINE [[__gnu__::__always_inline__]] inline
-#define _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA __attribute__((__always_inline__))
#define _GLIBCXX_SIMD_IS_UNLIKELY(__x) __builtin_expect(__x, 0)
#define _GLIBCXX_SIMD_IS_LIKELY(__x) __builtin_expect(__x, 1)
^ permalink raw reply [flat|nested] 19+ messages in thread
* [PATCH 3/8] libstdc++: More efficient masked inc-/decrement implementation
2023-02-23 8:48 [PATCH 0/8] std::experimental::simd patchset Matthias Kretz
2023-02-23 8:49 ` [PATCH 1/8] libstdc++: Simplify three helper functions into one Matthias Kretz
2023-02-23 8:49 ` [PATCH 2/8] libstdc++: Fix simd build failure on clang Matthias Kretz
@ 2023-02-23 8:49 ` Matthias Kretz
2023-02-24 17:12 ` Jonathan Wakely
2023-02-23 8:49 ` [PATCH 4/8] libstdc++: Add missing constexpr on simd shift implementation Matthias Kretz
` (4 subsequent siblings)
7 siblings, 1 reply; 19+ messages in thread
From: Matthias Kretz @ 2023-02-23 8:49 UTC (permalink / raw)
To: gcc-patches, libstdc++
[-- Attachment #1: Type: text/plain, Size: 1216 bytes --]
Signed-off-by: Matthias Kretz <m.kretz@gsi.de>
libstdc++-v3/ChangeLog:
PR libstdc++/108856
* include/experimental/bits/simd_builtin.h
(_SimdImplBuiltin::_S_masked_unary): More efficient
implementation of masked inc-/decrement for integers and floats
without AVX2.
* include/experimental/bits/simd_x86.h
(_SimdImplX86::_S_masked_unary): New. Use AVX512 masked subtract
builtins for masked inc-/decrement.
---
.../include/experimental/bits/simd_builtin.h | 27 +++++++-
.../include/experimental/bits/simd_x86.h | 68 +++++++++++++++++++
2 files changed, 93 insertions(+), 2 deletions(-)
--
──────────────────────────────────────────────────────────────────────────
Dr. Matthias Kretz https://mattkretz.github.io
GSI Helmholtz Centre for Heavy Ion Research https://gsi.de
stdₓ::simd
──────────────────────────────────────────────────────────────────────────
[-- Attachment #2: 0003-libstdc-More-efficient-masked-inc-decrement-implemen.patch --]
[-- Type: text/x-patch, Size: 4989 bytes --]
diff --git a/libstdc++-v3/include/experimental/bits/simd_builtin.h b/libstdc++-v3/include/experimental/bits/simd_builtin.h
index 792439a81bf..4a4de4534f3 100644
--- a/libstdc++-v3/include/experimental/bits/simd_builtin.h
+++ b/libstdc++-v3/include/experimental/bits/simd_builtin.h
@@ -2546,8 +2546,31 @@ _S_masked_unary(const _SimdWrapper<_K, _Np> __k,
_Op<decltype(__vv)> __op;
if (__k._M_is_constprop_all_of())
return __data(__op(__vv));
- else
- return _CommonImpl::_S_blend(__k, __v, __data(__op(__vv)));
+ else if constexpr (is_same_v<_Op<void>, __increment<void>>)
+ {
+ static_assert(not std::is_same_v<_K, bool>);
+ if constexpr (is_integral_v<_Tp>)
+ // Take a shortcut knowing that __k is an integer vector with values -1 or 0.
+ return __v._M_data - __vector_bitcast<_Tp>(__k._M_data);
+ else if constexpr (not __have_avx2)
+ return __v._M_data
+ + __vector_bitcast<_Tp>(__k._M_data & __builtin_bit_cast(
+ _K, _Tp(1)));
+ // starting with AVX2 it is more efficient to blend after add
+ }
+ else if constexpr (is_same_v<_Op<void>, __decrement<void>>)
+ {
+ static_assert(not std::is_same_v<_K, bool>);
+ if constexpr (is_integral_v<_Tp>)
+ // Take a shortcut knowing that __k is an integer vector with values -1 or 0.
+ return __v._M_data + __vector_bitcast<_Tp>(__k._M_data);
+ else if constexpr (not __have_avx2)
+ return __v._M_data
+ - __vector_bitcast<_Tp>(__k._M_data & __builtin_bit_cast(
+ _K, _Tp(1)));
+ // starting with AVX2 it is more efficient to blend after sub
+ }
+ return _CommonImpl::_S_blend(__k, __v, __data(__op(__vv)));
}
//}}}2
diff --git a/libstdc++-v3/include/experimental/bits/simd_x86.h b/libstdc++-v3/include/experimental/bits/simd_x86.h
index dcfdc2a9496..897a67829d1 100644
--- a/libstdc++-v3/include/experimental/bits/simd_x86.h
+++ b/libstdc++-v3/include/experimental/bits/simd_x86.h
@@ -3462,6 +3462,74 @@ _S_islessgreater(_SimdWrapper<_Tp, _Np> __x, _SimdWrapper<_Tp, _Np> __y)
}
//}}} }}}
+ template <template <typename> class _Op, typename _Tp, typename _K,
+ size_t _Np>
+ _GLIBCXX_SIMD_INTRINSIC static _SimdWrapper<_Tp, _Np>
+ _S_masked_unary(const _SimdWrapper<_K, _Np> __k,
+ const _SimdWrapper<_Tp, _Np> __v)
+ {
+ if (__k._M_is_constprop_none_of())
+ return __v;
+ else if (__k._M_is_constprop_all_of())
+ {
+ auto __vv = _Base::_M_make_simd(__v);
+ _Op<decltype(__vv)> __op;
+ return __data(__op(__vv));
+ }
+ else if constexpr (__is_bitmask_v<decltype(__k)>
+ && (is_same_v<_Op<void>, __increment<void>>
+ || is_same_v<_Op<void>, __decrement<void>>))
+ {
+ // optimize masked unary increment and decrement as masked sub +/-1
+ constexpr int __pm_one
+ = is_same_v<_Op<void>, __increment<void>> ? -1 : 1;
+ if constexpr (is_integral_v<_Tp>)
+ {
+ constexpr bool __lp64 = sizeof(long) == sizeof(long long);
+ using _Ip = std::make_signed_t<_Tp>;
+ using _Up = std::conditional_t<
+ std::is_same_v<_Ip, long>,
+ std::conditional_t<__lp64, long long, int>,
+ std::conditional_t<
+ std::is_same_v<_Ip, signed char>, char, _Ip>>;
+ const auto __value = __vector_bitcast<_Up>(__v._M_data);
+#define _GLIBCXX_SIMD_MASK_SUB(_Sizeof, _Width, _Instr) \
+ if constexpr (sizeof(_Tp) == _Sizeof && sizeof(__v) == _Width) \
+ return __vector_bitcast<_Tp>(__builtin_ia32_##_Instr##_mask(__value, \
+ __vector_broadcast<_Np>(_Up(__pm_one)), __value, __k._M_data))
+ _GLIBCXX_SIMD_MASK_SUB(1, 64, psubb512);
+ _GLIBCXX_SIMD_MASK_SUB(1, 32, psubb256);
+ _GLIBCXX_SIMD_MASK_SUB(1, 16, psubb128);
+ _GLIBCXX_SIMD_MASK_SUB(2, 64, psubw512);
+ _GLIBCXX_SIMD_MASK_SUB(2, 32, psubw256);
+ _GLIBCXX_SIMD_MASK_SUB(2, 16, psubw128);
+ _GLIBCXX_SIMD_MASK_SUB(4, 64, psubd512);
+ _GLIBCXX_SIMD_MASK_SUB(4, 32, psubd256);
+ _GLIBCXX_SIMD_MASK_SUB(4, 16, psubd128);
+ _GLIBCXX_SIMD_MASK_SUB(8, 64, psubq512);
+ _GLIBCXX_SIMD_MASK_SUB(8, 32, psubq256);
+ _GLIBCXX_SIMD_MASK_SUB(8, 16, psubq128);
+#undef _GLIBCXX_SIMD_MASK_SUB
+ }
+ else
+ {
+#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, _MM_FROUND_CUR_DIRECTION)
+ _GLIBCXX_SIMD_MASK_SUB(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(8, 32, subpd256);
+ _GLIBCXX_SIMD_MASK_SUB(8, 16, subpd128);
+#undef _GLIBCXX_SIMD_MASK_SUB
+ }
+ }
+ else
+ return _Base::template _S_masked_unary<_Op>(__k, __v);
+ }
};
// }}}
^ permalink raw reply [flat|nested] 19+ messages in thread
* [PATCH 4/8] libstdc++: Add missing constexpr on simd shift implementation
2023-02-23 8:48 [PATCH 0/8] std::experimental::simd patchset Matthias Kretz
` (2 preceding siblings ...)
2023-02-23 8:49 ` [PATCH 3/8] libstdc++: More efficient masked inc-/decrement implementation Matthias Kretz
@ 2023-02-23 8:49 ` Matthias Kretz
2023-02-23 11:07 ` Jonathan Wakely
2023-02-23 8:49 ` [PATCH 5/8] libstdc++: Always-inline most of non-cmath fixed_size implementation Matthias Kretz
` (3 subsequent siblings)
7 siblings, 1 reply; 19+ messages in thread
From: Matthias Kretz @ 2023-02-23 8:49 UTC (permalink / raw)
To: gcc-patches, libstdc++
[-- Attachment #1: Type: text/plain, Size: 1102 bytes --]
Resolves -Wtautological-compare warnings about `if
(__builtin_is_constant_evaluated())` in the implementations of these
functions.
Signed-off-by: Matthias Kretz <m.kretz@gsi.de>
libstdc++-v3/ChangeLog:
* include/experimental/bits/simd_x86.h (_S_bit_shift_left)
(_S_bit_shift_right): Declare constexpr. The implementation was
already expecting constexpr evaluation.
---
libstdc++-v3/include/experimental/bits/simd_x86.h | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
--
──────────────────────────────────────────────────────────────────────────
Dr. Matthias Kretz https://mattkretz.github.io
GSI Helmholtz Centre for Heavy Ion Research https://gsi.de
stdₓ::simd
──────────────────────────────────────────────────────────────────────────
[-- Attachment #2: 0004-libstdc-Add-missing-constexpr-on-simd-shift-implemen.patch --]
[-- Type: text/x-patch, Size: 1763 bytes --]
diff --git a/libstdc++-v3/include/experimental/bits/simd_x86.h b/libstdc++-v3/include/experimental/bits/simd_x86.h
index 897a67829d1..8872ca301b9 100644
--- a/libstdc++-v3/include/experimental/bits/simd_x86.h
+++ b/libstdc++-v3/include/experimental/bits/simd_x86.h
@@ -1526,7 +1526,7 @@ _S_modulus(_SimdWrapper<_Tp, _Np> __x, _SimdWrapper<_Tp, _Np> __y)
// values.
#ifndef _GLIBCXX_SIMD_NO_SHIFT_OPT
template <typename _Tp, typename _TVT = _VectorTraits<_Tp>>
- inline _GLIBCXX_CONST static typename _TVT::type
+ constexpr inline _GLIBCXX_CONST static typename _TVT::type
_S_bit_shift_left(_Tp __xx, int __y)
{
using _V = typename _TVT::type;
@@ -1631,7 +1631,7 @@ unsigned(
}
template <typename _Tp, typename _TVT = _VectorTraits<_Tp>>
- inline _GLIBCXX_CONST static typename _TVT::type
+ constexpr inline _GLIBCXX_CONST static typename _TVT::type
_S_bit_shift_left(_Tp __xx, typename _TVT::type __y)
{
using _V = typename _TVT::type;
@@ -1800,7 +1800,7 @@ _mm512_cvtepi16_epi8(
// _S_bit_shift_right {{{
#ifndef _GLIBCXX_SIMD_NO_SHIFT_OPT
template <typename _Tp, typename _TVT = _VectorTraits<_Tp>>
- inline _GLIBCXX_CONST static typename _TVT::type
+ constexpr inline _GLIBCXX_CONST static typename _TVT::type
_S_bit_shift_right(_Tp __xx, int __y)
{
using _V = typename _TVT::type;
@@ -1850,7 +1850,7 @@ _S_bit_shift_right(_Tp __xx, int __y)
}
template <typename _Tp, typename _TVT = _VectorTraits<_Tp>>
- inline _GLIBCXX_CONST static typename _TVT::type
+ constexpr inline _GLIBCXX_CONST static typename _TVT::type
_S_bit_shift_right(_Tp __xx, typename _TVT::type __y)
{
using _V = typename _TVT::type;
^ permalink raw reply [flat|nested] 19+ messages in thread
* [PATCH 5/8] libstdc++: Always-inline most of non-cmath fixed_size implementation
2023-02-23 8:48 [PATCH 0/8] std::experimental::simd patchset Matthias Kretz
` (3 preceding siblings ...)
2023-02-23 8:49 ` [PATCH 4/8] libstdc++: Add missing constexpr on simd shift implementation Matthias Kretz
@ 2023-02-23 8:49 ` Matthias Kretz
2023-02-24 17:10 ` Jonathan Wakely
2023-02-23 8:50 ` [PATCH 6/8] libstdc++: Fix formatting Matthias Kretz
` (2 subsequent siblings)
7 siblings, 1 reply; 19+ messages in thread
From: Matthias Kretz @ 2023-02-23 8:49 UTC (permalink / raw)
To: gcc-patches, libstdc++
[-- Attachment #1: Type: text/plain, Size: 2375 bytes --]
For simd, the inlining behavior should be similar to builtin types. (No
operator on buitin types is ever translated into a function call.)
Therefore, always_inline is the right choice (i.e. inline on -O0 as
well).
Signed-off-by: Matthias Kretz <m.kretz@gsi.de>
libstdc++-v3/ChangeLog:
PR libstdc++/108030
* include/experimental/bits/simd_fixed_size.h
(_SimdImplFixedSize::_S_broadcast): Replace inline with
_GLIBCXX_SIMD_INTRINSIC.
(_SimdImplFixedSize::_S_generate): Likewise.
(_SimdImplFixedSize::_S_load): Likewise.
(_SimdImplFixedSize::_S_masked_load): Likewise.
(_SimdImplFixedSize::_S_store): Likewise.
(_SimdImplFixedSize::_S_masked_store): Likewise.
(_SimdImplFixedSize::_S_min): Likewise.
(_SimdImplFixedSize::_S_max): Likewise.
(_SimdImplFixedSize::_S_complement): Likewise.
(_SimdImplFixedSize::_S_unary_minus): Likewise.
(_SimdImplFixedSize::_S_plus): Likewise.
(_SimdImplFixedSize::_S_minus): Likewise.
(_SimdImplFixedSize::_S_multiplies): Likewise.
(_SimdImplFixedSize::_S_divides): Likewise.
(_SimdImplFixedSize::_S_modulus): Likewise.
(_SimdImplFixedSize::_S_bit_and): Likewise.
(_SimdImplFixedSize::_S_bit_or): Likewise.
(_SimdImplFixedSize::_S_bit_xor): Likewise.
(_SimdImplFixedSize::_S_bit_shift_left): Likewise.
(_SimdImplFixedSize::_S_bit_shift_right): Likewise.
(_SimdImplFixedSize::_S_remquo): Add inline keyword (to be
explicit about not always-inline, yet).
(_SimdImplFixedSize::_S_isinf): Likewise.
(_SimdImplFixedSize::_S_isfinite): Likewise.
(_SimdImplFixedSize::_S_isnan): Likewise.
(_SimdImplFixedSize::_S_isnormal): Likewise.
(_SimdImplFixedSize::_S_signbit): Likewise.
---
.../experimental/bits/simd_fixed_size.h | 60 +++++++++----------
1 file changed, 30 insertions(+), 30 deletions(-)
--
──────────────────────────────────────────────────────────────────────────
Dr. Matthias Kretz https://mattkretz.github.io
GSI Helmholtz Centre for Heavy Ion Research https://gsi.de
stdₓ::simd
──────────────────────────────────────────────────────────────────────────
[-- Attachment #2: 0005-libstdc-Always-inline-most-of-non-cmath-fixed_size-i.patch --]
[-- Type: text/x-patch, Size: 8252 bytes --]
diff --git a/libstdc++-v3/include/experimental/bits/simd_fixed_size.h b/libstdc++-v3/include/experimental/bits/simd_fixed_size.h
index 3ac6eaa3f6b..88a9b27e359 100644
--- a/libstdc++-v3/include/experimental/bits/simd_fixed_size.h
+++ b/libstdc++-v3/include/experimental/bits/simd_fixed_size.h
@@ -1284,7 +1284,8 @@ struct _SimdImplFixedSize
// broadcast {{{2
template <typename _Tp>
- static constexpr inline _SimdMember<_Tp> _S_broadcast(_Tp __x) noexcept
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _SimdMember<_Tp>
+ _S_broadcast(_Tp __x) noexcept
{
return _SimdMember<_Tp>::_S_generate(
[&](auto __meta) constexpr _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA {
@@ -1294,8 +1295,8 @@ struct _SimdImplFixedSize
// _S_generator {{{2
template <typename _Fp, typename _Tp>
- static constexpr inline _SimdMember<_Tp> _S_generator(_Fp&& __gen,
- _TypeTag<_Tp>)
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _SimdMember<_Tp>
+ _S_generator(_Fp&& __gen, _TypeTag<_Tp>)
{
return _SimdMember<_Tp>::_S_generate(
[&__gen](auto __meta) constexpr _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA {
@@ -1310,8 +1311,8 @@ struct _SimdImplFixedSize
// _S_load {{{2
template <typename _Tp, typename _Up>
- static inline _SimdMember<_Tp> _S_load(const _Up* __mem,
- _TypeTag<_Tp>) noexcept
+ _GLIBCXX_SIMD_INTRINSIC static _SimdMember<_Tp>
+ _S_load(const _Up* __mem, _TypeTag<_Tp>) noexcept
{
return _SimdMember<_Tp>::_S_generate(
[&](auto __meta) _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA {
@@ -1321,7 +1322,7 @@ struct _SimdImplFixedSize
// _S_masked_load {{{2
template <typename _Tp, typename... _As, typename _Up>
- static inline _SimdTuple<_Tp, _As...>
+ _GLIBCXX_SIMD_INTRINSIC static _SimdTuple<_Tp, _As...>
_S_masked_load(const _SimdTuple<_Tp, _As...>& __old,
const _MaskMember __bits, const _Up* __mem) noexcept
{
@@ -1344,8 +1345,8 @@ _S_masked_load(const _SimdTuple<_Tp, _As...>& __old,
// _S_store {{{2
template <typename _Tp, typename _Up>
- static inline void _S_store(const _SimdMember<_Tp>& __v, _Up* __mem,
- _TypeTag<_Tp>) noexcept
+ _GLIBCXX_SIMD_INTRINSIC static void
+ _S_store(const _SimdMember<_Tp>& __v, _Up* __mem, _TypeTag<_Tp>) noexcept
{
__for_each(__v, [&](auto __meta, auto __native) _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA {
__meta._S_store(__native, &__mem[__meta._S_offset], _TypeTag<_Tp>());
@@ -1354,9 +1355,9 @@ _S_masked_load(const _SimdTuple<_Tp, _As...>& __old,
// _S_masked_store {{{2
template <typename _Tp, typename... _As, typename _Up>
- static inline void _S_masked_store(const _SimdTuple<_Tp, _As...>& __v,
- _Up* __mem,
- const _MaskMember __bits) noexcept
+ _GLIBCXX_SIMD_INTRINSIC static void
+ _S_masked_store(const _SimdTuple<_Tp, _As...>& __v, _Up* __mem,
+ const _MaskMember __bits) noexcept
{
__for_each(__v, [&](auto __meta, auto __native) _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA {
if (__meta._S_submask(__bits).any())
@@ -1464,7 +1465,7 @@ __for_each(
// _S_min, _S_max {{{2
template <typename _Tp, typename... _As>
- static inline constexpr _SimdTuple<_Tp, _As...>
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _SimdTuple<_Tp, _As...>
_S_min(const _SimdTuple<_Tp, _As...>& __a,
const _SimdTuple<_Tp, _As...>& __b)
{
@@ -1476,7 +1477,7 @@ _S_min(const _SimdTuple<_Tp, _As...>& __a,
}
template <typename _Tp, typename... _As>
- static inline constexpr _SimdTuple<_Tp, _As...>
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _SimdTuple<_Tp, _As...>
_S_max(const _SimdTuple<_Tp, _As...>& __a,
const _SimdTuple<_Tp, _As...>& __b)
{
@@ -1489,7 +1490,7 @@ _S_max(const _SimdTuple<_Tp, _As...>& __a,
// _S_complement {{{2
template <typename _Tp, typename... _As>
- static inline constexpr _SimdTuple<_Tp, _As...>
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _SimdTuple<_Tp, _As...>
_S_complement(const _SimdTuple<_Tp, _As...>& __x) noexcept
{
return __x._M_apply_per_chunk(
@@ -1500,7 +1501,7 @@ _S_complement(const _SimdTuple<_Tp, _As...>& __x) noexcept
// _S_unary_minus {{{2
template <typename _Tp, typename... _As>
- static inline constexpr _SimdTuple<_Tp, _As...>
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _SimdTuple<_Tp, _As...>
_S_unary_minus(const _SimdTuple<_Tp, _As...>& __x) noexcept
{
return __x._M_apply_per_chunk(
@@ -1513,7 +1514,7 @@ _S_unary_minus(const _SimdTuple<_Tp, _As...>& __x) noexcept
#define _GLIBCXX_SIMD_FIXED_OP(name_, op_) \
template <typename _Tp, typename... _As> \
- static inline constexpr _SimdTuple<_Tp, _As...> name_( \
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _SimdTuple<_Tp, _As...> name_( \
const _SimdTuple<_Tp, _As...>& __x, const _SimdTuple<_Tp, _As...>& __y) \
{ \
return __x._M_apply_per_chunk( \
@@ -1536,7 +1537,7 @@ _S_unary_minus(const _SimdTuple<_Tp, _As...>& __x) noexcept
#undef _GLIBCXX_SIMD_FIXED_OP
template <typename _Tp, typename... _As>
- static inline constexpr _SimdTuple<_Tp, _As...>
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _SimdTuple<_Tp, _As...>
_S_bit_shift_left(const _SimdTuple<_Tp, _As...>& __x, int __y)
{
return __x._M_apply_per_chunk(
@@ -1546,7 +1547,7 @@ _S_bit_shift_left(const _SimdTuple<_Tp, _As...>& __x, int __y)
}
template <typename _Tp, typename... _As>
- static inline constexpr _SimdTuple<_Tp, _As...>
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _SimdTuple<_Tp, _As...>
_S_bit_shift_right(const _SimdTuple<_Tp, _As...>& __x, int __y)
{
return __x._M_apply_per_chunk(
@@ -1665,10 +1666,9 @@ _S_bit_shift_right(const _SimdTuple<_Tp, _As...>& __x, int __y)
#undef _GLIBCXX_SIMD_APPLY_ON_TUPLE
template <typename _Tp, typename... _Abis>
- static _SimdTuple<_Tp, _Abis...> _S_remquo(
- const _SimdTuple<_Tp, _Abis...>& __x,
- const _SimdTuple<_Tp, _Abis...>& __y,
- __fixed_size_storage_t<int, _SimdTuple<_Tp, _Abis...>::_S_size()>* __z)
+ static inline _SimdTuple<_Tp, _Abis...>
+ _S_remquo(const _SimdTuple<_Tp, _Abis...>& __x, const _SimdTuple<_Tp, _Abis...>& __y,
+ __fixed_size_storage_t<int, _SimdTuple<_Tp, _Abis...>::_S_size()>* __z)
{
return __x._M_apply_per_chunk(
[](auto __impl, const auto __xx, const auto __yy, auto& __zz)
@@ -1689,14 +1689,14 @@ _S_frexp(const _SimdTuple<_Tp, _As...>& __x,
}, __exp);
}
-#define _GLIBCXX_SIMD_TEST_ON_TUPLE_(name_) \
- template <typename _Tp, typename... _As> \
- static inline _MaskMember \
- _S_##name_(const _SimdTuple<_Tp, _As...>& __x) noexcept \
- { \
- return _M_test([](auto __impl, \
- auto __xx) { return __impl._S_##name_(__xx); }, \
- __x); \
+#define _GLIBCXX_SIMD_TEST_ON_TUPLE_(name_) \
+ template <typename _Tp, typename... _As> \
+ static inline _MaskMember \
+ _S_##name_(const _SimdTuple<_Tp, _As...>& __x) noexcept \
+ { \
+ return _M_test([] (auto __impl, auto __xx) _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA { \
+ return __impl._S_##name_(__xx); \
+ }, __x); \
}
_GLIBCXX_SIMD_TEST_ON_TUPLE_(isinf)
^ permalink raw reply [flat|nested] 19+ messages in thread
* [PATCH 6/8] libstdc++: Fix formatting
2023-02-23 8:48 [PATCH 0/8] std::experimental::simd patchset Matthias Kretz
` (4 preceding siblings ...)
2023-02-23 8:49 ` [PATCH 5/8] libstdc++: Always-inline most of non-cmath fixed_size implementation Matthias Kretz
@ 2023-02-23 8:50 ` Matthias Kretz
2023-02-24 17:14 ` Jonathan Wakely
2023-02-23 8:50 ` [PATCH 7/8] libstdc++: Fix -Wsign-compare issue Matthias Kretz
2023-02-23 8:50 ` [PATCH 8/8] libstdc++: Test that integral simd reductions are precise Matthias Kretz
7 siblings, 1 reply; 19+ messages in thread
From: Matthias Kretz @ 2023-02-23 8:50 UTC (permalink / raw)
To: gcc-patches, libstdc++
[-- Attachment #1: Type: text/plain, Size: 1626 bytes --]
Whitespace changes only.
Signed-off-by: Matthias Kretz <m.kretz@gsi.de>
libstdc++-v3/ChangeLog:
* include/experimental/bits/simd.h: Line breaks and indenting
fixed to follow the libstdc++ standard.
* include/experimental/bits/simd_builtin.h: Likewise.
* include/experimental/bits/simd_fixed_size.h: Likewise.
* include/experimental/bits/simd_neon.h: Likewise.
* include/experimental/bits/simd_ppc.h: Likewise.
* include/experimental/bits/simd_scalar.h: Likewise.
* include/experimental/bits/simd_x86.h: Likewise.
---
libstdc++-v3/include/experimental/bits/simd.h | 473 ++++++------
.../include/experimental/bits/simd_builtin.h | 692 +++++++++---------
.../experimental/bits/simd_fixed_size.h | 228 +++---
.../include/experimental/bits/simd_neon.h | 24 +-
.../include/experimental/bits/simd_ppc.h | 3 +-
.../include/experimental/bits/simd_scalar.h | 362 +++++----
.../include/experimental/bits/simd_x86.h | 90 ++-
7 files changed, 942 insertions(+), 930 deletions(-)
--
──────────────────────────────────────────────────────────────────────────
Dr. Matthias Kretz https://mattkretz.github.io
GSI Helmholtz Centre for Heavy Ion Research https://gsi.de
stdₓ::simd
──────────────────────────────────────────────────────────────────────────
[-- Attachment #2: 0006-libstdc-Fix-formatting.patch --]
[-- Type: text/x-patch, Size: 143058 bytes --]
diff --git a/libstdc++-v3/include/experimental/bits/simd.h b/libstdc++-v3/include/experimental/bits/simd.h
index 7482d109291..fb661c9657f 100644
--- a/libstdc++-v3/include/experimental/bits/simd.h
+++ b/libstdc++-v3/include/experimental/bits/simd.h
@@ -180,10 +180,7 @@ struct vector_aligned_tag
template <typename _Tp, typename _Up>
_GLIBCXX_SIMD_INTRINSIC static constexpr _Up*
_S_apply(_Up* __ptr)
- {
- return static_cast<_Up*>(
- __builtin_assume_aligned(__ptr, _S_alignment<_Tp, _Up>));
- }
+ { return static_cast<_Up*>( __builtin_assume_aligned(__ptr, _S_alignment<_Tp, _Up>)); }
};
template <size_t _Np> struct overaligned_tag
@@ -288,13 +285,15 @@ namespace __detail
// expression. math_errhandling may expand to an extern symbol, in which case a constexpr value
// must be guessed.
template <int = math_errhandling>
- constexpr bool __handle_fpexcept_impl(int)
+ constexpr bool
+ __handle_fpexcept_impl(int)
{ return math_errhandling & MATH_ERREXCEPT; }
#endif
// Fallback if math_errhandling doesn't work: with fast-math assume floating-point exceptions are
// ignored, otherwise implement correct exception behavior.
- constexpr bool __handle_fpexcept_impl(float)
+ constexpr bool
+ __handle_fpexcept_impl(float)
{
#if defined __FAST_MATH__
return false;
@@ -749,8 +748,7 @@ struct __make_dependent
// __invoke_ub{{{
template <typename... _Args>
[[noreturn]] _GLIBCXX_SIMD_ALWAYS_INLINE void
- __invoke_ub([[maybe_unused]] const char* __msg,
- [[maybe_unused]] const _Args&... __args)
+ __invoke_ub([[maybe_unused]] const char* __msg, [[maybe_unused]] const _Args&... __args)
{
#ifdef _GLIBCXX_DEBUG_UB
__builtin_fprintf(stderr, __msg, __args...);
@@ -795,11 +793,14 @@ class _ExactBool
const bool _M_data;
public:
- _GLIBCXX_SIMD_INTRINSIC constexpr _ExactBool(bool __b) : _M_data(__b) {}
+ _GLIBCXX_SIMD_INTRINSIC constexpr
+ _ExactBool(bool __b) : _M_data(__b) {}
_ExactBool(int) = delete;
- _GLIBCXX_SIMD_INTRINSIC constexpr operator bool() const { return _M_data; }
+ _GLIBCXX_SIMD_INTRINSIC constexpr
+ operator bool() const
+ { return _M_data; }
};
// }}}
@@ -1488,8 +1489,7 @@ struct __vector_type_n<_Tp, 1, enable_if_t<__is_vectorizable_v<_Tp>>>
// else, use GNU-style builtin vector types
template <typename _Tp, size_t _Np>
- struct __vector_type_n<_Tp, _Np,
- enable_if_t<__is_vectorizable_v<_Tp> && _Np >= 2>>
+ struct __vector_type_n<_Tp, _Np, enable_if_t<__is_vectorizable_v<_Tp> && _Np >= 2>>
{
static constexpr size_t _S_Np2 = std::__bit_ceil(_Np * sizeof(_Tp));
@@ -1770,8 +1770,7 @@ __bit_cast(const _From __x)
// }}}
// __to_intrin {{{
template <typename _Tp, typename _TVT = _VectorTraits<_Tp>,
- typename _R
- = __intrinsic_type_t<typename _TVT::value_type, _TVT::_S_full_size>>
+ typename _R = __intrinsic_type_t<typename _TVT::value_type, _TVT::_S_full_size>>
_GLIBCXX_SIMD_INTRINSIC constexpr _R
__to_intrin(_Tp __x)
{
@@ -1792,9 +1791,7 @@ __to_intrin(_Tp __x)
template <typename _Tp, typename... _Args>
_GLIBCXX_SIMD_INTRINSIC constexpr __vector_type_t<_Tp, sizeof...(_Args)>
__make_vector(const _Args&... __args)
- {
- return __vector_type_t<_Tp, sizeof...(_Args)>{static_cast<_Tp>(__args)...};
- }
+ { return __vector_type_t<_Tp, sizeof...(_Args)>{static_cast<_Tp>(__args)...}; }
// }}}
// __vector_broadcast{{{
@@ -1813,10 +1810,7 @@ __vector_broadcast(_Tp __x)
template <typename _Tp, size_t _Np, typename _Gp, size_t... _I>
_GLIBCXX_SIMD_INTRINSIC constexpr __vector_type_t<_Tp, _Np>
__generate_vector_impl(_Gp&& __gen, index_sequence<_I...>)
- {
- return __vector_type_t<_Tp, _Np>{
- static_cast<_Tp>(__gen(_SizeConstant<_I>()))...};
- }
+ { return __vector_type_t<_Tp, _Np>{ static_cast<_Tp>(__gen(_SizeConstant<_I>()))...}; }
template <typename _V, typename _VVT = _VectorTraits<_V>, typename _Gp>
_GLIBCXX_SIMD_INTRINSIC constexpr _V
@@ -2029,8 +2023,7 @@ __not(_Tp __a) noexcept
// }}}
// __concat{{{
template <typename _Tp, typename _TVT = _VectorTraits<_Tp>,
- typename _R = __vector_type_t<typename _TVT::value_type,
- _TVT::_S_full_size * 2>>
+ typename _R = __vector_type_t<typename _TVT::value_type, _TVT::_S_full_size * 2>>
constexpr _R
__concat(_Tp a_, _Tp b_)
{
@@ -2174,8 +2167,7 @@ __zero_extend(_Tp __x)
int _SplitBy,
typename _Tp,
typename _TVT = _VectorTraits<_Tp>,
- typename _R = __vector_type_t<typename _TVT::value_type,
- _TVT::_S_full_size / _SplitBy>>
+ typename _R = __vector_type_t<typename _TVT::value_type, _TVT::_S_full_size / _SplitBy>>
_GLIBCXX_SIMD_INTRINSIC constexpr _R
__extract(_Tp __in)
{
@@ -2221,8 +2213,7 @@ __extract(_Tp __in)
// }}}
// __lo/__hi64[z]{{{
template <typename _Tp,
- typename _R
- = __vector_type8_t<typename _VectorTraits<_Tp>::value_type>>
+ typename _R = __vector_type8_t<typename _VectorTraits<_Tp>::value_type>>
_GLIBCXX_SIMD_INTRINSIC constexpr _R
__lo64(_Tp __x)
{
@@ -2232,8 +2223,7 @@ __lo64(_Tp __x)
}
template <typename _Tp,
- typename _R
- = __vector_type8_t<typename _VectorTraits<_Tp>::value_type>>
+ typename _R = __vector_type8_t<typename _VectorTraits<_Tp>::value_type>>
_GLIBCXX_SIMD_INTRINSIC constexpr _R
__hi64(_Tp __x)
{
@@ -2244,8 +2234,7 @@ __hi64(_Tp __x)
}
template <typename _Tp,
- typename _R
- = __vector_type8_t<typename _VectorTraits<_Tp>::value_type>>
+ typename _R = __vector_type8_t<typename _VectorTraits<_Tp>::value_type>>
_GLIBCXX_SIMD_INTRINSIC constexpr _R
__hi64z([[maybe_unused]] _Tp __x)
{
@@ -2356,18 +2345,15 @@ struct __bool_storage_member_type<64>
// the following excludes bool via __is_vectorizable
#if _GLIBCXX_SIMD_HAVE_SSE
template <typename _Tp, size_t _Bytes>
- struct __intrinsic_type<_Tp, _Bytes,
- enable_if_t<__is_vectorizable_v<_Tp> && _Bytes <= 64>>
+ struct __intrinsic_type<_Tp, _Bytes, enable_if_t<__is_vectorizable_v<_Tp> && _Bytes <= 64>>
{
static_assert(!is_same_v<_Tp, long double>,
"no __intrinsic_type support for long double on x86");
- static constexpr size_t _S_VBytes = _Bytes <= 16 ? 16
- : _Bytes <= 32 ? 32
- : 64;
+ static constexpr size_t _S_VBytes = _Bytes <= 16 ? 16 : _Bytes <= 32 ? 32 : 64;
using type [[__gnu__::__vector_size__(_S_VBytes)]]
- = conditional_t<is_integral_v<_Tp>, long long int, _Tp>;
+ = conditional_t<is_integral_v<_Tp>, long long int, _Tp>;
};
#endif // _GLIBCXX_SIMD_HAVE_SSE
@@ -2413,16 +2399,19 @@ struct __intrinsic_type<make_unsigned_t<__int_with_sizeof_t<_Bits / 8>>
#undef _GLIBCXX_SIMD_ARM_INTRIN
template <typename _Tp, size_t _Bytes>
- struct __intrinsic_type<_Tp, _Bytes,
- enable_if_t<__is_vectorizable_v<_Tp> && _Bytes <= 16>>
+ struct __intrinsic_type<_Tp, _Bytes, enable_if_t<__is_vectorizable_v<_Tp> && _Bytes <= 16>>
{
static constexpr int _SVecBytes = _Bytes <= 8 ? 8 : 16;
+
using _Ip = __int_for_sizeof_t<_Tp>;
+
using _Up = conditional_t<
is_floating_point_v<_Tp>, _Tp,
conditional_t<is_unsigned_v<_Tp>, make_unsigned_t<_Ip>, _Ip>>;
+
static_assert(!is_same_v<_Tp, _Up> || _SVecBytes != _Bytes,
"should use explicit specialization above");
+
using type = typename __intrinsic_type<_Up, _SVecBytes>::type;
};
#endif // _GLIBCXX_SIMD_HAVE_NEON
@@ -2457,18 +2446,20 @@ struct __intrinsic_type_impl<_Tp>
#undef _GLIBCXX_SIMD_PPC_INTRIN
template <typename _Tp, size_t _Bytes>
- struct __intrinsic_type<_Tp, _Bytes,
- enable_if_t<__is_vectorizable_v<_Tp> && _Bytes <= 16>>
+ struct __intrinsic_type<_Tp, _Bytes, enable_if_t<__is_vectorizable_v<_Tp> && _Bytes <= 16>>
{
static constexpr bool _S_is_ldouble = is_same_v<_Tp, long double>;
+
// allow _Tp == long double with -mlong-double-64
static_assert(!(_S_is_ldouble && sizeof(long double) > sizeof(double)),
"no __intrinsic_type support for 128-bit floating point on PowerPC");
+
#ifndef __VSX__
static_assert(!(is_same_v<_Tp, double>
|| (_S_is_ldouble && sizeof(long double) == sizeof(double))),
"no __intrinsic_type support for 64-bit floating point on PowerPC w/o VSX");
#endif
+
using type =
typename __intrinsic_type_impl<
conditional_t<is_floating_point_v<_Tp>,
@@ -2489,22 +2480,29 @@ struct _SimdWrapper
static constexpr size_t _S_full_size = sizeof(_BuiltinType) * __CHAR_BIT__;
_GLIBCXX_SIMD_INTRINSIC constexpr _SimdWrapper<bool, _S_full_size>
- __as_full_vector() const { return _M_data; }
+ __as_full_vector() const
+ { return _M_data; }
+
+ _GLIBCXX_SIMD_INTRINSIC constexpr
+ _SimdWrapper() = default;
- _GLIBCXX_SIMD_INTRINSIC constexpr _SimdWrapper() = default;
- _GLIBCXX_SIMD_INTRINSIC constexpr _SimdWrapper(_BuiltinType __k)
- : _M_data(__k) {};
+ _GLIBCXX_SIMD_INTRINSIC constexpr
+ _SimdWrapper(_BuiltinType __k) : _M_data(__k) {};
- _GLIBCXX_SIMD_INTRINSIC operator const _BuiltinType&() const
+ _GLIBCXX_SIMD_INTRINSIC
+ operator const _BuiltinType&() const
{ return _M_data; }
- _GLIBCXX_SIMD_INTRINSIC operator _BuiltinType&()
+ _GLIBCXX_SIMD_INTRINSIC
+ operator _BuiltinType&()
{ return _M_data; }
- _GLIBCXX_SIMD_INTRINSIC _BuiltinType __intrin() const
+ _GLIBCXX_SIMD_INTRINSIC _BuiltinType
+ __intrin() const
{ return _M_data; }
- _GLIBCXX_SIMD_INTRINSIC constexpr value_type operator[](size_t __i) const
+ _GLIBCXX_SIMD_INTRINSIC constexpr value_type
+ operator[](size_t __i) const
{ return _M_data & (_BuiltinType(1) << __i); }
template <size_t __i>
@@ -2512,7 +2510,8 @@ struct _SimdWrapper
operator[](_SizeConstant<__i>) const
{ return _M_data & (_BuiltinType(1) << __i); }
- _GLIBCXX_SIMD_INTRINSIC constexpr void _M_set(size_t __i, value_type __x)
+ _GLIBCXX_SIMD_INTRINSIC constexpr void
+ _M_set(size_t __i, value_type __x)
{
if (__x)
_M_data |= (_BuiltinType(1) << __i);
@@ -2520,11 +2519,12 @@ struct _SimdWrapper
_M_data &= ~(_BuiltinType(1) << __i);
}
- _GLIBCXX_SIMD_INTRINSIC
- constexpr bool _M_is_constprop() const
+ _GLIBCXX_SIMD_INTRINSIC constexpr bool
+ _M_is_constprop() const
{ return __builtin_constant_p(_M_data); }
- _GLIBCXX_SIMD_INTRINSIC constexpr bool _M_is_constprop_none_of() const
+ _GLIBCXX_SIMD_INTRINSIC constexpr bool
+ _M_is_constprop_none_of() const
{
if (__builtin_constant_p(_M_data))
{
@@ -2536,7 +2536,8 @@ struct _SimdWrapper
return false;
}
- _GLIBCXX_SIMD_INTRINSIC constexpr bool _M_is_constprop_all_of() const
+ _GLIBCXX_SIMD_INTRINSIC constexpr bool
+ _M_is_constprop_all_of() const
{
if (__builtin_constant_p(_M_data))
{
@@ -2558,10 +2559,11 @@ struct _SimdWrapperBase
template <typename _BuiltinType>
struct _SimdWrapperBase<false, _BuiltinType> // no padding or no SNaNs
{
- _GLIBCXX_SIMD_INTRINSIC constexpr _SimdWrapperBase() = default;
- _GLIBCXX_SIMD_INTRINSIC constexpr _SimdWrapperBase(_BuiltinType __init)
- : _M_data(__init)
- {}
+ _GLIBCXX_SIMD_INTRINSIC constexpr
+ _SimdWrapperBase() = default;
+
+ _GLIBCXX_SIMD_INTRINSIC constexpr
+ _SimdWrapperBase(_BuiltinType __init) : _M_data(__init) {}
_BuiltinType _M_data;
};
@@ -2570,10 +2572,11 @@ struct _SimdWrapperBase<false, _BuiltinType>
struct _SimdWrapperBase<true, _BuiltinType> // with padding that needs to
// never become SNaN
{
- _GLIBCXX_SIMD_INTRINSIC constexpr _SimdWrapperBase() : _M_data() {}
- _GLIBCXX_SIMD_INTRINSIC constexpr _SimdWrapperBase(_BuiltinType __init)
- : _M_data(__init)
- {}
+ _GLIBCXX_SIMD_INTRINSIC constexpr
+ _SimdWrapperBase() : _M_data() {}
+
+ _GLIBCXX_SIMD_INTRINSIC constexpr
+ _SimdWrapperBase(_BuiltinType __init) : _M_data(__init) {}
_BuiltinType _M_data;
};
@@ -2612,24 +2615,33 @@ struct _SimdWrapper
__as_full_vector() const
{ return _M_data; }
- _GLIBCXX_SIMD_INTRINSIC constexpr _SimdWrapper(initializer_list<_Tp> __init)
- : _Base(__generate_from_n_evaluations<_Width, _BuiltinType>(
- [&](auto __i) _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA { return __init.begin()[__i.value]; })) {}
+ _GLIBCXX_SIMD_INTRINSIC constexpr
+ _SimdWrapper(initializer_list<_Tp> __init)
+ : _Base(__generate_from_n_evaluations<_Width, _BuiltinType>(
+ [&](auto __i) _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA {
+ return __init.begin()[__i.value];
+ })) {}
- _GLIBCXX_SIMD_INTRINSIC constexpr _SimdWrapper() = default;
- _GLIBCXX_SIMD_INTRINSIC constexpr _SimdWrapper(const _SimdWrapper&)
- = default;
- _GLIBCXX_SIMD_INTRINSIC constexpr _SimdWrapper(_SimdWrapper&&) = default;
+ _GLIBCXX_SIMD_INTRINSIC constexpr
+ _SimdWrapper() = default;
+
+ _GLIBCXX_SIMD_INTRINSIC constexpr
+ _SimdWrapper(const _SimdWrapper&) = default;
+
+ _GLIBCXX_SIMD_INTRINSIC constexpr
+ _SimdWrapper(_SimdWrapper&&) = default;
_GLIBCXX_SIMD_INTRINSIC constexpr _SimdWrapper&
operator=(const _SimdWrapper&) = default;
+
_GLIBCXX_SIMD_INTRINSIC constexpr _SimdWrapper&
operator=(_SimdWrapper&&) = default;
template <typename _V, typename = enable_if_t<disjunction_v<
is_same<_V, __vector_type_t<_Tp, _Width>>,
is_same<_V, __intrinsic_type_t<_Tp, _Width>>>>>
- _GLIBCXX_SIMD_INTRINSIC constexpr _SimdWrapper(_V __x)
+ _GLIBCXX_SIMD_INTRINSIC constexpr
+ _SimdWrapper(_V __x)
// __vector_bitcast can convert e.g. __m128 to __vector(2) float
: _Base(__vector_bitcast<_Tp, _Width>(__x)) {}
@@ -2644,27 +2656,34 @@ __as_full_vector() const
{ return _M_data[int(__i)]; });
}
- _GLIBCXX_SIMD_INTRINSIC constexpr operator const _BuiltinType&() const
+ _GLIBCXX_SIMD_INTRINSIC constexpr
+ operator const _BuiltinType&() const
{ return _M_data; }
- _GLIBCXX_SIMD_INTRINSIC constexpr operator _BuiltinType&()
+ _GLIBCXX_SIMD_INTRINSIC constexpr
+ operator _BuiltinType&()
{ return _M_data; }
- _GLIBCXX_SIMD_INTRINSIC constexpr _Tp operator[](size_t __i) const
+ _GLIBCXX_SIMD_INTRINSIC constexpr _Tp
+ operator[](size_t __i) const
{ return _M_data[__i]; }
template <size_t __i>
- _GLIBCXX_SIMD_INTRINSIC constexpr _Tp operator[](_SizeConstant<__i>) const
+ _GLIBCXX_SIMD_INTRINSIC constexpr _Tp
+ operator[](_SizeConstant<__i>) const
{ return _M_data[__i]; }
- _GLIBCXX_SIMD_INTRINSIC constexpr void _M_set(size_t __i, _Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC constexpr void
+ _M_set(size_t __i, _Tp __x)
{ _M_data[__i] = __x; }
_GLIBCXX_SIMD_INTRINSIC
- constexpr bool _M_is_constprop() const
+ constexpr bool
+ _M_is_constprop() const
{ return __builtin_constant_p(_M_data); }
- _GLIBCXX_SIMD_INTRINSIC constexpr bool _M_is_constprop_none_of() const
+ _GLIBCXX_SIMD_INTRINSIC constexpr bool
+ _M_is_constprop_none_of() const
{
if (__builtin_constant_p(_M_data))
{
@@ -2685,7 +2704,8 @@ __as_full_vector() const
return false;
}
- _GLIBCXX_SIMD_INTRINSIC constexpr bool _M_is_constprop_all_of() const
+ _GLIBCXX_SIMD_INTRINSIC constexpr bool
+ _M_is_constprop_all_of() const
{
if (__builtin_constant_p(_M_data))
{
@@ -2883,22 +2903,14 @@ struct deduce
struct rebind_simd;
template <typename _Tp, typename _Up, typename _Abi>
- struct rebind_simd<
- _Tp, simd<_Up, _Abi>,
- void_t<simd_abi::deduce_t<_Tp, simd_size_v<_Up, _Abi>, _Abi>>>
- {
- using type
- = simd<_Tp, simd_abi::deduce_t<_Tp, simd_size_v<_Up, _Abi>, _Abi>>;
- };
+ struct rebind_simd<_Tp, simd<_Up, _Abi>,
+ void_t<simd_abi::deduce_t<_Tp, simd_size_v<_Up, _Abi>, _Abi>>>
+ { using type = simd<_Tp, simd_abi::deduce_t<_Tp, simd_size_v<_Up, _Abi>, _Abi>>; };
template <typename _Tp, typename _Up, typename _Abi>
- struct rebind_simd<
- _Tp, simd_mask<_Up, _Abi>,
- void_t<simd_abi::deduce_t<_Tp, simd_size_v<_Up, _Abi>, _Abi>>>
- {
- using type
- = simd_mask<_Tp, simd_abi::deduce_t<_Tp, simd_size_v<_Up, _Abi>, _Abi>>;
- };
+ struct rebind_simd<_Tp, simd_mask<_Up, _Abi>,
+ void_t<simd_abi::deduce_t<_Tp, simd_size_v<_Up, _Abi>, _Abi>>>
+ { using type = simd_mask<_Tp, simd_abi::deduce_t<_Tp, simd_size_v<_Up, _Abi>, _Abi>>; };
template <typename _Tp, typename _V>
using rebind_simd_t = typename rebind_simd<_Tp, _V>::type;
@@ -2908,13 +2920,11 @@ struct rebind_simd
struct resize_simd;
template <int _Np, typename _Tp, typename _Abi>
- struct resize_simd<_Np, simd<_Tp, _Abi>,
- void_t<simd_abi::deduce_t<_Tp, _Np, _Abi>>>
+ struct resize_simd<_Np, simd<_Tp, _Abi>, void_t<simd_abi::deduce_t<_Tp, _Np, _Abi>>>
{ using type = simd<_Tp, simd_abi::deduce_t<_Tp, _Np, _Abi>>; };
template <int _Np, typename _Tp, typename _Abi>
- struct resize_simd<_Np, simd_mask<_Tp, _Abi>,
- void_t<simd_abi::deduce_t<_Tp, _Np, _Abi>>>
+ struct resize_simd<_Np, simd_mask<_Tp, _Abi>, void_t<simd_abi::deduce_t<_Tp, _Np, _Abi>>>
{ using type = simd_mask<_Tp, simd_abi::deduce_t<_Tp, _Np, _Abi>>; };
template <int _Np, typename _V>
@@ -2963,13 +2973,11 @@ struct is_simd_mask<simd_mask<_Tp, _Abi>>
// casts [simd.casts] {{{1
// static_simd_cast {{{2
-template <typename _Tp, typename _Up, typename _Ap, bool = is_simd_v<_Tp>,
- typename = void>
+template <typename _Tp, typename _Up, typename _Ap, bool = is_simd_v<_Tp>, typename = void>
struct __static_simd_cast_return_type;
template <typename _Tp, typename _A0, typename _Up, typename _Ap>
- struct __static_simd_cast_return_type<simd_mask<_Tp, _A0>, _Up, _Ap, false,
- void>
+ struct __static_simd_cast_return_type<simd_mask<_Tp, _A0>, _Up, _Ap, false, void>
: __static_simd_cast_return_type<simd<_Tp, _A0>, _Up, _Ap> {};
template <typename _Tp, typename _Up, typename _Ap>
@@ -3284,6 +3292,7 @@ __get_lvalue(const const_where_expression& __x)
public:
const_where_expression(const const_where_expression&) = delete;
+
const_where_expression& operator=(const const_where_expression&) = delete;
_GLIBCXX_SIMD_INTRINSIC const_where_expression(const _M& __kk, const _Tp& dd)
@@ -3328,8 +3337,8 @@ class const_where_expression<bool, _Tp>
struct _Wrapper { using value_type = _V; };
protected:
- using value_type =
- typename conditional_t<is_arithmetic_v<_V>, _Wrapper, _V>::value_type;
+ using value_type
+ = typename conditional_t<is_arithmetic_v<_V>, _Wrapper, _V>::value_type;
_GLIBCXX_SIMD_INTRINSIC friend const _M&
__get_mask(const const_where_expression& __x)
@@ -3426,32 +3435,32 @@ static_assert(
_GLIBCXX_SIMD_OP_(>>, _S_shift_right);
#undef _GLIBCXX_SIMD_OP_
- _GLIBCXX_SIMD_INTRINSIC void operator++() &&
+ _GLIBCXX_SIMD_INTRINSIC void
+ operator++() &&
{
__data(_M_value)
- = _Impl::template _S_masked_unary<__increment>(__data(_M_k),
- __data(_M_value));
+ = _Impl::template _S_masked_unary<__increment>(__data(_M_k), __data(_M_value));
}
- _GLIBCXX_SIMD_INTRINSIC void operator++(int) &&
+ _GLIBCXX_SIMD_INTRINSIC void
+ operator++(int) &&
{
__data(_M_value)
- = _Impl::template _S_masked_unary<__increment>(__data(_M_k),
- __data(_M_value));
+ = _Impl::template _S_masked_unary<__increment>(__data(_M_k), __data(_M_value));
}
- _GLIBCXX_SIMD_INTRINSIC void operator--() &&
+ _GLIBCXX_SIMD_INTRINSIC void
+ operator--() &&
{
__data(_M_value)
- = _Impl::template _S_masked_unary<__decrement>(__data(_M_k),
- __data(_M_value));
+ = _Impl::template _S_masked_unary<__decrement>(__data(_M_k), __data(_M_value));
}
- _GLIBCXX_SIMD_INTRINSIC void operator--(int) &&
+ _GLIBCXX_SIMD_INTRINSIC void
+ operator--(int) &&
{
__data(_M_value)
- = _Impl::template _S_masked_unary<__decrement>(__data(_M_k),
- __data(_M_value));
+ = _Impl::template _S_masked_unary<__decrement>(__data(_M_k), __data(_M_value));
}
// intentionally hides const_where_expression::copy_from
@@ -3459,15 +3468,15 @@ static_assert(
_GLIBCXX_SIMD_INTRINSIC void
copy_from(const _LoadStorePtr<_Up, value_type>* __mem, _Flags) &&
{
- __data(_M_value)
- = _Impl::_S_masked_load(__data(_M_value), __data(_M_k),
- _Flags::template _S_apply<_Tp>(__mem));
+ __data(_M_value) = _Impl::_S_masked_load(__data(_M_value), __data(_M_k),
+ _Flags::template _S_apply<_Tp>(__mem));
}
};
// where_expression<bool, T> {{{2
template <typename _Tp>
- class where_expression<bool, _Tp> : public const_where_expression<bool, _Tp>
+ class where_expression<bool, _Tp>
+ : public const_where_expression<bool, _Tp>
{
using _M = bool;
using typename const_where_expression<_M, _Tp>::value_type;
@@ -3478,12 +3487,14 @@ class where_expression<bool, _Tp> : public const_where_expression<bool, _Tp>
where_expression(const where_expression&) = delete;
where_expression& operator=(const where_expression&) = delete;
- _GLIBCXX_SIMD_INTRINSIC where_expression(const _M& __kk, _Tp& dd)
- : const_where_expression<_M, _Tp>(__kk, dd) {}
+ _GLIBCXX_SIMD_INTRINSIC
+ where_expression(const _M& __kk, _Tp& dd)
+ : const_where_expression<_M, _Tp>(__kk, dd) {}
#define _GLIBCXX_SIMD_OP_(__op) \
template <typename _Up> \
- _GLIBCXX_SIMD_INTRINSIC void operator __op(_Up&& __x)&& \
+ _GLIBCXX_SIMD_INTRINSIC void \
+ operator __op(_Up&& __x)&& \
{ if (_M_k) _M_value __op static_cast<_Up&&>(__x); }
_GLIBCXX_SIMD_OP_(=)
@@ -3499,16 +3510,20 @@ class where_expression<bool, _Tp> : public const_where_expression<bool, _Tp>
_GLIBCXX_SIMD_OP_(>>=)
#undef _GLIBCXX_SIMD_OP_
- _GLIBCXX_SIMD_INTRINSIC void operator++() &&
+ _GLIBCXX_SIMD_INTRINSIC void
+ operator++() &&
{ if (_M_k) ++_M_value; }
- _GLIBCXX_SIMD_INTRINSIC void operator++(int) &&
+ _GLIBCXX_SIMD_INTRINSIC void
+ operator++(int) &&
{ if (_M_k) ++_M_value; }
- _GLIBCXX_SIMD_INTRINSIC void operator--() &&
+ _GLIBCXX_SIMD_INTRINSIC void
+ operator--() &&
{ if (_M_k) --_M_value; }
- _GLIBCXX_SIMD_INTRINSIC void operator--(int) &&
+ _GLIBCXX_SIMD_INTRINSIC void
+ operator--(int) &&
{ if (_M_k) --_M_value; }
// intentionally hides const_where_expression::copy_from
@@ -3526,23 +3541,20 @@ class where_expression<bool, _Tp> : public const_where_expression<bool, _Tp>
template <typename _Tp, typename _Ap>
_GLIBCXX_SIMD_INTRINSIC
- const_where_expression<simd_mask<_Tp, _Ap>, simd<_Tp, _Ap>>
- where(const typename simd<_Tp, _Ap>::mask_type& __k,
- const simd<_Tp, _Ap>& __value)
+ const_where_expression<simd_mask<_Tp, _Ap>, simd<_Tp, _Ap>>
+ where(const typename simd<_Tp, _Ap>::mask_type& __k, const simd<_Tp, _Ap>& __value)
{ return {__k, __value}; }
template <typename _Tp, typename _Ap>
_GLIBCXX_SIMD_INTRINSIC
- where_expression<simd_mask<_Tp, _Ap>, simd_mask<_Tp, _Ap>>
- where(const remove_const_t<simd_mask<_Tp, _Ap>>& __k,
- simd_mask<_Tp, _Ap>& __value)
+ where_expression<simd_mask<_Tp, _Ap>, simd_mask<_Tp, _Ap>>
+ where(const remove_const_t<simd_mask<_Tp, _Ap>>& __k, simd_mask<_Tp, _Ap>& __value)
{ return {__k, __value}; }
template <typename _Tp, typename _Ap>
_GLIBCXX_SIMD_INTRINSIC
- const_where_expression<simd_mask<_Tp, _Ap>, simd_mask<_Tp, _Ap>>
- where(const remove_const_t<simd_mask<_Tp, _Ap>>& __k,
- const simd_mask<_Tp, _Ap>& __value)
+ const_where_expression<simd_mask<_Tp, _Ap>, simd_mask<_Tp, _Ap>>
+ where(const remove_const_t<simd_mask<_Tp, _Ap>>& __k, const simd_mask<_Tp, _Ap>& __value)
{ return {__k, __value}; }
template <typename _Tp>
@@ -3555,11 +3567,11 @@ where(_ExactBool __k, _Tp& __value)
where(_ExactBool __k, const _Tp& __value)
{ return {__k, __value}; }
- template <typename _Tp, typename _Ap>
- void where(bool __k, simd<_Tp, _Ap>& __value) = delete;
+template <typename _Tp, typename _Ap>
+ void where(bool __k, simd<_Tp, _Ap>& __value) = delete;
- template <typename _Tp, typename _Ap>
- void where(bool __k, const simd<_Tp, _Ap>& __value) = delete;
+template <typename _Tp, typename _Ap>
+ void where(bool __k, const simd<_Tp, _Ap>& __value) = delete;
// proposed mask iterations {{{1
namespace __proposed {
@@ -3576,10 +3588,12 @@ class iterator
size_t __mask;
size_t __bit;
- _GLIBCXX_SIMD_INTRINSIC void __next_bit()
+ _GLIBCXX_SIMD_INTRINSIC void
+ __next_bit()
{ __bit = __builtin_ctzl(__mask); }
- _GLIBCXX_SIMD_INTRINSIC void __reset_lsb()
+ _GLIBCXX_SIMD_INTRINSIC void
+ __reset_lsb()
{
// 01100100 - 1 = 01100011
__mask &= (__mask - 1);
@@ -3591,20 +3605,24 @@ class iterator
iterator(const iterator&) = default;
iterator(iterator&&) = default;
- _GLIBCXX_SIMD_ALWAYS_INLINE size_t operator->() const
+ _GLIBCXX_SIMD_ALWAYS_INLINE size_t
+ operator->() const
{ return __bit; }
- _GLIBCXX_SIMD_ALWAYS_INLINE size_t operator*() const
+ _GLIBCXX_SIMD_ALWAYS_INLINE size_t
+ operator*() const
{ return __bit; }
- _GLIBCXX_SIMD_ALWAYS_INLINE iterator& operator++()
+ _GLIBCXX_SIMD_ALWAYS_INLINE iterator&
+ operator++()
{
__reset_lsb();
__next_bit();
return *this;
}
- _GLIBCXX_SIMD_ALWAYS_INLINE iterator operator++(int)
+ _GLIBCXX_SIMD_ALWAYS_INLINE iterator
+ operator++(int)
{
iterator __tmp = *this;
__reset_lsb();
@@ -3612,17 +3630,21 @@ class iterator
return __tmp;
}
- _GLIBCXX_SIMD_ALWAYS_INLINE bool operator==(const iterator& __rhs) const
+ _GLIBCXX_SIMD_ALWAYS_INLINE bool
+ operator==(const iterator& __rhs) const
{ return __mask == __rhs.__mask; }
- _GLIBCXX_SIMD_ALWAYS_INLINE bool operator!=(const iterator& __rhs) const
+ _GLIBCXX_SIMD_ALWAYS_INLINE bool
+ operator!=(const iterator& __rhs) const
{ return __mask != __rhs.__mask; }
};
- iterator begin() const
+ iterator
+ begin() const
{ return __bits.to_ullong(); }
- iterator end() const
+ iterator
+ end() const
{ return 0; }
};
@@ -3637,15 +3659,13 @@ where(const simd_mask<_Tp, _Ap>& __k)
// reductions [simd.reductions] {{{1
template <typename _Tp, typename _Abi, typename _BinaryOperation = plus<>>
_GLIBCXX_SIMD_INTRINSIC _GLIBCXX_SIMD_CONSTEXPR _Tp
- reduce(const simd<_Tp, _Abi>& __v,
- _BinaryOperation __binary_op = _BinaryOperation())
+ reduce(const simd<_Tp, _Abi>& __v, _BinaryOperation __binary_op = _BinaryOperation())
{ return _Abi::_SimdImpl::_S_reduce(__v, __binary_op); }
template <typename _M, typename _V, typename _BinaryOperation = plus<>>
_GLIBCXX_SIMD_INTRINSIC typename _V::value_type
reduce(const const_where_expression<_M, _V>& __x,
- typename _V::value_type __identity_element,
- _BinaryOperation __binary_op)
+ typename _V::value_type __identity_element, _BinaryOperation __binary_op)
{
if (__builtin_expect(none_of(__get_mask(__x)), false))
return __identity_element;
@@ -3684,16 +3704,12 @@ reduce(const const_where_expression<_M, _V>& __x, bit_xor<> __binary_op)
template <typename _Tp, typename _Abi>
_GLIBCXX_SIMD_INTRINSIC _GLIBCXX_SIMD_CONSTEXPR _Tp
hmin(const simd<_Tp, _Abi>& __v) noexcept
- {
- return _Abi::_SimdImpl::_S_reduce(__v, __detail::_Minimum());
- }
+ { return _Abi::_SimdImpl::_S_reduce(__v, __detail::_Minimum()); }
template <typename _Tp, typename _Abi>
_GLIBCXX_SIMD_INTRINSIC _GLIBCXX_SIMD_CONSTEXPR _Tp
hmax(const simd<_Tp, _Abi>& __v) noexcept
- {
- return _Abi::_SimdImpl::_S_reduce(__v, __detail::_Maximum());
- }
+ { return _Abi::_SimdImpl::_S_reduce(__v, __detail::_Maximum()); }
template <typename _M, typename _V>
_GLIBCXX_SIMD_INTRINSIC _GLIBCXX_SIMD_CONSTEXPR
@@ -3761,8 +3777,7 @@ minmax(const simd<_Tp, _Ap>& __a, const simd<_Tp, _Ap>& __b)
template <typename _Tp, typename _Ap>
_GLIBCXX_SIMD_INTRINSIC _GLIBCXX_SIMD_CONSTEXPR simd<_Tp, _Ap>
- clamp(const simd<_Tp, _Ap>& __v, const simd<_Tp, _Ap>& __lo,
- const simd<_Tp, _Ap>& __hi)
+ clamp(const simd<_Tp, _Ap>& __v, const simd<_Tp, _Ap>& __lo, const simd<_Tp, _Ap>& __hi)
{
using _Impl = typename _Ap::_SimdImpl;
return {__private_init,
@@ -3783,8 +3798,7 @@ clamp(const simd<_Tp, _Ap>& __v, const simd<_Tp, _Ap>& __lo,
_SimdWrapper<_Tp, _Np / _Total * _Combine>
__extract_part(const _SimdWrapper<_Tp, _Np> __x);
-template <int _Index, int _Parts, int _Combine = 1, typename _Tp, typename _A0,
- typename... _As>
+template <int _Index, int _Parts, int _Combine = 1, typename _Tp, typename _A0, typename... _As>
_GLIBCXX_SIMD_INTRINSIC auto
__extract_part(const _SimdTuple<_Tp, _A0, _As...>& __x);
@@ -3794,7 +3808,8 @@ clamp(const simd<_Tp, _Ap>& __v, const simd<_Tp, _Ap>& __lo,
struct _SizeList
{
template <size_t _I>
- static constexpr size_t _S_at(_SizeConstant<_I> = {})
+ static constexpr size_t
+ _S_at(_SizeConstant<_I> = {})
{
if constexpr (_I == 0)
return _V0;
@@ -3803,7 +3818,8 @@ struct _SizeList
}
template <size_t _I>
- static constexpr auto _S_before(_SizeConstant<_I> = {})
+ static constexpr auto
+ _S_before(_SizeConstant<_I> = {})
{
if constexpr (_I == 0)
return _SizeConstant<0>();
@@ -3813,7 +3829,8 @@ struct _SizeList
}
template <size_t _Np>
- static constexpr auto _S_pop_front(_SizeConstant<_Np> = {})
+ static constexpr auto
+ _S_pop_front(_SizeConstant<_Np> = {})
{
if constexpr (_Np == 0)
return _SizeList();
@@ -3965,8 +3982,7 @@ split(const simd<typename _V::value_type, _Ap>& __x)
// }}}
// split<simd_mask>(simd_mask) {{{
template <typename _V, typename _Ap,
- size_t _Parts
- = simd_size_v<typename _V::simd_type::value_type, _Ap> / _V::size()>
+ size_t _Parts = simd_size_v<typename _V::simd_type::value_type, _Ap> / _V::size()>
enable_if_t<is_simd_mask_v<_V> && simd_size_v<typename
_V::simd_type::value_type, _Ap> == _Parts * _V::size(), array<_V, _Parts>>
split(const simd_mask<typename _V::simd_type::value_type, _Ap>& __x)
@@ -4131,8 +4147,7 @@ static_assert(
// __store_pack_of_simd {{{
template <typename _Tp, typename _A0, typename... _As>
_GLIBCXX_SIMD_INTRINSIC void
- __store_pack_of_simd(char* __mem, const simd<_Tp, _A0>& __x0,
- const simd<_Tp, _As>&... __xs)
+ __store_pack_of_simd(char* __mem, const simd<_Tp, _A0>& __x0, const simd<_Tp, _As>&... __xs)
{
constexpr size_t __n_bytes = sizeof(_Tp) * simd_size_v<_Tp, _A0>;
__builtin_memcpy(__mem, &__data(__x0), __n_bytes);
@@ -4188,7 +4203,8 @@ class _SmartReference
int _M_index;
_Up& _M_obj;
- _GLIBCXX_SIMD_INTRINSIC constexpr _ValueType _M_read() const noexcept
+ _GLIBCXX_SIMD_INTRINSIC constexpr _ValueType
+ _M_read() const noexcept
{
if constexpr (is_arithmetic_v<_Up>)
return _M_obj;
@@ -4197,7 +4213,8 @@ class _SmartReference
}
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC constexpr void _M_write(_Tp&& __x) const
+ _GLIBCXX_SIMD_INTRINSIC constexpr void
+ _M_write(_Tp&& __x) const
{ _Accessor::_S_set(_M_obj, _M_index, static_cast<_Tp&&>(__x)); }
public:
@@ -4207,32 +4224,32 @@ _SmartReference(_Up& __o, int __i) noexcept
using value_type = _ValueType;
- _GLIBCXX_SIMD_INTRINSIC _SmartReference(const _SmartReference&) = delete;
+ _GLIBCXX_SIMD_INTRINSIC
+ _SmartReference(const _SmartReference&) = delete;
- _GLIBCXX_SIMD_INTRINSIC constexpr operator value_type() const noexcept
+ _GLIBCXX_SIMD_INTRINSIC constexpr
+ operator value_type() const noexcept
{ return _M_read(); }
- template <typename _Tp,
- typename
- = _ValuePreservingOrInt<__remove_cvref_t<_Tp>, value_type>>
- _GLIBCXX_SIMD_INTRINSIC constexpr _SmartReference operator=(_Tp&& __x) &&
+ template <typename _Tp, typename = _ValuePreservingOrInt<__remove_cvref_t<_Tp>, value_type>>
+ _GLIBCXX_SIMD_INTRINSIC constexpr _SmartReference
+ operator=(_Tp&& __x) &&
{
_M_write(static_cast<_Tp&&>(__x));
return {_M_obj, _M_index};
}
-#define _GLIBCXX_SIMD_OP_(__op) \
- template <typename _Tp, \
- typename _TT \
- = decltype(declval<value_type>() __op declval<_Tp>()), \
- typename = _ValuePreservingOrInt<__remove_cvref_t<_Tp>, _TT>, \
- typename = _ValuePreservingOrInt<_TT, value_type>> \
- _GLIBCXX_SIMD_INTRINSIC constexpr _SmartReference \
- operator __op##=(_Tp&& __x) && \
- { \
- const value_type& __lhs = _M_read(); \
- _M_write(__lhs __op __x); \
- return {_M_obj, _M_index}; \
+#define _GLIBCXX_SIMD_OP_(__op) \
+ template <typename _Tp, \
+ typename _TT = decltype(declval<value_type>() __op declval<_Tp>()), \
+ typename = _ValuePreservingOrInt<__remove_cvref_t<_Tp>, _TT>, \
+ typename = _ValuePreservingOrInt<_TT, value_type>> \
+ _GLIBCXX_SIMD_INTRINSIC constexpr _SmartReference \
+ operator __op##=(_Tp&& __x) && \
+ { \
+ const value_type& __lhs = _M_read(); \
+ _M_write(__lhs __op __x); \
+ return {_M_obj, _M_index}; \
}
_GLIBCXX_SIMD_ALL_ARITHMETICS(_GLIBCXX_SIMD_OP_);
_GLIBCXX_SIMD_ALL_SHIFTS(_GLIBCXX_SIMD_OP_);
@@ -4240,9 +4257,9 @@ _SmartReference(_Up& __o, int __i) noexcept
#undef _GLIBCXX_SIMD_OP_
template <typename _Tp = void,
- typename
- = decltype(++declval<conditional_t<true, value_type, _Tp>&>())>
- _GLIBCXX_SIMD_INTRINSIC constexpr _SmartReference operator++() &&
+ typename = decltype(++declval<conditional_t<true, value_type, _Tp>&>())>
+ _GLIBCXX_SIMD_INTRINSIC constexpr _SmartReference
+ operator++() &&
{
value_type __x = _M_read();
_M_write(++__x);
@@ -4250,9 +4267,9 @@ _SmartReference(_Up& __o, int __i) noexcept
}
template <typename _Tp = void,
- typename
- = decltype(declval<conditional_t<true, value_type, _Tp>&>()++)>
- _GLIBCXX_SIMD_INTRINSIC constexpr value_type operator++(int) &&
+ typename = decltype(declval<conditional_t<true, value_type, _Tp>&>()++)>
+ _GLIBCXX_SIMD_INTRINSIC constexpr value_type
+ operator++(int) &&
{
const value_type __r = _M_read();
value_type __x = __r;
@@ -4261,9 +4278,9 @@ _SmartReference(_Up& __o, int __i) noexcept
}
template <typename _Tp = void,
- typename
- = decltype(--declval<conditional_t<true, value_type, _Tp>&>())>
- _GLIBCXX_SIMD_INTRINSIC constexpr _SmartReference operator--() &&
+ typename = decltype(--declval<conditional_t<true, value_type, _Tp>&>())>
+ _GLIBCXX_SIMD_INTRINSIC constexpr _SmartReference
+ operator--() &&
{
value_type __x = _M_read();
_M_write(--__x);
@@ -4271,9 +4288,9 @@ _SmartReference(_Up& __o, int __i) noexcept
}
template <typename _Tp = void,
- typename
- = decltype(declval<conditional_t<true, value_type, _Tp>&>()--)>
- _GLIBCXX_SIMD_INTRINSIC constexpr value_type operator--(int) &&
+ typename = decltype(declval<conditional_t<true, value_type, _Tp>&>()--)>
+ _GLIBCXX_SIMD_INTRINSIC constexpr value_type
+ operator--(int) &&
{
const value_type __r = _M_read();
value_type __x = __r;
@@ -4349,7 +4366,8 @@ struct __decay_abi<__scalar_abi_wrapper<_Bytes>>
template <template <int> class _Abi, int _Bytes, typename _Tp>
struct __find_next_valid_abi
{
- static constexpr auto _S_choose()
+ static constexpr auto
+ _S_choose()
{
constexpr int _NextBytes = std::__bit_ceil(_Bytes) / 2;
using _NextAbi = _Abi<_NextBytes>;
@@ -4393,7 +4411,8 @@ struct _AbiList<_A0, _Rest...>
typename _AbiList<_Rest...>::template _FirstValidAbi<_Tp, _Np>>;
template <typename _Tp, int _Np>
- static constexpr auto _S_determine_best_abi()
+ static constexpr auto
+ _S_determine_best_abi()
{
static_assert(_Np >= 1);
constexpr int _Bytes = sizeof(_Tp) * _Np;
@@ -4556,17 +4575,15 @@ simd_mask(const simd_mask<_Up, simd_abi::fixed_size<size()>>& __x)
template <typename _Flags>
_GLIBCXX_SIMD_ALWAYS_INLINE
simd_mask(const value_type* __mem, _Flags)
- : _M_data(_Impl::template _S_load<_Ip>(
- _Flags::template _S_apply<simd_mask>(__mem))) {}
+ : _M_data(_Impl::template _S_load<_Ip>(_Flags::template _S_apply<simd_mask>(__mem))) {}
template <typename _Flags>
_GLIBCXX_SIMD_ALWAYS_INLINE
simd_mask(const value_type* __mem, simd_mask __k, _Flags)
: _M_data{}
{
- _M_data
- = _Impl::_S_masked_load(_M_data, __k._M_data,
- _Flags::template _S_apply<simd_mask>(__mem));
+ _M_data = _Impl::_S_masked_load(_M_data, __k._M_data,
+ _Flags::template _S_apply<simd_mask>(__mem));
}
// }}}
@@ -4574,10 +4591,7 @@ simd_mask(const simd_mask<_Up, simd_abi::fixed_size<size()>>& __x)
template <typename _Flags>
_GLIBCXX_SIMD_ALWAYS_INLINE void
copy_from(const value_type* __mem, _Flags)
- {
- _M_data = _Impl::template _S_load<_Ip>(
- _Flags::template _S_apply<simd_mask>(__mem));
- }
+ { _M_data = _Impl::template _S_load<_Ip>(_Flags::template _S_apply<simd_mask>(__mem)); }
// }}}
// stores [simd_mask.store] {{{
@@ -4618,8 +4632,7 @@ simd_mask(const simd_mask<_Up, simd_abi::fixed_size<size()>>& __x)
#ifdef _GLIBCXX_SIMD_ENABLE_IMPLICIT_MASK_CAST
// simd_mask<int> && simd_mask<uint> needs disambiguation
template <typename _Up, typename _A2,
- typename
- = enable_if_t<is_convertible_v<simd_mask<_Up, _A2>, simd_mask>>>
+ typename = enable_if_t<is_convertible_v<simd_mask<_Up, _A2>, simd_mask>>>
_GLIBCXX_SIMD_ALWAYS_INLINE friend simd_mask
operator&&(const simd_mask& __x, const simd_mask<_Up, _A2>& __y)
{
@@ -4628,8 +4641,7 @@ simd_mask(const simd_mask<_Up, simd_abi::fixed_size<size()>>& __x)
}
template <typename _Up, typename _A2,
- typename
- = enable_if_t<is_convertible_v<simd_mask<_Up, _A2>, simd_mask>>>
+ typename = enable_if_t<is_convertible_v<simd_mask<_Up, _A2>, simd_mask>>>
_GLIBCXX_SIMD_ALWAYS_INLINE friend simd_mask
operator||(const simd_mask& __x, const simd_mask<_Up, _A2>& __y)
{
@@ -4640,15 +4652,11 @@ simd_mask(const simd_mask<_Up, simd_abi::fixed_size<size()>>& __x)
_GLIBCXX_SIMD_ALWAYS_INLINE friend simd_mask
operator&&(const simd_mask& __x, const simd_mask& __y)
- {
- return {__private_init, _Impl::_S_logical_and(__x._M_data, __y._M_data)};
- }
+ { return {__private_init, _Impl::_S_logical_and(__x._M_data, __y._M_data)}; }
_GLIBCXX_SIMD_ALWAYS_INLINE friend simd_mask
operator||(const simd_mask& __x, const simd_mask& __y)
- {
- return {__private_init, _Impl::_S_logical_or(__x._M_data, __y._M_data)};
- }
+ { return {__private_init, _Impl::_S_logical_or(__x._M_data, __y._M_data)}; }
_GLIBCXX_SIMD_ALWAYS_INLINE friend simd_mask
operator&(const simd_mask& __x, const simd_mask& __y)
@@ -4714,8 +4722,7 @@ simd_mask(const simd_mask<_Up, simd_abi::fixed_size<size()>>& __x)
// }}}
// bitset_init ctor {{{
_GLIBCXX_SIMD_INTRINSIC simd_mask(_BitsetInit, bitset<size()> __init)
- : _M_data(
- _Impl::_S_from_bitmask(_SanitizedBitMask<size()>(__init), _S_type_tag))
+ : _M_data(_Impl::_S_from_bitmask(_SanitizedBitMask<size()>(__init), _S_type_tag))
{}
// }}}
@@ -4727,8 +4734,7 @@ simd_mask(const simd_mask<_Up, simd_abi::fixed_size<size()>>& __x)
struct _CvtProxy
{
template <typename _Up, typename _A2,
- typename
- = enable_if_t<simd_size_v<_Up, _A2> == simd_size_v<_Tp, _Abi>>>
+ typename = enable_if_t<simd_size_v<_Up, _A2> == simd_size_v<_Tp, _Abi>>>
_GLIBCXX_SIMD_ALWAYS_INLINE
operator simd_mask<_Up, _A2>() &&
{
@@ -5419,26 +5425,17 @@ namespace __float_bitwise_operators
template <typename _Tp, typename _Ap>
_GLIBCXX_SIMD_INTRINSIC _GLIBCXX_SIMD_CONSTEXPR simd<_Tp, _Ap>
operator^(const simd<_Tp, _Ap>& __a, const simd<_Tp, _Ap>& __b)
- {
- return {__private_init,
- _Ap::_SimdImpl::_S_bit_xor(__data(__a), __data(__b))};
- }
+ { return {__private_init, _Ap::_SimdImpl::_S_bit_xor(__data(__a), __data(__b))}; }
template <typename _Tp, typename _Ap>
_GLIBCXX_SIMD_INTRINSIC _GLIBCXX_SIMD_CONSTEXPR simd<_Tp, _Ap>
operator|(const simd<_Tp, _Ap>& __a, const simd<_Tp, _Ap>& __b)
- {
- return {__private_init,
- _Ap::_SimdImpl::_S_bit_or(__data(__a), __data(__b))};
- }
+ { return {__private_init, _Ap::_SimdImpl::_S_bit_or(__data(__a), __data(__b))}; }
template <typename _Tp, typename _Ap>
_GLIBCXX_SIMD_INTRINSIC _GLIBCXX_SIMD_CONSTEXPR simd<_Tp, _Ap>
operator&(const simd<_Tp, _Ap>& __a, const simd<_Tp, _Ap>& __b)
- {
- return {__private_init,
- _Ap::_SimdImpl::_S_bit_and(__data(__a), __data(__b))};
- }
+ { return {__private_init, _Ap::_SimdImpl::_S_bit_and(__data(__a), __data(__b))}; }
template <typename _Tp, typename _Ap>
_GLIBCXX_SIMD_INTRINSIC _GLIBCXX_SIMD_CONSTEXPR
diff --git a/libstdc++-v3/include/experimental/bits/simd_builtin.h b/libstdc++-v3/include/experimental/bits/simd_builtin.h
index 4a4de4534f3..0e75f941288 100644
--- a/libstdc++-v3/include/experimental/bits/simd_builtin.h
+++ b/libstdc++-v3/include/experimental/bits/simd_builtin.h
@@ -836,22 +836,19 @@ struct _GnuTraits
// _SimdBase / base class for simd, providing extra conversions {{{
struct _SimdBase2
{
- _GLIBCXX_SIMD_ALWAYS_INLINE
- explicit operator __intrinsic_type_t<_Tp, _Np>() const
- {
- return __to_intrin(static_cast<const simd<_Tp, _Abi>*>(this)->_M_data);
- }
- _GLIBCXX_SIMD_ALWAYS_INLINE
- explicit operator __vector_type_t<_Tp, _Np>() const
- {
- return static_cast<const simd<_Tp, _Abi>*>(this)->_M_data.__builtin();
- }
+ _GLIBCXX_SIMD_ALWAYS_INLINE explicit
+ operator __intrinsic_type_t<_Tp, _Np>() const
+ { return __to_intrin(static_cast<const simd<_Tp, _Abi>*>(this)->_M_data); }
+
+ _GLIBCXX_SIMD_ALWAYS_INLINE explicit
+ operator __vector_type_t<_Tp, _Np>() const
+ { return static_cast<const simd<_Tp, _Abi>*>(this)->_M_data.__builtin(); }
};
struct _SimdBase1
{
- _GLIBCXX_SIMD_ALWAYS_INLINE
- explicit operator __intrinsic_type_t<_Tp, _Np>() const
+ _GLIBCXX_SIMD_ALWAYS_INLINE explicit
+ operator __intrinsic_type_t<_Tp, _Np>() const
{ return __data(*static_cast<const simd<_Tp, _Abi>*>(this)); }
};
@@ -863,23 +860,19 @@ struct _SimdBase1
// _MaskBase {{{
struct _MaskBase2
{
- _GLIBCXX_SIMD_ALWAYS_INLINE
- explicit operator __intrinsic_type_t<_Tp, _Np>() const
- {
- return static_cast<const simd_mask<_Tp, _Abi>*>(this)
- ->_M_data.__intrin();
- }
- _GLIBCXX_SIMD_ALWAYS_INLINE
- explicit operator __vector_type_t<_Tp, _Np>() const
- {
- return static_cast<const simd_mask<_Tp, _Abi>*>(this)->_M_data._M_data;
- }
+ _GLIBCXX_SIMD_ALWAYS_INLINE explicit
+ operator __intrinsic_type_t<_Tp, _Np>() const
+ { return static_cast<const simd_mask<_Tp, _Abi>*>(this) ->_M_data.__intrin(); }
+
+ _GLIBCXX_SIMD_ALWAYS_INLINE explicit
+ operator __vector_type_t<_Tp, _Np>() const
+ { return static_cast<const simd_mask<_Tp, _Abi>*>(this)->_M_data._M_data; }
};
struct _MaskBase1
{
- _GLIBCXX_SIMD_ALWAYS_INLINE
- explicit operator __intrinsic_type_t<_Tp, _Np>() const
+ _GLIBCXX_SIMD_ALWAYS_INLINE explicit
+ operator __intrinsic_type_t<_Tp, _Np>() const
{ return __data(*static_cast<const simd_mask<_Tp, _Abi>*>(this)); }
};
@@ -898,6 +891,7 @@ class _MaskCastType
public:
_GLIBCXX_SIMD_ALWAYS_INLINE
_MaskCastType(_Up __x) : _M_data(__x) {}
+
_GLIBCXX_SIMD_ALWAYS_INLINE
operator _MaskMember() const { return _M_data; }
};
@@ -913,6 +907,7 @@ class _SimdCastType1
public:
_GLIBCXX_SIMD_ALWAYS_INLINE
_SimdCastType1(_Ap __a) : _M_data(__vector_bitcast<_Tp>(__a)) {}
+
_GLIBCXX_SIMD_ALWAYS_INLINE
operator _SimdMember() const { return _M_data; }
};
@@ -926,8 +921,10 @@ class _SimdCastType2
public:
_GLIBCXX_SIMD_ALWAYS_INLINE
_SimdCastType2(_Ap __a) : _M_data(__vector_bitcast<_Tp>(__a)) {}
+
_GLIBCXX_SIMD_ALWAYS_INLINE
_SimdCastType2(_Bp __b) : _M_data(__b) {}
+
_GLIBCXX_SIMD_ALWAYS_INLINE
operator _SimdMember() const { return _M_data; }
};
@@ -1039,16 +1036,13 @@ _S_implicit_mask()
}
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static constexpr __intrinsic_type_t<_Tp,
- _S_size<_Tp>>
+ _GLIBCXX_SIMD_INTRINSIC static constexpr __intrinsic_type_t<_Tp, _S_size<_Tp>>
_S_implicit_mask_intrin()
- {
- return __to_intrin(
- __vector_bitcast<_Tp>(_S_implicit_mask<_Tp>()._M_data));
- }
+ { return __to_intrin(__vector_bitcast<_Tp>(_S_implicit_mask<_Tp>()._M_data)); }
template <typename _TW, typename _TVT = _VectorTraits<_TW>>
- _GLIBCXX_SIMD_INTRINSIC static constexpr _TW _S_masked(_TW __x)
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _TW
+ _S_masked(_TW __x)
{
using _Tp = typename _TVT::value_type;
if constexpr (!_MaskMember<_Tp>::_S_is_partial)
@@ -1170,8 +1164,7 @@ _S_implicit_mask()
{ return __implicit_mask_n<_S_size<_Tp>>(); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static constexpr __bool_storage_member_type_t<
- _S_size<_Tp>>
+ _GLIBCXX_SIMD_INTRINSIC static constexpr __bool_storage_member_type_t<_S_size<_Tp>>
_S_implicit_mask_intrin()
{ return __implicit_mask_n<_S_size<_Tp>>(); }
@@ -1303,7 +1296,8 @@ _S_load(const void* __p)
// }}}
// _S_store {{{
template <size_t _ReqBytes = 0, typename _TV>
- _GLIBCXX_SIMD_INTRINSIC static void _S_store(_TV __x, void* __addr)
+ _GLIBCXX_SIMD_INTRINSIC static void
+ _S_store(_TV __x, void* __addr)
{
constexpr size_t _Bytes = _ReqBytes == 0 ? sizeof(__x) : _ReqBytes;
static_assert(sizeof(__x) >= _Bytes);
@@ -1339,8 +1333,8 @@ _S_load(const void* __p)
}
template <typename _Tp, size_t _Np>
- _GLIBCXX_SIMD_INTRINSIC static void _S_store(_SimdWrapper<_Tp, _Np> __x,
- void* __addr)
+ _GLIBCXX_SIMD_INTRINSIC static void
+ _S_store(_SimdWrapper<_Tp, _Np> __x, void* __addr)
{ _S_store<_Np * sizeof(_Tp)>(__x._M_data, __addr); }
// }}}
@@ -1447,8 +1441,8 @@ _S_broadcast(_Tp __x) noexcept
// _S_generator {{{2
template <typename _Fp, typename _Tp>
- inline static constexpr _SimdMember<_Tp> _S_generator(_Fp&& __gen,
- _TypeTag<_Tp>)
+ inline static constexpr _SimdMember<_Tp>
+ _S_generator(_Fp&& __gen, _TypeTag<_Tp>)
{
return __generate_vector<_Tp, _S_full_size<_Tp>>(
[&](auto __i) constexpr _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA {
@@ -1569,8 +1563,7 @@ _S_masked_load(_SimdWrapper<_Tp, _Np> __merge, _MaskMember<_Tp> __k,
// _S_masked_store_nocvt {{{2
template <typename _Tp, size_t _Np>
_GLIBCXX_SIMD_INTRINSIC static void
- _S_masked_store_nocvt(_SimdWrapper<_Tp, _Np> __v, _Tp* __mem,
- _MaskMember<_Tp> __k)
+ _S_masked_store_nocvt(_SimdWrapper<_Tp, _Np> __v, _Tp* __mem, _MaskMember<_Tp> __k)
{
_BitOps::_S_bit_iteration(
_MaskImpl::_S_to_bits(__k),
@@ -1583,8 +1576,7 @@ _S_masked_store_nocvt(_SimdWrapper<_Tp, _Np> __v, _Tp* __mem,
template <typename _TW, typename _TVT = _VectorTraits<_TW>,
typename _Tp = typename _TVT::value_type, typename _Up>
static inline void
- _S_masked_store(const _TW __v, _Up* __mem, const _MaskMember<_Tp> __k)
- noexcept
+ _S_masked_store(const _TW __v, _Up* __mem, const _MaskMember<_Tp> __k) noexcept
{
constexpr size_t _TV_size = _S_size<_Tp>;
[[maybe_unused]] const auto __vi = __to_intrin(__v);
@@ -1946,7 +1938,8 @@ _S_reduce(simd<_Tp, _Abi> __x, _BinaryOperation&& __binary_op)
// frexp, modf and copysign implemented in simd_math.h
#define _GLIBCXX_SIMD_MATH_FALLBACK(__name) \
template <typename _Tp, typename... _More> \
- static _Tp _S_##__name(const _Tp& __x, const _More&... __more) \
+ static _Tp \
+ _S_##__name(const _Tp& __x, const _More&... __more) \
{ \
return __generate_vector<_Tp>( \
[&](auto __i) _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA { \
@@ -1956,8 +1949,8 @@ _S_reduce(simd<_Tp, _Abi> __x, _BinaryOperation&& __binary_op)
#define _GLIBCXX_SIMD_MATH_FALLBACK_MASKRET(__name) \
template <typename _Tp, typename... _More> \
- static typename _Tp::mask_type _S_##__name(const _Tp& __x, \
- const _More&... __more) \
+ static typename _Tp::mask_type \
+ _S_##__name(const _Tp& __x, const _More&... __more) \
{ \
return __generate_vector<_Tp>( \
[&](auto __i) _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA { \
@@ -1967,7 +1960,8 @@ _S_reduce(simd<_Tp, _Abi> __x, _BinaryOperation&& __binary_op)
#define _GLIBCXX_SIMD_MATH_FALLBACK_FIXEDRET(_RetTp, __name) \
template <typename _Tp, typename... _More> \
- static auto _S_##__name(const _Tp& __x, const _More&... __more) \
+ static auto \
+ _S_##__name(const _Tp& __x, const _More&... __more) \
{ \
return __fixed_size_storage_t<_RetTp, \
_VectorTraits<_Tp>::_S_partial_width>:: \
@@ -2115,22 +2109,22 @@ _S_islessgreater(_SimdWrapper<_Tp, _Np> __x,
#undef _GLIBCXX_SIMD_MATH_FALLBACK_FIXEDRET
// _S_abs {{{3
template <typename _Tp, size_t _Np>
- _GLIBCXX_SIMD_INTRINSIC static _SimdWrapper<_Tp, _Np>
- _S_abs(_SimdWrapper<_Tp, _Np> __x) noexcept
- {
- // if (__builtin_is_constant_evaluated())
- // {
- // return __x._M_data < 0 ? -__x._M_data : __x._M_data;
- // }
- if constexpr (is_floating_point_v<_Tp>)
- // `v < 0 ? -v : v` cannot compile to the efficient implementation of
- // masking the signbit off because it must consider v == -0
-
- // ~(-0.) & v would be easy, but breaks with fno-signed-zeros
- return __and(_S_absmask<__vector_type_t<_Tp, _Np>>, __x._M_data);
- else
- return __x._M_data < 0 ? -__x._M_data : __x._M_data;
- }
+ _GLIBCXX_SIMD_INTRINSIC static _SimdWrapper<_Tp, _Np>
+ _S_abs(_SimdWrapper<_Tp, _Np> __x) noexcept
+ {
+ // if (__builtin_is_constant_evaluated())
+ // {
+ // return __x._M_data < 0 ? -__x._M_data : __x._M_data;
+ // }
+ if constexpr (is_floating_point_v<_Tp>)
+ // `v < 0 ? -v : v` cannot compile to the efficient implementation of
+ // masking the signbit off because it must consider v == -0
+
+ // ~(-0.) & v would be easy, but breaks with fno-signed-zeros
+ return __and(_S_absmask<__vector_type_t<_Tp, _Np>>, __x._M_data);
+ else
+ return __x._M_data < 0 ? -__x._M_data : __x._M_data;
+ }
// }}}3
// _S_plus_minus {{{
@@ -2138,318 +2132,316 @@ _S_abs(_SimdWrapper<_Tp, _Np> __x) noexcept
// - _TV must be __vector_type_t<floating-point type, N>.
// - _UV must be _TV or floating-point type.
template <typename _TV, typename _UV>
- _GLIBCXX_SIMD_INTRINSIC static constexpr _TV _S_plus_minus(_TV __x,
- _UV __y) noexcept
- {
- #if defined __i386__ && !defined __SSE_MATH__
- if constexpr (sizeof(__x) == 8)
- { // operations on __x would use the FPU
- static_assert(is_same_v<_TV, __vector_type_t<float, 2>>);
- const auto __x4 = __vector_bitcast<float, 4>(__x);
- if constexpr (is_same_v<_TV, _UV>)
- return __vector_bitcast<float, 2>(
- _S_plus_minus(__x4, __vector_bitcast<float, 4>(__y)));
- else
- return __vector_bitcast<float, 2>(_S_plus_minus(__x4, __y));
- }
- #endif
- #if !defined __clang__ && __GCC_IEC_559 == 0
- if (__builtin_is_constant_evaluated()
- || (__builtin_constant_p(__x) && __builtin_constant_p(__y)))
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _TV
+ _S_plus_minus(_TV __x, _UV __y) noexcept
+ {
+#if defined __i386__ && !defined __SSE_MATH__
+ if constexpr (sizeof(__x) == 8)
+ { // operations on __x would use the FPU
+ static_assert(is_same_v<_TV, __vector_type_t<float, 2>>);
+ const auto __x4 = __vector_bitcast<float, 4>(__x);
+ if constexpr (is_same_v<_TV, _UV>)
+ return __vector_bitcast<float, 2>(
+ _S_plus_minus(__x4, __vector_bitcast<float, 4>(__y)));
+ else
+ return __vector_bitcast<float, 2>(_S_plus_minus(__x4, __y));
+ }
+#endif
+#if !defined __clang__ && __GCC_IEC_559 == 0
+ if (__builtin_is_constant_evaluated()
+ || (__builtin_constant_p(__x) && __builtin_constant_p(__y)))
+ return (__x + __y) - __y;
+ else
+ return [&] {
+ __x += __y;
+ if constexpr(__have_sse)
+ {
+ if constexpr (sizeof(__x) >= 16)
+ asm("" : "+x"(__x));
+ else if constexpr (is_same_v<__vector_type_t<float, 2>, _TV>)
+ asm("" : "+x"(__x[0]), "+x"(__x[1]));
+ else
+ __assert_unreachable<_TV>();
+ }
+ else if constexpr(__have_neon)
+ asm("" : "+w"(__x));
+ else if constexpr (__have_power_vmx)
+ {
+ if constexpr (is_same_v<__vector_type_t<float, 2>, _TV>)
+ asm("" : "+fgr"(__x[0]), "+fgr"(__x[1]));
+ else
+ asm("" : "+v"(__x));
+ }
+ else
+ asm("" : "+g"(__x));
+ return __x - __y;
+ }();
+#else
return (__x + __y) - __y;
- else
- return [&] {
- __x += __y;
- if constexpr(__have_sse)
- {
- if constexpr (sizeof(__x) >= 16)
- asm("" : "+x"(__x));
- else if constexpr (is_same_v<__vector_type_t<float, 2>, _TV>)
- asm("" : "+x"(__x[0]), "+x"(__x[1]));
- else
- __assert_unreachable<_TV>();
- }
- else if constexpr(__have_neon)
- asm("" : "+w"(__x));
- else if constexpr (__have_power_vmx)
- {
- if constexpr (is_same_v<__vector_type_t<float, 2>, _TV>)
- asm("" : "+fgr"(__x[0]), "+fgr"(__x[1]));
- else
- asm("" : "+v"(__x));
- }
- else
- asm("" : "+g"(__x));
- return __x - __y;
- }();
- #else
- return (__x + __y) - __y;
- #endif
- }
+#endif
+ }
// }}}
// _S_nearbyint {{{3
template <typename _Tp, typename _TVT = _VectorTraits<_Tp>>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_nearbyint(_Tp __x_) noexcept
- {
- using value_type = typename _TVT::value_type;
- using _V = typename _TVT::type;
- const _V __x = __x_;
- const _V __absx = __and(__x, _S_absmask<_V>);
- static_assert(__CHAR_BIT__ * sizeof(1ull) >= __digits_v<value_type>);
- _GLIBCXX_SIMD_USE_CONSTEXPR _V __shifter_abs
- = _V() + (1ull << (__digits_v<value_type> - 1));
- const _V __shifter = __or(__and(_S_signmask<_V>, __x), __shifter_abs);
- const _V __shifted = _S_plus_minus(__x, __shifter);
- return __absx < __shifter_abs ? __shifted : __x;
- }
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_nearbyint(_Tp __x_) noexcept
+ {
+ using value_type = typename _TVT::value_type;
+ using _V = typename _TVT::type;
+ const _V __x = __x_;
+ const _V __absx = __and(__x, _S_absmask<_V>);
+ static_assert(__CHAR_BIT__ * sizeof(1ull) >= __digits_v<value_type>);
+ _GLIBCXX_SIMD_USE_CONSTEXPR _V __shifter_abs
+ = _V() + (1ull << (__digits_v<value_type> - 1));
+ const _V __shifter = __or(__and(_S_signmask<_V>, __x), __shifter_abs);
+ const _V __shifted = _S_plus_minus(__x, __shifter);
+ return __absx < __shifter_abs ? __shifted : __x;
+ }
// _S_rint {{{3
template <typename _Tp, typename _TVT = _VectorTraits<_Tp>>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_rint(_Tp __x) noexcept
- {
- return _SuperImpl::_S_nearbyint(__x);
- }
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_rint(_Tp __x) noexcept
+ { return _SuperImpl::_S_nearbyint(__x); }
// _S_trunc {{{3
template <typename _Tp, size_t _Np>
- _GLIBCXX_SIMD_INTRINSIC static _SimdWrapper<_Tp, _Np>
- _S_trunc(_SimdWrapper<_Tp, _Np> __x)
- {
- using _V = __vector_type_t<_Tp, _Np>;
- const _V __absx = __and(__x._M_data, _S_absmask<_V>);
- static_assert(__CHAR_BIT__ * sizeof(1ull) >= __digits_v<_Tp>);
- constexpr _Tp __shifter = 1ull << (__digits_v<_Tp> - 1);
- _V __truncated = _S_plus_minus(__absx, __shifter);
- __truncated -= __truncated > __absx ? _V() + 1 : _V();
- return __absx < __shifter ? __or(__xor(__absx, __x._M_data), __truncated)
- : __x._M_data;
- }
+ _GLIBCXX_SIMD_INTRINSIC static _SimdWrapper<_Tp, _Np>
+ _S_trunc(_SimdWrapper<_Tp, _Np> __x)
+ {
+ using _V = __vector_type_t<_Tp, _Np>;
+ const _V __absx = __and(__x._M_data, _S_absmask<_V>);
+ static_assert(__CHAR_BIT__ * sizeof(1ull) >= __digits_v<_Tp>);
+ constexpr _Tp __shifter = 1ull << (__digits_v<_Tp> - 1);
+ _V __truncated = _S_plus_minus(__absx, __shifter);
+ __truncated -= __truncated > __absx ? _V() + 1 : _V();
+ return __absx < __shifter ? __or(__xor(__absx, __x._M_data), __truncated)
+ : __x._M_data;
+ }
// _S_round {{{3
template <typename _Tp, size_t _Np>
- _GLIBCXX_SIMD_INTRINSIC static _SimdWrapper<_Tp, _Np>
- _S_round(_SimdWrapper<_Tp, _Np> __x)
- {
- const auto __abs_x = _SuperImpl::_S_abs(__x);
- const auto __t_abs = _SuperImpl::_S_trunc(__abs_x)._M_data;
- const auto __r_abs // round(abs(x)) =
- = __t_abs + (__abs_x._M_data - __t_abs >= _Tp(.5) ? _Tp(1) : 0);
- return __or(__xor(__abs_x._M_data, __x._M_data), __r_abs);
- }
+ _GLIBCXX_SIMD_INTRINSIC static _SimdWrapper<_Tp, _Np>
+ _S_round(_SimdWrapper<_Tp, _Np> __x)
+ {
+ const auto __abs_x = _SuperImpl::_S_abs(__x);
+ const auto __t_abs = _SuperImpl::_S_trunc(__abs_x)._M_data;
+ const auto __r_abs // round(abs(x)) =
+ = __t_abs + (__abs_x._M_data - __t_abs >= _Tp(.5) ? _Tp(1) : 0);
+ return __or(__xor(__abs_x._M_data, __x._M_data), __r_abs);
+ }
// _S_floor {{{3
template <typename _Tp, size_t _Np>
- _GLIBCXX_SIMD_INTRINSIC static _SimdWrapper<_Tp, _Np>
- _S_floor(_SimdWrapper<_Tp, _Np> __x)
- {
- const auto __y = _SuperImpl::_S_trunc(__x)._M_data;
- const auto __negative_input
- = __vector_bitcast<_Tp>(__x._M_data < __vector_broadcast<_Np, _Tp>(0));
- const auto __mask
- = __andnot(__vector_bitcast<_Tp>(__y == __x._M_data), __negative_input);
- return __or(__andnot(__mask, __y),
- __and(__mask, __y - __vector_broadcast<_Np, _Tp>(1)));
- }
+ _GLIBCXX_SIMD_INTRINSIC static _SimdWrapper<_Tp, _Np>
+ _S_floor(_SimdWrapper<_Tp, _Np> __x)
+ {
+ const auto __y = _SuperImpl::_S_trunc(__x)._M_data;
+ const auto __negative_input
+ = __vector_bitcast<_Tp>(__x._M_data < __vector_broadcast<_Np, _Tp>(0));
+ const auto __mask
+ = __andnot(__vector_bitcast<_Tp>(__y == __x._M_data), __negative_input);
+ return __or(__andnot(__mask, __y),
+ __and(__mask, __y - __vector_broadcast<_Np, _Tp>(1)));
+ }
// _S_ceil {{{3
template <typename _Tp, size_t _Np>
- _GLIBCXX_SIMD_INTRINSIC static _SimdWrapper<_Tp, _Np>
- _S_ceil(_SimdWrapper<_Tp, _Np> __x)
- {
- const auto __y = _SuperImpl::_S_trunc(__x)._M_data;
- const auto __negative_input
- = __vector_bitcast<_Tp>(__x._M_data < __vector_broadcast<_Np, _Tp>(0));
- const auto __inv_mask
- = __or(__vector_bitcast<_Tp>(__y == __x._M_data), __negative_input);
- return __or(__and(__inv_mask, __y),
- __andnot(__inv_mask, __y + __vector_broadcast<_Np, _Tp>(1)));
- }
+ _GLIBCXX_SIMD_INTRINSIC static _SimdWrapper<_Tp, _Np>
+ _S_ceil(_SimdWrapper<_Tp, _Np> __x)
+ {
+ const auto __y = _SuperImpl::_S_trunc(__x)._M_data;
+ const auto __negative_input
+ = __vector_bitcast<_Tp>(__x._M_data < __vector_broadcast<_Np, _Tp>(0));
+ const auto __inv_mask
+ = __or(__vector_bitcast<_Tp>(__y == __x._M_data), __negative_input);
+ return __or(__and(__inv_mask, __y),
+ __andnot(__inv_mask, __y + __vector_broadcast<_Np, _Tp>(1)));
+ }
// _S_isnan {{{3
template <typename _Tp, size_t _Np>
- _GLIBCXX_SIMD_INTRINSIC static _MaskMember<_Tp>
- _S_isnan([[maybe_unused]] _SimdWrapper<_Tp, _Np> __x)
- {
- #if __FINITE_MATH_ONLY__
- return {}; // false
- #elif !defined __SUPPORT_SNAN__
- return ~(__x._M_data == __x._M_data);
- #elif defined __STDC_IEC_559__
- using _Ip = __int_for_sizeof_t<_Tp>;
- const auto __absn = __vector_bitcast<_Ip>(_SuperImpl::_S_abs(__x));
- const auto __infn
- = __vector_bitcast<_Ip>(__vector_broadcast<_Np>(__infinity_v<_Tp>));
- return __infn < __absn;
- #else
- #error "Not implemented: how to support SNaN but non-IEC559 floating-point?"
- #endif
- }
+ _GLIBCXX_SIMD_INTRINSIC static _MaskMember<_Tp>
+ _S_isnan([[maybe_unused]] _SimdWrapper<_Tp, _Np> __x)
+ {
+#if __FINITE_MATH_ONLY__
+ return {}; // false
+#elif !defined __SUPPORT_SNAN__
+ return ~(__x._M_data == __x._M_data);
+#elif defined __STDC_IEC_559__
+ using _Ip = __int_for_sizeof_t<_Tp>;
+ const auto __absn = __vector_bitcast<_Ip>(_SuperImpl::_S_abs(__x));
+ const auto __infn
+ = __vector_bitcast<_Ip>(__vector_broadcast<_Np>(__infinity_v<_Tp>));
+ return __infn < __absn;
+#else
+#error "Not implemented: how to support SNaN but non-IEC559 floating-point?"
+#endif
+ }
// _S_isfinite {{{3
template <typename _Tp, size_t _Np>
- _GLIBCXX_SIMD_INTRINSIC static _MaskMember<_Tp>
- _S_isfinite([[maybe_unused]] _SimdWrapper<_Tp, _Np> __x)
- {
- #if __FINITE_MATH_ONLY__
- using _UV = typename _MaskMember<_Tp>::_BuiltinType;
- _GLIBCXX_SIMD_USE_CONSTEXPR _UV __alltrue = ~_UV();
- return __alltrue;
- #else
- // if all exponent bits are set, __x is either inf or NaN
- using _Ip = __int_for_sizeof_t<_Tp>;
- const auto __absn = __vector_bitcast<_Ip>(_SuperImpl::_S_abs(__x));
- const auto __maxn
- = __vector_bitcast<_Ip>(__vector_broadcast<_Np>(__finite_max_v<_Tp>));
- return __absn <= __maxn;
- #endif
- }
+ _GLIBCXX_SIMD_INTRINSIC static _MaskMember<_Tp>
+ _S_isfinite([[maybe_unused]] _SimdWrapper<_Tp, _Np> __x)
+ {
+#if __FINITE_MATH_ONLY__
+ using _UV = typename _MaskMember<_Tp>::_BuiltinType;
+ _GLIBCXX_SIMD_USE_CONSTEXPR _UV __alltrue = ~_UV();
+ return __alltrue;
+#else
+ // if all exponent bits are set, __x is either inf or NaN
+ using _Ip = __int_for_sizeof_t<_Tp>;
+ const auto __absn = __vector_bitcast<_Ip>(_SuperImpl::_S_abs(__x));
+ const auto __maxn
+ = __vector_bitcast<_Ip>(__vector_broadcast<_Np>(__finite_max_v<_Tp>));
+ return __absn <= __maxn;
+#endif
+ }
// _S_isunordered {{{3
template <typename _Tp, size_t _Np>
- _GLIBCXX_SIMD_INTRINSIC static _MaskMember<_Tp>
- _S_isunordered(_SimdWrapper<_Tp, _Np> __x, _SimdWrapper<_Tp, _Np> __y)
- {
- return __or(_S_isnan(__x), _S_isnan(__y));
- }
+ _GLIBCXX_SIMD_INTRINSIC static _MaskMember<_Tp>
+ _S_isunordered(_SimdWrapper<_Tp, _Np> __x, _SimdWrapper<_Tp, _Np> __y)
+ { return __or(_S_isnan(__x), _S_isnan(__y)); }
// _S_signbit {{{3
template <typename _Tp, size_t _Np>
- _GLIBCXX_SIMD_INTRINSIC static _MaskMember<_Tp>
- _S_signbit(_SimdWrapper<_Tp, _Np> __x)
- {
- using _Ip = __int_for_sizeof_t<_Tp>;
- return __vector_bitcast<_Ip>(__x) < 0;
- // Arithmetic right shift (SRA) would also work (instead of compare), but
- // 64-bit SRA isn't available on x86 before AVX512. And in general,
- // compares are more likely to be efficient than SRA.
- }
+ _GLIBCXX_SIMD_INTRINSIC static _MaskMember<_Tp>
+ _S_signbit(_SimdWrapper<_Tp, _Np> __x)
+ {
+ using _Ip = __int_for_sizeof_t<_Tp>;
+ return __vector_bitcast<_Ip>(__x) < 0;
+ // Arithmetic right shift (SRA) would also work (instead of compare), but
+ // 64-bit SRA isn't available on x86 before AVX512. And in general,
+ // compares are more likely to be efficient than SRA.
+ }
// _S_isinf {{{3
template <typename _Tp, size_t _Np>
- _GLIBCXX_SIMD_INTRINSIC static _MaskMember<_Tp>
- _S_isinf([[maybe_unused]] _SimdWrapper<_Tp, _Np> __x)
- {
- #if __FINITE_MATH_ONLY__
- return {}; // false
- #else
- return _SuperImpl::template _S_equal_to<_Tp, _Np>(_SuperImpl::_S_abs(__x),
- __vector_broadcast<_Np>(
- __infinity_v<_Tp>));
- // alternative:
- // compare to inf using the corresponding integer type
- /*
- return
- __vector_bitcast<_Tp>(__vector_bitcast<__int_for_sizeof_t<_Tp>>(
- _S_abs(__x)._M_data)
- ==
- __vector_bitcast<__int_for_sizeof_t<_Tp>>(__vector_broadcast<_Np>(
- __infinity_v<_Tp>)));
- */
- #endif
- }
+ _GLIBCXX_SIMD_INTRINSIC static _MaskMember<_Tp>
+ _S_isinf([[maybe_unused]] _SimdWrapper<_Tp, _Np> __x)
+ {
+#if __FINITE_MATH_ONLY__
+ return {}; // false
+#else
+ return _SuperImpl::template _S_equal_to<_Tp, _Np>(_SuperImpl::_S_abs(__x),
+ __vector_broadcast<_Np>(
+ __infinity_v<_Tp>));
+ // alternative:
+ // compare to inf using the corresponding integer type
+ /*
+ return
+ __vector_bitcast<_Tp>(__vector_bitcast<__int_for_sizeof_t<_Tp>>(
+ _S_abs(__x)._M_data)
+ ==
+ __vector_bitcast<__int_for_sizeof_t<_Tp>>(__vector_broadcast<_Np>(
+ __infinity_v<_Tp>)));
+ */
+#endif
+ }
// _S_isnormal {{{3
template <typename _Tp, size_t _Np>
- _GLIBCXX_SIMD_INTRINSIC static _MaskMember<_Tp>
- _S_isnormal(_SimdWrapper<_Tp, _Np> __x)
- {
- using _Ip = __int_for_sizeof_t<_Tp>;
- const auto __absn = __vector_bitcast<_Ip>(_SuperImpl::_S_abs(__x));
- const auto __minn
- = __vector_bitcast<_Ip>(__vector_broadcast<_Np>(__norm_min_v<_Tp>));
- #if __FINITE_MATH_ONLY__
- return __absn >= __minn;
- #else
- const auto __maxn
- = __vector_bitcast<_Ip>(__vector_broadcast<_Np>(__finite_max_v<_Tp>));
- return __minn <= __absn && __absn <= __maxn;
- #endif
- }
+ _GLIBCXX_SIMD_INTRINSIC static _MaskMember<_Tp>
+ _S_isnormal(_SimdWrapper<_Tp, _Np> __x)
+ {
+ using _Ip = __int_for_sizeof_t<_Tp>;
+ const auto __absn = __vector_bitcast<_Ip>(_SuperImpl::_S_abs(__x));
+ const auto __minn
+ = __vector_bitcast<_Ip>(__vector_broadcast<_Np>(__norm_min_v<_Tp>));
+#if __FINITE_MATH_ONLY__
+ return __absn >= __minn;
+#else
+ const auto __maxn
+ = __vector_bitcast<_Ip>(__vector_broadcast<_Np>(__finite_max_v<_Tp>));
+ return __minn <= __absn && __absn <= __maxn;
+#endif
+ }
// _S_fpclassify {{{3
template <typename _Tp, size_t _Np>
- _GLIBCXX_SIMD_INTRINSIC static __fixed_size_storage_t<int, _Np>
- _S_fpclassify(_SimdWrapper<_Tp, _Np> __x)
- {
- using _I = __int_for_sizeof_t<_Tp>;
- const auto __xn
- = __vector_bitcast<_I>(__to_intrin(_SuperImpl::_S_abs(__x)));
- constexpr size_t _NI = sizeof(__xn) / sizeof(_I);
- _GLIBCXX_SIMD_USE_CONSTEXPR auto __minn
- = __vector_bitcast<_I>(__vector_broadcast<_NI>(__norm_min_v<_Tp>));
- _GLIBCXX_SIMD_USE_CONSTEXPR auto __infn
- = __vector_bitcast<_I>(__vector_broadcast<_NI>(__infinity_v<_Tp>));
-
- _GLIBCXX_SIMD_USE_CONSTEXPR auto __fp_normal
- = __vector_broadcast<_NI, _I>(FP_NORMAL);
- #if !__FINITE_MATH_ONLY__
- _GLIBCXX_SIMD_USE_CONSTEXPR auto __fp_nan
- = __vector_broadcast<_NI, _I>(FP_NAN);
- _GLIBCXX_SIMD_USE_CONSTEXPR auto __fp_infinite
- = __vector_broadcast<_NI, _I>(FP_INFINITE);
- #endif
- #ifndef __FAST_MATH__
- _GLIBCXX_SIMD_USE_CONSTEXPR auto __fp_subnormal
- = __vector_broadcast<_NI, _I>(FP_SUBNORMAL);
- #endif
- _GLIBCXX_SIMD_USE_CONSTEXPR auto __fp_zero
- = __vector_broadcast<_NI, _I>(FP_ZERO);
+ _GLIBCXX_SIMD_INTRINSIC static __fixed_size_storage_t<int, _Np>
+ _S_fpclassify(_SimdWrapper<_Tp, _Np> __x)
+ {
+ using _I = __int_for_sizeof_t<_Tp>;
+ const auto __xn
+ = __vector_bitcast<_I>(__to_intrin(_SuperImpl::_S_abs(__x)));
+ constexpr size_t _NI = sizeof(__xn) / sizeof(_I);
+ _GLIBCXX_SIMD_USE_CONSTEXPR auto __minn
+ = __vector_bitcast<_I>(__vector_broadcast<_NI>(__norm_min_v<_Tp>));
+ _GLIBCXX_SIMD_USE_CONSTEXPR auto __infn
+ = __vector_bitcast<_I>(__vector_broadcast<_NI>(__infinity_v<_Tp>));
+
+ _GLIBCXX_SIMD_USE_CONSTEXPR auto __fp_normal
+ = __vector_broadcast<_NI, _I>(FP_NORMAL);
+#if !__FINITE_MATH_ONLY__
+ _GLIBCXX_SIMD_USE_CONSTEXPR auto __fp_nan
+ = __vector_broadcast<_NI, _I>(FP_NAN);
+ _GLIBCXX_SIMD_USE_CONSTEXPR auto __fp_infinite
+ = __vector_broadcast<_NI, _I>(FP_INFINITE);
+#endif
+#ifndef __FAST_MATH__
+ _GLIBCXX_SIMD_USE_CONSTEXPR auto __fp_subnormal
+ = __vector_broadcast<_NI, _I>(FP_SUBNORMAL);
+#endif
+ _GLIBCXX_SIMD_USE_CONSTEXPR auto __fp_zero
+ = __vector_broadcast<_NI, _I>(FP_ZERO);
- __vector_type_t<_I, _NI>
- __tmp = __xn < __minn
+ __vector_type_t<_I, _NI>
+ __tmp = __xn < __minn
#ifdef __FAST_MATH__
- ? __fp_zero
+ ? __fp_zero
#else
- ? (__xn == 0 ? __fp_zero : __fp_subnormal)
+ ? (__xn == 0 ? __fp_zero : __fp_subnormal)
#endif
#if __FINITE_MATH_ONLY__
- : __fp_normal;
+ : __fp_normal;
#else
- : (__xn < __infn ? __fp_normal
- : (__xn == __infn ? __fp_infinite : __fp_nan));
+ : (__xn < __infn ? __fp_normal
+ : (__xn == __infn ? __fp_infinite : __fp_nan));
#endif
- if constexpr (sizeof(_I) == sizeof(int))
- {
- using _FixedInt = __fixed_size_storage_t<int, _Np>;
- const auto __as_int = __vector_bitcast<int, _Np>(__tmp);
- if constexpr (_FixedInt::_S_tuple_size == 1)
- return {__as_int};
- else if constexpr (_FixedInt::_S_tuple_size == 2
- && is_same_v<
- typename _FixedInt::_SecondType::_FirstAbi,
- simd_abi::scalar>)
- return {__extract<0, 2>(__as_int), __as_int[_Np - 1]};
- else if constexpr (_FixedInt::_S_tuple_size == 2)
- return {__extract<0, 2>(__as_int),
- __auto_bitcast(__extract<1, 2>(__as_int))};
- else
- __assert_unreachable<_Tp>();
- }
- else if constexpr (_Np == 2 && sizeof(_I) == 8
- && __fixed_size_storage_t<int, _Np>::_S_tuple_size == 2)
- {
- const auto __aslong = __vector_bitcast<_LLong>(__tmp);
- return {int(__aslong[0]), {int(__aslong[1])}};
- }
- #if _GLIBCXX_SIMD_X86INTRIN
- else if constexpr (sizeof(_Tp) == 8 && sizeof(__tmp) == 32
- && __fixed_size_storage_t<int, _Np>::_S_tuple_size == 1)
- return {_mm_packs_epi32(__to_intrin(__lo128(__tmp)),
- __to_intrin(__hi128(__tmp)))};
- else if constexpr (sizeof(_Tp) == 8 && sizeof(__tmp) == 64
- && __fixed_size_storage_t<int, _Np>::_S_tuple_size == 1)
- return {_mm512_cvtepi64_epi32(__to_intrin(__tmp))};
- #endif // _GLIBCXX_SIMD_X86INTRIN
- else if constexpr (__fixed_size_storage_t<int, _Np>::_S_tuple_size == 1)
- return {__call_with_subscripts<_Np>(__vector_bitcast<_LLong>(__tmp),
- [](auto... __l) _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA {
- return __make_wrapper<int>(__l...);
- })};
- else
- __assert_unreachable<_Tp>();
- }
+ if constexpr (sizeof(_I) == sizeof(int))
+ {
+ using _FixedInt = __fixed_size_storage_t<int, _Np>;
+ const auto __as_int = __vector_bitcast<int, _Np>(__tmp);
+ if constexpr (_FixedInt::_S_tuple_size == 1)
+ return {__as_int};
+ else if constexpr (_FixedInt::_S_tuple_size == 2
+ && is_same_v<
+ typename _FixedInt::_SecondType::_FirstAbi,
+ simd_abi::scalar>)
+ return {__extract<0, 2>(__as_int), __as_int[_Np - 1]};
+ else if constexpr (_FixedInt::_S_tuple_size == 2)
+ return {__extract<0, 2>(__as_int),
+ __auto_bitcast(__extract<1, 2>(__as_int))};
+ else
+ __assert_unreachable<_Tp>();
+ }
+ else if constexpr (_Np == 2 && sizeof(_I) == 8
+ && __fixed_size_storage_t<int, _Np>::_S_tuple_size == 2)
+ {
+ const auto __aslong = __vector_bitcast<_LLong>(__tmp);
+ return {int(__aslong[0]), {int(__aslong[1])}};
+ }
+#if _GLIBCXX_SIMD_X86INTRIN
+ else if constexpr (sizeof(_Tp) == 8 && sizeof(__tmp) == 32
+ && __fixed_size_storage_t<int, _Np>::_S_tuple_size == 1)
+ return {_mm_packs_epi32(__to_intrin(__lo128(__tmp)),
+ __to_intrin(__hi128(__tmp)))};
+ else if constexpr (sizeof(_Tp) == 8 && sizeof(__tmp) == 64
+ && __fixed_size_storage_t<int, _Np>::_S_tuple_size == 1)
+ return {_mm512_cvtepi64_epi32(__to_intrin(__tmp))};
+#endif // _GLIBCXX_SIMD_X86INTRIN
+ else if constexpr (__fixed_size_storage_t<int, _Np>::_S_tuple_size == 1)
+ return {__call_with_subscripts<_Np>(__vector_bitcast<_LLong>(__tmp),
+ [](auto... __l) _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA {
+ return __make_wrapper<int>(__l...);
+ })};
+ else
+ __assert_unreachable<_Tp>();
+ }
// _S_increment & _S_decrement{{{2
template <typename _Tp, size_t _Np>
@@ -2703,10 +2695,7 @@ struct _MaskImplBuiltin
template <typename _Tp>
_GLIBCXX_SIMD_INTRINSIC static constexpr _MaskMember<_Tp>
_S_broadcast(bool __x)
- {
- return __x ? _Abi::template _S_implicit_mask<_Tp>()
- : _MaskMember<_Tp>();
- }
+ { return __x ? _Abi::template _S_implicit_mask<_Tp>() : _MaskMember<_Tp>(); }
// }}}
// _S_load {{{
@@ -2808,8 +2797,8 @@ _S_masked_load(_SimdWrapper<_Tp, _Np> __merge,
// _S_store {{{2
template <typename _Tp, size_t _Np>
- _GLIBCXX_SIMD_INTRINSIC static void _S_store(_SimdWrapper<_Tp, _Np> __v,
- bool* __mem) noexcept
+ _GLIBCXX_SIMD_INTRINSIC static void
+ _S_store(_SimdWrapper<_Tp, _Np> __v, bool* __mem) noexcept
{
__execute_n_times<_Np>([&](auto __i) constexpr _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA {
__mem[__i] = __v[__i];
@@ -2832,21 +2821,17 @@ _S_masked_store(const _SimdWrapper<_Tp, _Np> __v, bool* __mem,
template <size_t _Np, typename _Tp>
_GLIBCXX_SIMD_INTRINSIC static _MaskMember<_Tp>
_S_from_bitmask(_SanitizedBitMask<_Np> __bits, _TypeTag<_Tp>)
- {
- return _SuperImpl::template _S_to_maskvector<_Tp, _S_size<_Tp>>(__bits);
- }
+ { return _SuperImpl::template _S_to_maskvector<_Tp, _S_size<_Tp>>(__bits); }
// logical and bitwise operators {{{2
template <typename _Tp, size_t _Np>
_GLIBCXX_SIMD_INTRINSIC static constexpr _SimdWrapper<_Tp, _Np>
- _S_logical_and(const _SimdWrapper<_Tp, _Np>& __x,
- const _SimdWrapper<_Tp, _Np>& __y)
+ _S_logical_and(const _SimdWrapper<_Tp, _Np>& __x, const _SimdWrapper<_Tp, _Np>& __y)
{ return __and(__x._M_data, __y._M_data); }
template <typename _Tp, size_t _Np>
_GLIBCXX_SIMD_INTRINSIC static constexpr _SimdWrapper<_Tp, _Np>
- _S_logical_or(const _SimdWrapper<_Tp, _Np>& __x,
- const _SimdWrapper<_Tp, _Np>& __y)
+ _S_logical_or(const _SimdWrapper<_Tp, _Np>& __x, const _SimdWrapper<_Tp, _Np>& __y)
{ return __or(__x._M_data, __y._M_data); }
template <typename _Tp, size_t _Np>
@@ -2862,26 +2847,23 @@ _S_bit_not(const _SimdWrapper<_Tp, _Np>& __x)
template <typename _Tp, size_t _Np>
_GLIBCXX_SIMD_INTRINSIC static constexpr _SimdWrapper<_Tp, _Np>
- _S_bit_and(const _SimdWrapper<_Tp, _Np>& __x,
- const _SimdWrapper<_Tp, _Np>& __y)
+ _S_bit_and(const _SimdWrapper<_Tp, _Np>& __x, const _SimdWrapper<_Tp, _Np>& __y)
{ return __and(__x._M_data, __y._M_data); }
template <typename _Tp, size_t _Np>
_GLIBCXX_SIMD_INTRINSIC static constexpr _SimdWrapper<_Tp, _Np>
- _S_bit_or(const _SimdWrapper<_Tp, _Np>& __x,
- const _SimdWrapper<_Tp, _Np>& __y)
+ _S_bit_or(const _SimdWrapper<_Tp, _Np>& __x, const _SimdWrapper<_Tp, _Np>& __y)
{ return __or(__x._M_data, __y._M_data); }
template <typename _Tp, size_t _Np>
_GLIBCXX_SIMD_INTRINSIC static constexpr _SimdWrapper<_Tp, _Np>
- _S_bit_xor(const _SimdWrapper<_Tp, _Np>& __x,
- const _SimdWrapper<_Tp, _Np>& __y)
+ _S_bit_xor(const _SimdWrapper<_Tp, _Np>& __x, const _SimdWrapper<_Tp, _Np>& __y)
{ return __xor(__x._M_data, __y._M_data); }
// smart_reference access {{{2
template <typename _Tp, size_t _Np>
- static constexpr void _S_set(_SimdWrapper<_Tp, _Np>& __k, int __i,
- bool __x) noexcept
+ static constexpr void
+ _S_set(_SimdWrapper<_Tp, _Np>& __k, int __i, bool __x) noexcept
{
if constexpr (is_same_v<_Tp, bool>)
__k._M_set(__i, __x);
@@ -2907,15 +2889,13 @@ _S_bit_xor(const _SimdWrapper<_Tp, _Np>& __x,
// _S_masked_assign{{{2
template <typename _Tp, size_t _Np>
_GLIBCXX_SIMD_INTRINSIC static void
- _S_masked_assign(_SimdWrapper<_Tp, _Np> __k,
- _SimdWrapper<_Tp, _Np>& __lhs,
+ _S_masked_assign(_SimdWrapper<_Tp, _Np> __k, _SimdWrapper<_Tp, _Np>& __lhs,
__type_identity_t<_SimdWrapper<_Tp, _Np>> __rhs)
{ __lhs = _CommonImpl::_S_blend(__k, __lhs, __rhs); }
template <typename _Tp, size_t _Np>
_GLIBCXX_SIMD_INTRINSIC static void
- _S_masked_assign(_SimdWrapper<_Tp, _Np> __k,
- _SimdWrapper<_Tp, _Np>& __lhs, bool __rhs)
+ _S_masked_assign(_SimdWrapper<_Tp, _Np> __k, _SimdWrapper<_Tp, _Np>& __lhs, bool __rhs)
{
if (__builtin_constant_p(__rhs))
{
@@ -2995,20 +2975,14 @@ _S_popcount(simd_mask<_Tp, _Abi> __k)
template <typename _Tp>
_GLIBCXX_SIMD_INTRINSIC static int
_S_find_first_set(simd_mask<_Tp, _Abi> __k)
- {
- return std::__countr_zero(
- _SuperImpl::_S_to_bits(__data(__k))._M_to_bits());
- }
+ { return std::__countr_zero( _SuperImpl::_S_to_bits(__data(__k))._M_to_bits()); }
// }}}
// _S_find_last_set {{{
template <typename _Tp>
_GLIBCXX_SIMD_INTRINSIC static int
_S_find_last_set(simd_mask<_Tp, _Abi> __k)
- {
- return std::__bit_width(
- _SuperImpl::_S_to_bits(__data(__k))._M_to_bits()) - 1;
- }
+ { return std::__bit_width( _SuperImpl::_S_to_bits(__data(__k))._M_to_bits()) - 1; }
// }}}
};
diff --git a/libstdc++-v3/include/experimental/bits/simd_fixed_size.h b/libstdc++-v3/include/experimental/bits/simd_fixed_size.h
index 88a9b27e359..123e714b528 100644
--- a/libstdc++-v3/include/experimental/bits/simd_fixed_size.h
+++ b/libstdc++-v3/include/experimental/bits/simd_fixed_size.h
@@ -55,10 +55,7 @@ struct __simd_tuple_element<0, _SimdTuple<_Tp, _A0, _As...>>
template <size_t _I, typename _Tp, typename _A0, typename... _As>
struct __simd_tuple_element<_I, _SimdTuple<_Tp, _A0, _As...>>
- {
- using type =
- typename __simd_tuple_element<_I - 1, _SimdTuple<_Tp, _As...>>::type;
- };
+ { using type = typename __simd_tuple_element<_I - 1, _SimdTuple<_Tp, _As...>>::type; };
template <size_t _I, typename _Tp>
using __simd_tuple_element_t = typename __simd_tuple_element<_I, _Tp>::type;
@@ -80,10 +77,8 @@ __simd_tuple_concat(const _SimdTuple<_Tp, _A0s...>& __left,
}
template <typename _Tp, typename _A10, typename... _A1s>
- _GLIBCXX_SIMD_INTRINSIC constexpr _SimdTuple<_Tp, simd_abi::scalar, _A10,
- _A1s...>
- __simd_tuple_concat(const _Tp& __left,
- const _SimdTuple<_Tp, _A10, _A1s...>& __right)
+ _GLIBCXX_SIMD_INTRINSIC constexpr _SimdTuple<_Tp, simd_abi::scalar, _A10, _A1s...>
+ __simd_tuple_concat(const _Tp& __left, const _SimdTuple<_Tp, _A10, _A1s...>& __right)
{ return {__left, __right}; }
// }}}
@@ -112,37 +107,29 @@ struct __as_simd_tuple
template <typename _Tp, typename _A0, typename... _Abis>
_GLIBCXX_SIMD_INTRINSIC constexpr simd<_Tp, _A0>
- __simd_tuple_get_impl(__as_simd, const _SimdTuple<_Tp, _A0, _Abis...>& __t,
- _SizeConstant<0>)
+ __simd_tuple_get_impl(__as_simd, const _SimdTuple<_Tp, _A0, _Abis...>& __t, _SizeConstant<0>)
{ return {__private_init, __t.first}; }
template <typename _Tp, typename _A0, typename... _Abis>
_GLIBCXX_SIMD_INTRINSIC constexpr const auto&
- __simd_tuple_get_impl(__as_simd_tuple,
- const _SimdTuple<_Tp, _A0, _Abis...>& __t,
+ __simd_tuple_get_impl(__as_simd_tuple, const _SimdTuple<_Tp, _A0, _Abis...>& __t,
_SizeConstant<0>)
{ return __t.first; }
template <typename _Tp, typename _A0, typename... _Abis>
_GLIBCXX_SIMD_INTRINSIC constexpr auto&
- __simd_tuple_get_impl(__as_simd_tuple, _SimdTuple<_Tp, _A0, _Abis...>& __t,
- _SizeConstant<0>)
+ __simd_tuple_get_impl(__as_simd_tuple, _SimdTuple<_Tp, _A0, _Abis...>& __t, _SizeConstant<0>)
{ return __t.first; }
template <typename _R, size_t _Np, typename _Tp, typename... _Abis>
_GLIBCXX_SIMD_INTRINSIC constexpr auto
- __simd_tuple_get_impl(_R, const _SimdTuple<_Tp, _Abis...>& __t,
- _SizeConstant<_Np>)
+ __simd_tuple_get_impl(_R, const _SimdTuple<_Tp, _Abis...>& __t, _SizeConstant<_Np>)
{ return __simd_tuple_get_impl(_R(), __t.second, _SizeConstant<_Np - 1>()); }
template <size_t _Np, typename _Tp, typename... _Abis>
_GLIBCXX_SIMD_INTRINSIC constexpr auto&
- __simd_tuple_get_impl(__as_simd_tuple, _SimdTuple<_Tp, _Abis...>& __t,
- _SizeConstant<_Np>)
- {
- return __simd_tuple_get_impl(__as_simd_tuple(), __t.second,
- _SizeConstant<_Np - 1>());
- }
+ __simd_tuple_get_impl(__as_simd_tuple, _SimdTuple<_Tp, _Abis...>& __t, _SizeConstant<_Np>)
+ { return __simd_tuple_get_impl(__as_simd_tuple(), __t.second, _SizeConstant<_Np - 1>()); }
template <size_t _Np, typename _Tp, typename... _Abis>
_GLIBCXX_SIMD_INTRINSIC constexpr auto
@@ -154,16 +141,12 @@ __get_simd_at(const _SimdTuple<_Tp, _Abis...>& __t)
template <size_t _Np, typename _Tp, typename... _Abis>
_GLIBCXX_SIMD_INTRINSIC constexpr auto
__get_tuple_at(const _SimdTuple<_Tp, _Abis...>& __t)
- {
- return __simd_tuple_get_impl(__as_simd_tuple(), __t, _SizeConstant<_Np>());
- }
+ { return __simd_tuple_get_impl(__as_simd_tuple(), __t, _SizeConstant<_Np>()); }
template <size_t _Np, typename _Tp, typename... _Abis>
_GLIBCXX_SIMD_INTRINSIC constexpr auto&
__get_tuple_at(_SimdTuple<_Tp, _Abis...>& __t)
- {
- return __simd_tuple_get_impl(__as_simd_tuple(), __t, _SizeConstant<_Np>());
- }
+ { return __simd_tuple_get_impl(__as_simd_tuple(), __t, _SizeConstant<_Np>()); }
// __tuple_element_meta {{{1
template <typename _Tp, typename _Abi, size_t _Offset>
@@ -213,17 +196,13 @@ struct _WithOffset
{
static inline constexpr size_t _S_offset = _Offset;
- _GLIBCXX_SIMD_INTRINSIC char* _M_as_charptr()
- {
- return reinterpret_cast<char*>(this)
- + _S_offset * sizeof(typename _Base::value_type);
- }
+ _GLIBCXX_SIMD_INTRINSIC char*
+ _M_as_charptr()
+ { return reinterpret_cast<char*>(this) + _S_offset * sizeof(typename _Base::value_type); }
- _GLIBCXX_SIMD_INTRINSIC const char* _M_as_charptr() const
- {
- return reinterpret_cast<const char*>(this)
- + _S_offset * sizeof(typename _Base::value_type);
- }
+ _GLIBCXX_SIMD_INTRINSIC const char*
+ _M_as_charptr() const
+ { return reinterpret_cast<const char*>(this) + _S_offset * sizeof(typename _Base::value_type); }
};
// make _WithOffset<_WithOffset> ill-formed to use:
@@ -240,19 +219,13 @@ __add_offset(_Tp& __base)
_GLIBCXX_SIMD_INTRINSIC
decltype(auto)
__add_offset(const _Tp& __base)
- {
- return static_cast<const _WithOffset<_Offset, __remove_cvref_t<_Tp>>&>(
- __base);
- }
+ { return static_cast<const _WithOffset<_Offset, __remove_cvref_t<_Tp>>&>(__base); }
template <size_t _Offset, size_t _ExistingOffset, typename _Tp>
_GLIBCXX_SIMD_INTRINSIC
decltype(auto)
__add_offset(_WithOffset<_ExistingOffset, _Tp>& __base)
- {
- return static_cast<_WithOffset<_Offset + _ExistingOffset, _Tp>&>(
- static_cast<_Tp&>(__base));
- }
+ { return static_cast<_WithOffset<_Offset + _ExistingOffset, _Tp>&>(static_cast<_Tp&>(__base)); }
template <size_t _Offset, size_t _ExistingOffset, typename _Tp>
_GLIBCXX_SIMD_INTRINSIC
@@ -298,7 +271,8 @@ struct _SimdTupleData
_SecondType second;
_GLIBCXX_SIMD_INTRINSIC
- constexpr bool _M_is_constprop() const
+ constexpr bool
+ _M_is_constprop() const
{
if constexpr (is_class_v<_FirstType>)
return first._M_is_constprop() && second._M_is_constprop();
@@ -314,7 +288,8 @@ struct _SimdTupleData<_FirstType, _SimdTuple<_Tp>>
static constexpr _SimdTuple<_Tp> second = {};
_GLIBCXX_SIMD_INTRINSIC
- constexpr bool _M_is_constprop() const
+ constexpr bool
+ _M_is_constprop() const
{
if constexpr (is_class_v<_FirstType>)
return first._M_is_constprop();
@@ -353,25 +328,31 @@ struct _SimdTuple<_Tp, _Abi0, _Abis...>
= default;
template <typename _Up>
- _GLIBCXX_SIMD_INTRINSIC constexpr _SimdTuple(_Up&& __x)
+ _GLIBCXX_SIMD_INTRINSIC constexpr
+ _SimdTuple(_Up&& __x)
: _Base{static_cast<_Up&&>(__x)} {}
template <typename _Up, typename _Up2>
- _GLIBCXX_SIMD_INTRINSIC constexpr _SimdTuple(_Up&& __x, _Up2&& __y)
+ _GLIBCXX_SIMD_INTRINSIC constexpr
+ _SimdTuple(_Up&& __x, _Up2&& __y)
: _Base{static_cast<_Up&&>(__x), static_cast<_Up2&&>(__y)} {}
template <typename _Up>
- _GLIBCXX_SIMD_INTRINSIC constexpr _SimdTuple(_Up&& __x, _SimdTuple<_Tp>)
+ _GLIBCXX_SIMD_INTRINSIC constexpr
+ _SimdTuple(_Up&& __x, _SimdTuple<_Tp>)
: _Base{static_cast<_Up&&>(__x)} {}
- _GLIBCXX_SIMD_INTRINSIC char* _M_as_charptr()
+ _GLIBCXX_SIMD_INTRINSIC char*
+ _M_as_charptr()
{ return reinterpret_cast<char*>(this); }
- _GLIBCXX_SIMD_INTRINSIC const char* _M_as_charptr() const
+ _GLIBCXX_SIMD_INTRINSIC const char*
+ _M_as_charptr() const
{ return reinterpret_cast<const char*>(this); }
template <size_t _Np>
- _GLIBCXX_SIMD_INTRINSIC constexpr auto& _M_at()
+ _GLIBCXX_SIMD_INTRINSIC constexpr auto&
+ _M_at()
{
if constexpr (_Np == 0)
return first;
@@ -380,7 +361,8 @@ struct _SimdTuple<_Tp, _Abi0, _Abis...>
}
template <size_t _Np>
- _GLIBCXX_SIMD_INTRINSIC constexpr const auto& _M_at() const
+ _GLIBCXX_SIMD_INTRINSIC constexpr const auto&
+ _M_at() const
{
if constexpr (_Np == 0)
return first;
@@ -389,7 +371,8 @@ struct _SimdTuple<_Tp, _Abi0, _Abis...>
}
template <size_t _Np>
- _GLIBCXX_SIMD_INTRINSIC constexpr auto _M_simd_at() const
+ _GLIBCXX_SIMD_INTRINSIC constexpr auto
+ _M_simd_at() const
{
if constexpr (_Np == 0)
return simd<_Tp, _Abi0>(__private_init, first);
@@ -552,8 +535,8 @@ _M_assign_front(const _SimdTuple<_Tp, _As...>& __x) &
}
template <typename _R = _Tp, typename _Fp, typename... _More>
- _GLIBCXX_SIMD_INTRINSIC auto _M_apply_r(_Fp&& __fun,
- const _More&... __more) const
+ _GLIBCXX_SIMD_INTRINSIC auto
+ _M_apply_r(_Fp&& __fun, const _More&... __more) const
{
auto&& __first = __fun(__tuple_element_meta<_Tp, _Abi0, 0>(), first,
__more.first...);
@@ -590,8 +573,8 @@ _M_assign_front(const _SimdTuple<_Tp, _As...>& __x) &
return second[integral_constant<_Up, _I - simd_size_v<_Tp, _Abi0>>()];
}
- _GLIBCXX_SIMD_INTRINSIC
- _Tp operator[](size_t __i) const noexcept
+ _GLIBCXX_SIMD_INTRINSIC _Tp
+ operator[](size_t __i) const noexcept
{
if constexpr (_S_tuple_size == 1)
return _M_subscript_read(__i);
@@ -613,8 +596,8 @@ _M_assign_front(const _SimdTuple<_Tp, _As...>& __x) &
}
}
- _GLIBCXX_SIMD_INTRINSIC
- void _M_set(size_t __i, _Tp __val) noexcept
+ _GLIBCXX_SIMD_INTRINSIC void
+ _M_set(size_t __i, _Tp __val) noexcept
{
if constexpr (_S_tuple_size == 1)
return _M_subscript_write(__i, __val);
@@ -633,8 +616,8 @@ _M_assign_front(const _SimdTuple<_Tp, _As...>& __x) &
private:
// _M_subscript_read/_write {{{
- _GLIBCXX_SIMD_INTRINSIC
- _Tp _M_subscript_read([[maybe_unused]] size_t __i) const noexcept
+ _GLIBCXX_SIMD_INTRINSIC _Tp
+ _M_subscript_read([[maybe_unused]] size_t __i) const noexcept
{
if constexpr (__is_vectorizable_v<_FirstType>)
return first;
@@ -642,8 +625,8 @@ _M_assign_front(const _SimdTuple<_Tp, _As...>& __x) &
return first[__i];
}
- _GLIBCXX_SIMD_INTRINSIC
- void _M_subscript_write([[maybe_unused]] size_t __i, _Tp __y) noexcept
+ _GLIBCXX_SIMD_INTRINSIC void
+ _M_subscript_write([[maybe_unused]] size_t __i, _Tp __y) noexcept
{
if constexpr (__is_vectorizable_v<_FirstType>)
first = __y;
@@ -687,8 +670,7 @@ __make_simd_tuple(
size_t _Offset = 0, // skip this many elements in __from0
typename _R = __fixed_size_storage_t<_Tp, _Np>, typename _V0,
typename _V0VT = _VectorTraits<_V0>, typename... _VX>
- _GLIBCXX_SIMD_INTRINSIC _R constexpr __to_simd_tuple(const _V0 __from0,
- const _VX... __fromX)
+ _GLIBCXX_SIMD_INTRINSIC _R constexpr __to_simd_tuple(const _V0 __from0, const _VX... __fromX)
{
static_assert(is_same_v<typename _V0VT::value_type, _Tp>);
static_assert(_Offset < _V0VT::_S_full_size);
@@ -900,11 +882,8 @@ __for_each(_SimdTuple<_Tp, _A0, _A1, _As...>& __t, _Fp&& __fun)
// __for_each(_SimdTuple &, const _SimdTuple &, Fun) {{{1
template <size_t _Offset = 0, typename _Tp, typename _A0, typename _Fp>
_GLIBCXX_SIMD_INTRINSIC constexpr void
- __for_each(_SimdTuple<_Tp, _A0>& __a, const _SimdTuple<_Tp, _A0>& __b,
- _Fp&& __fun)
- {
- static_cast<_Fp&&>(__fun)(__make_meta<_Offset>(__a), __a.first, __b.first);
- }
+ __for_each(_SimdTuple<_Tp, _A0>& __a, const _SimdTuple<_Tp, _A0>& __b, _Fp&& __fun)
+ { static_cast<_Fp&&>(__fun)(__make_meta<_Offset>(__a), __a.first, __b.first); }
template <size_t _Offset = 0, typename _Tp, typename _A0, typename _A1,
typename... _As, typename _Fp>
@@ -920,11 +899,8 @@ __for_each(_SimdTuple<_Tp, _A0, _A1, _As...>& __a,
// __for_each(const _SimdTuple &, const _SimdTuple &, Fun) {{{1
template <size_t _Offset = 0, typename _Tp, typename _A0, typename _Fp>
_GLIBCXX_SIMD_INTRINSIC constexpr void
- __for_each(const _SimdTuple<_Tp, _A0>& __a, const _SimdTuple<_Tp, _A0>& __b,
- _Fp&& __fun)
- {
- static_cast<_Fp&&>(__fun)(__make_meta<_Offset>(__a), __a.first, __b.first);
- }
+ __for_each(const _SimdTuple<_Tp, _A0>& __a, const _SimdTuple<_Tp, _A0>& __b, _Fp&& __fun)
+ { static_cast<_Fp&&>(__fun)(__make_meta<_Offset>(__a), __a.first, __b.first); }
template <size_t _Offset = 0, typename _Tp, typename _A0, typename _A1,
typename... _As, typename _Fp>
@@ -939,8 +915,7 @@ __for_each(const _SimdTuple<_Tp, _A0, _A1, _As...>& __a,
// }}}1
// __extract_part(_SimdTuple) {{{
-template <int _Index, int _Total, int _Combine, typename _Tp, typename _A0,
- typename... _As>
+template <int _Index, int _Total, int _Combine, typename _Tp, typename _A0, typename... _As>
_GLIBCXX_SIMD_INTRINSIC auto // __vector_type_t or _SimdTuple
__extract_part(const _SimdTuple<_Tp, _A0, _As...>& __x)
{
@@ -1062,8 +1037,8 @@ struct __autocvt_to_simd
return &_M_data;
}
- _GLIBCXX_SIMD_INTRINSIC
- constexpr __autocvt_to_simd(_Tp dd) : _M_data(dd) {}
+ _GLIBCXX_SIMD_INTRINSIC constexpr
+ __autocvt_to_simd(_Tp dd) : _M_data(dd) {}
template <typename _Abi>
_GLIBCXX_SIMD_INTRINSIC
@@ -1073,18 +1048,12 @@ struct __autocvt_to_simd
template <typename _Abi>
_GLIBCXX_SIMD_INTRINSIC
operator simd<typename _TT::value_type, _Abi>&()
- {
- return *reinterpret_cast<simd<typename _TT::value_type, _Abi>*>(
- &_M_data);
- }
+ { return *reinterpret_cast<simd<typename _TT::value_type, _Abi>*>(&_M_data); }
template <typename _Abi>
_GLIBCXX_SIMD_INTRINSIC
operator simd<typename _TT::value_type, _Abi>*()
- {
- return reinterpret_cast<simd<typename _TT::value_type, _Abi>*>(
- &_M_data);
- }
+ { return reinterpret_cast<simd<typename _TT::value_type, _Abi>*>(&_M_data); }
};
template <typename _Tp>
@@ -1197,12 +1166,12 @@ struct _SimdBase
_SimdBase(const _SimdBase&) {}
_SimdBase() = default;
- _GLIBCXX_SIMD_ALWAYS_INLINE
- explicit operator const _SimdMember &() const
+ _GLIBCXX_SIMD_ALWAYS_INLINE explicit
+ operator const _SimdMember &() const
{ return static_cast<const simd<_Tp, _Fixed>*>(this)->_M_data; }
- _GLIBCXX_SIMD_ALWAYS_INLINE
- explicit operator array<_Tp, _Np>() const
+ _GLIBCXX_SIMD_ALWAYS_INLINE explicit
+ operator array<_Tp, _Np>() const
{
array<_Tp, _Np> __r;
// _SimdMember can be larger because of higher alignment
@@ -1224,10 +1193,12 @@ struct _SimdCastType
{
_GLIBCXX_SIMD_ALWAYS_INLINE
_SimdCastType(const array<_Tp, _Np>&);
+
_GLIBCXX_SIMD_ALWAYS_INLINE
_SimdCastType(const _SimdMember& dd) : _M_data(dd) {}
- _GLIBCXX_SIMD_ALWAYS_INLINE
- explicit operator const _SimdMember &() const { return _M_data; }
+
+ _GLIBCXX_SIMD_ALWAYS_INLINE explicit
+ operator const _SimdMember &() const { return _M_data; }
private:
const _SimdMember& _M_data;
@@ -1466,8 +1437,7 @@ __for_each(
// _S_min, _S_max {{{2
template <typename _Tp, typename... _As>
_GLIBCXX_SIMD_INTRINSIC static constexpr _SimdTuple<_Tp, _As...>
- _S_min(const _SimdTuple<_Tp, _As...>& __a,
- const _SimdTuple<_Tp, _As...>& __b)
+ _S_min(const _SimdTuple<_Tp, _As...>& __a, const _SimdTuple<_Tp, _As...>& __b)
{
return __a._M_apply_per_chunk(
[](auto __impl, auto __aa, auto __bb) constexpr _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA {
@@ -1478,8 +1448,7 @@ _S_min(const _SimdTuple<_Tp, _As...>& __a,
template <typename _Tp, typename... _As>
_GLIBCXX_SIMD_INTRINSIC static constexpr _SimdTuple<_Tp, _As...>
- _S_max(const _SimdTuple<_Tp, _As...>& __a,
- const _SimdTuple<_Tp, _As...>& __b)
+ _S_max(const _SimdTuple<_Tp, _As...>& __a, const _SimdTuple<_Tp, _As...>& __b)
{
return __a._M_apply_per_chunk(
[](auto __impl, auto __aa, auto __bb) constexpr _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA {
@@ -1692,7 +1661,7 @@ _S_frexp(const _SimdTuple<_Tp, _As...>& __x,
#define _GLIBCXX_SIMD_TEST_ON_TUPLE_(name_) \
template <typename _Tp, typename... _As> \
static inline _MaskMember \
- _S_##name_(const _SimdTuple<_Tp, _As...>& __x) noexcept \
+ _S_##name_(const _SimdTuple<_Tp, _As...>& __x) noexcept \
{ \
return _M_test([] (auto __impl, auto __xx) _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA { \
return __impl._S_##name_(__xx); \
@@ -1754,8 +1723,8 @@ __for_each(
// smart_reference access {{{2
template <typename _Tp, typename... _As, typename _Up>
- _GLIBCXX_SIMD_INTRINSIC static void _S_set(_SimdTuple<_Tp, _As...>& __v,
- int __i, _Up&& __x) noexcept
+ _GLIBCXX_SIMD_INTRINSIC static void
+ _S_set(_SimdTuple<_Tp, _As...>& __v, int __i, _Up&& __x) noexcept
{ __v._M_set(__i, static_cast<_Up&&>(__x)); }
// _S_masked_assign {{{2
@@ -1789,10 +1758,9 @@ __for_each(
// _S_masked_cassign {{{2
template <typename _Op, typename _Tp, typename... _As>
- static inline void _S_masked_cassign(const _MaskMember __bits,
- _SimdTuple<_Tp, _As...>& __lhs,
- const _SimdTuple<_Tp, _As...>& __rhs,
- _Op __op)
+ static inline void
+ _S_masked_cassign(const _MaskMember __bits, _SimdTuple<_Tp, _As...>& __lhs,
+ const _SimdTuple<_Tp, _As...>& __rhs, _Op __op)
{
__for_each(__lhs, __rhs,
[&](auto __meta, auto& __native_lhs, auto __native_rhs)
@@ -1806,9 +1774,9 @@ __for_each(
// Optimization for the case where the RHS is a scalar. No need to broadcast
// the scalar to a simd first.
template <typename _Op, typename _Tp, typename... _As>
- static inline void _S_masked_cassign(const _MaskMember __bits,
- _SimdTuple<_Tp, _As...>& __lhs,
- const _Tp& __rhs, _Op __op)
+ static inline void
+ _S_masked_cassign(const _MaskMember __bits, _SimdTuple<_Tp, _As...>& __lhs,
+ const _Tp& __rhs, _Op __op)
{
__for_each(
__lhs, [&](auto __meta, auto& __native_lhs) constexpr _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA {
@@ -1906,7 +1874,8 @@ _S_convert(simd_mask<_Up, _UAbi> __x)
{ return __bits; }
// _S_load {{{2
- static inline _MaskMember _S_load(const bool* __mem) noexcept
+ static inline _MaskMember
+ _S_load(const bool* __mem) noexcept
{
// TODO: _UChar is not necessarily the best type to use here. For smaller
// _Np _UShort, _UInt, _ULLong, float, and double can be more efficient.
@@ -1921,9 +1890,8 @@ _S_convert(simd_mask<_Up, _UAbi> __x)
}
// _S_masked_load {{{2
- static inline _MaskMember _S_masked_load(_MaskMember __merge,
- _MaskMember __mask,
- const bool* __mem) noexcept
+ static inline _MaskMember
+ _S_masked_load(_MaskMember __merge, _MaskMember __mask, const bool* __mem) noexcept
{
_BitOps::_S_bit_iteration(__mask.to_ullong(),
[&](auto __i) _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA {
@@ -1933,8 +1901,8 @@ _S_convert(simd_mask<_Up, _UAbi> __x)
}
// _S_store {{{2
- static inline void _S_store(const _MaskMember __bitmask,
- bool* __mem) noexcept
+ static inline void
+ _S_store(const _MaskMember __bitmask, bool* __mem) noexcept
{
if constexpr (_Np == 1)
__mem[0] = __bitmask[0];
@@ -1943,8 +1911,8 @@ _S_convert(simd_mask<_Up, _UAbi> __x)
}
// _S_masked_store {{{2
- static inline void _S_masked_store(const _MaskMember __v, bool* __mem,
- const _MaskMember __k) noexcept
+ static inline void
+ _S_masked_store(const _MaskMember __v, bool* __mem, const _MaskMember __k) noexcept
{
_BitOps::_S_bit_iteration(
__k, [&](auto __i) _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA { __mem[__i] = __v[__i]; });
@@ -1976,20 +1944,18 @@ _S_bit_xor(const _MaskMember& __x, const _MaskMember& __y) noexcept
{ return __x ^ __y; }
// smart_reference access {{{2
- _GLIBCXX_SIMD_INTRINSIC static void _S_set(_MaskMember& __k, int __i,
- bool __x) noexcept
+ _GLIBCXX_SIMD_INTRINSIC static void
+ _S_set(_MaskMember& __k, int __i, bool __x) noexcept
{ __k.set(__i, __x); }
// _S_masked_assign {{{2
_GLIBCXX_SIMD_INTRINSIC static void
- _S_masked_assign(const _MaskMember __k, _MaskMember& __lhs,
- const _MaskMember __rhs)
+ _S_masked_assign(const _MaskMember __k, _MaskMember& __lhs, const _MaskMember __rhs)
{ __lhs = (__lhs & ~__k) | (__rhs & __k); }
// Optimization for the case where the RHS is a scalar.
- _GLIBCXX_SIMD_INTRINSIC static void _S_masked_assign(const _MaskMember __k,
- _MaskMember& __lhs,
- const bool __rhs)
+ _GLIBCXX_SIMD_INTRINSIC static void
+ _S_masked_assign(const _MaskMember __k, _MaskMember& __lhs, const bool __rhs)
{
if (__rhs)
__lhs |= __k;
@@ -2000,19 +1966,22 @@ _S_masked_assign(const _MaskMember __k, _MaskMember& __lhs,
// }}}2
// _S_all_of {{{
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static bool _S_all_of(simd_mask<_Tp, _Abi> __k)
+ _GLIBCXX_SIMD_INTRINSIC static bool
+ _S_all_of(simd_mask<_Tp, _Abi> __k)
{ return __data(__k).all(); }
// }}}
// _S_any_of {{{
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static bool _S_any_of(simd_mask<_Tp, _Abi> __k)
+ _GLIBCXX_SIMD_INTRINSIC static bool
+ _S_any_of(simd_mask<_Tp, _Abi> __k)
{ return __data(__k).any(); }
// }}}
// _S_none_of {{{
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static bool _S_none_of(simd_mask<_Tp, _Abi> __k)
+ _GLIBCXX_SIMD_INTRINSIC static bool
+ _S_none_of(simd_mask<_Tp, _Abi> __k)
{ return __data(__k).none(); }
// }}}
@@ -2030,7 +1999,8 @@ _S_masked_assign(const _MaskMember __k, _MaskMember& __lhs,
// }}}
// _S_popcount {{{
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static int _S_popcount(simd_mask<_Tp, _Abi> __k)
+ _GLIBCXX_SIMD_INTRINSIC static int
+ _S_popcount(simd_mask<_Tp, _Abi> __k)
{ return __data(__k).count(); }
// }}}
diff --git a/libstdc++-v3/include/experimental/bits/simd_neon.h b/libstdc++-v3/include/experimental/bits/simd_neon.h
index 7e4cb17b205..637b121b130 100644
--- a/libstdc++-v3/include/experimental/bits/simd_neon.h
+++ b/libstdc++-v3/include/experimental/bits/simd_neon.h
@@ -134,7 +134,8 @@ _S_reduce(simd<_Tp, _Abi> __x, _BinaryOperation&& __binary_op)
// math {{{
// _S_sqrt {{{
template <typename _Tp, typename _TVT = _VectorTraits<_Tp>>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_sqrt(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_sqrt(_Tp __x)
{
if constexpr (__have_neon_a64)
{
@@ -157,7 +158,8 @@ _S_reduce(simd<_Tp, _Abi> __x, _BinaryOperation&& __binary_op)
// }}}
// _S_trunc {{{
template <typename _TW, typename _TVT = _VectorTraits<_TW>>
- _GLIBCXX_SIMD_INTRINSIC static _TW _S_trunc(_TW __x)
+ _GLIBCXX_SIMD_INTRINSIC static _TW
+ _S_trunc(_TW __x)
{
using _Tp = typename _TVT::value_type;
if constexpr (__have_neon_a32)
@@ -216,7 +218,8 @@ _S_round(_SimdWrapper<_Tp, _Np> __x)
// }}}
// _S_floor {{{
template <typename _Tp, typename _TVT = _VectorTraits<_Tp>>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_floor(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_floor(_Tp __x)
{
if constexpr (__have_neon_a32)
{
@@ -239,7 +242,8 @@ _S_round(_SimdWrapper<_Tp, _Np> __x)
// }}}
// _S_ceil {{{
template <typename _Tp, typename _TVT = _VectorTraits<_Tp>>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_ceil(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_ceil(_Tp __x)
{
if constexpr (__have_neon_a32)
{
@@ -400,7 +404,8 @@ struct _MaskImplNeon
// _S_all_of {{{
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static bool _S_all_of(simd_mask<_Tp, _Abi> __k)
+ _GLIBCXX_SIMD_INTRINSIC static bool
+ _S_all_of(simd_mask<_Tp, _Abi> __k)
{
const auto __kk
= __vector_bitcast<char>(__k._M_data)
@@ -419,7 +424,8 @@ struct _MaskImplNeon
// }}}
// _S_any_of {{{
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static bool _S_any_of(simd_mask<_Tp, _Abi> __k)
+ _GLIBCXX_SIMD_INTRINSIC static bool
+ _S_any_of(simd_mask<_Tp, _Abi> __k)
{
const auto __kk
= __vector_bitcast<char>(__k._M_data)
@@ -438,7 +444,8 @@ struct _MaskImplNeon
// }}}
// _S_none_of {{{
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static bool _S_none_of(simd_mask<_Tp, _Abi> __k)
+ _GLIBCXX_SIMD_INTRINSIC static bool
+ _S_none_of(simd_mask<_Tp, _Abi> __k)
{
const auto __kk = _Abi::_S_masked(__k._M_data);
if constexpr (sizeof(__k) == 16)
@@ -472,7 +479,8 @@ struct _MaskImplNeon
// }}}
// _S_popcount {{{
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static int _S_popcount(simd_mask<_Tp, _Abi> __k)
+ _GLIBCXX_SIMD_INTRINSIC static int
+ _S_popcount(simd_mask<_Tp, _Abi> __k)
{
if constexpr (sizeof(_Tp) == 1)
{
diff --git a/libstdc++-v3/include/experimental/bits/simd_ppc.h b/libstdc++-v3/include/experimental/bits/simd_ppc.h
index 0d56912894a..eca1b34241b 100644
--- a/libstdc++-v3/include/experimental/bits/simd_ppc.h
+++ b/libstdc++-v3/include/experimental/bits/simd_ppc.h
@@ -124,7 +124,8 @@ struct _MaskImplPpc
// _S_popcount {{{
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static int _S_popcount(simd_mask<_Tp, _Abi> __k)
+ _GLIBCXX_SIMD_INTRINSIC static int
+ _S_popcount(simd_mask<_Tp, _Abi> __k)
{
const auto __kv = __as_vector(__k);
if constexpr (__have_power10vec)
diff --git a/libstdc++-v3/include/experimental/bits/simd_scalar.h b/libstdc++-v3/include/experimental/bits/simd_scalar.h
index f9ae70430db..1a1cc46fbe0 100644
--- a/libstdc++-v3/include/experimental/bits/simd_scalar.h
+++ b/libstdc++-v3/include/experimental/bits/simd_scalar.h
@@ -74,7 +74,8 @@ struct _IsValid
template <typename _Tp>
static constexpr bool _S_is_valid_v = _IsValid<_Tp>::value;
- _GLIBCXX_SIMD_INTRINSIC static constexpr bool _S_masked(bool __x)
+ _GLIBCXX_SIMD_INTRINSIC static constexpr bool
+ _S_masked(bool __x)
{ return __x; }
using _CommonImpl = _CommonImplScalar;
@@ -110,7 +111,8 @@ struct _CommonImplScalar
{
// _S_store {{{
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static void _S_store(_Tp __x, void* __addr)
+ _GLIBCXX_SIMD_INTRINSIC static void
+ _S_store(_Tp __x, void* __addr)
{ __builtin_memcpy(__addr, &__x, sizeof(_Tp)); }
// }}}
@@ -138,26 +140,26 @@ struct _SimdImplScalar
// _S_broadcast {{{2
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static constexpr _Tp _S_broadcast(_Tp __x) noexcept
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _Tp
+ _S_broadcast(_Tp __x) noexcept
{ return __x; }
// _S_generator {{{2
template <typename _Fp, typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static constexpr _Tp _S_generator(_Fp&& __gen,
- _TypeTag<_Tp>)
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _Tp
+ _S_generator(_Fp&& __gen, _TypeTag<_Tp>)
{ return __gen(_SizeConstant<0>()); }
// _S_load {{{2
template <typename _Tp, typename _Up>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_load(const _Up* __mem,
- _TypeTag<_Tp>) noexcept
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_load(const _Up* __mem, _TypeTag<_Tp>) noexcept
{ return static_cast<_Tp>(__mem[0]); }
// _S_masked_load {{{2
template <typename _Tp, typename _Up>
- _GLIBCXX_SIMD_INTRINSIC
- static _Tp _S_masked_load(_Tp __merge, bool __k,
- const _Up* __mem) noexcept
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_masked_load(_Tp __merge, bool __k, const _Up* __mem) noexcept
{
if (__k)
__merge = static_cast<_Tp>(__mem[0]);
@@ -166,97 +168,95 @@ struct _SimdImplScalar
// _S_store {{{2
template <typename _Tp, typename _Up>
- _GLIBCXX_SIMD_INTRINSIC
- static void _S_store(_Tp __v, _Up* __mem, _TypeTag<_Tp>) noexcept
+ _GLIBCXX_SIMD_INTRINSIC static void
+ _S_store(_Tp __v, _Up* __mem, _TypeTag<_Tp>) noexcept
{ __mem[0] = static_cast<_Up>(__v); }
// _S_masked_store {{{2
template <typename _Tp, typename _Up>
- _GLIBCXX_SIMD_INTRINSIC
- static void _S_masked_store(const _Tp __v, _Up* __mem,
- const bool __k) noexcept
+ _GLIBCXX_SIMD_INTRINSIC static void
+ _S_masked_store(const _Tp __v, _Up* __mem, const bool __k) noexcept
{ if (__k) __mem[0] = __v; }
// _S_negate {{{2
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC
- static constexpr bool _S_negate(_Tp __x) noexcept
+ _GLIBCXX_SIMD_INTRINSIC static constexpr bool
+ _S_negate(_Tp __x) noexcept
{ return !__x; }
// _S_reduce {{{2
template <typename _Tp, typename _BinaryOperation>
- _GLIBCXX_SIMD_INTRINSIC
- static constexpr _Tp
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _Tp
_S_reduce(const simd<_Tp, simd_abi::scalar>& __x, const _BinaryOperation&)
{ return __x._M_data; }
// _S_min, _S_max {{{2
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC
- static constexpr _Tp _S_min(const _Tp __a, const _Tp __b)
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _Tp
+ _S_min(const _Tp __a, const _Tp __b)
{ return std::min(__a, __b); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC
- static constexpr _Tp _S_max(const _Tp __a, const _Tp __b)
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _Tp
+ _S_max(const _Tp __a, const _Tp __b)
{ return std::max(__a, __b); }
// _S_complement {{{2
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC
- static constexpr _Tp _S_complement(_Tp __x) noexcept
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _Tp
+ _S_complement(_Tp __x) noexcept
{ return static_cast<_Tp>(~__x); }
// _S_unary_minus {{{2
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC
- static constexpr _Tp _S_unary_minus(_Tp __x) noexcept
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _Tp
+ _S_unary_minus(_Tp __x) noexcept
{ return static_cast<_Tp>(-__x); }
// arithmetic operators {{{2
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC
- static constexpr _Tp _S_plus(_Tp __x, _Tp __y)
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _Tp
+ _S_plus(_Tp __x, _Tp __y)
{
return static_cast<_Tp>(__promote_preserving_unsigned(__x)
+ __promote_preserving_unsigned(__y));
}
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC
- static constexpr _Tp _S_minus(_Tp __x, _Tp __y)
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _Tp
+ _S_minus(_Tp __x, _Tp __y)
{
return static_cast<_Tp>(__promote_preserving_unsigned(__x)
- __promote_preserving_unsigned(__y));
}
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC
- static constexpr _Tp _S_multiplies(_Tp __x, _Tp __y)
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _Tp
+ _S_multiplies(_Tp __x, _Tp __y)
{
return static_cast<_Tp>(__promote_preserving_unsigned(__x)
* __promote_preserving_unsigned(__y));
}
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC
- static constexpr _Tp _S_divides(_Tp __x, _Tp __y)
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _Tp
+ _S_divides(_Tp __x, _Tp __y)
{
return static_cast<_Tp>(__promote_preserving_unsigned(__x)
/ __promote_preserving_unsigned(__y));
}
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC
- static constexpr _Tp _S_modulus(_Tp __x, _Tp __y)
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _Tp
+ _S_modulus(_Tp __x, _Tp __y)
{
return static_cast<_Tp>(__promote_preserving_unsigned(__x)
% __promote_preserving_unsigned(__y));
}
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC
- static constexpr _Tp _S_bit_and(_Tp __x, _Tp __y)
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _Tp
+ _S_bit_and(_Tp __x, _Tp __y)
{
if constexpr (is_floating_point_v<_Tp>)
{
@@ -269,8 +269,8 @@ struct _SimdImplScalar
}
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC
- static constexpr _Tp _S_bit_or(_Tp __x, _Tp __y)
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _Tp
+ _S_bit_or(_Tp __x, _Tp __y)
{
if constexpr (is_floating_point_v<_Tp>)
{
@@ -283,8 +283,8 @@ struct _SimdImplScalar
}
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC
- static constexpr _Tp _S_bit_xor(_Tp __x, _Tp __y)
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _Tp
+ _S_bit_xor(_Tp __x, _Tp __y)
{
if constexpr (is_floating_point_v<_Tp>)
{
@@ -297,13 +297,13 @@ struct _SimdImplScalar
}
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC
- static constexpr _Tp _S_bit_shift_left(_Tp __x, int __y)
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _Tp
+ _S_bit_shift_left(_Tp __x, int __y)
{ return static_cast<_Tp>(__promote_preserving_unsigned(__x) << __y); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC
- static constexpr _Tp _S_bit_shift_right(_Tp __x, int __y)
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _Tp
+ _S_bit_shift_right(_Tp __x, int __y)
{ return static_cast<_Tp>(__promote_preserving_unsigned(__x) >> __y); }
// math {{{2
@@ -312,300 +312,362 @@ struct _SimdImplScalar
using _ST = _SimdTuple<_Tp, simd_abi::scalar>;
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_acos(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_acos(_Tp __x)
{ return std::acos(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_asin(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_asin(_Tp __x)
{ return std::asin(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_atan(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_atan(_Tp __x)
{ return std::atan(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_cos(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_cos(_Tp __x)
{ return std::cos(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_sin(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_sin(_Tp __x)
{ return std::sin(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_tan(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_tan(_Tp __x)
{ return std::tan(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_acosh(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_acosh(_Tp __x)
{ return std::acosh(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_asinh(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_asinh(_Tp __x)
{ return std::asinh(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_atanh(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_atanh(_Tp __x)
{ return std::atanh(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_cosh(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_cosh(_Tp __x)
{ return std::cosh(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_sinh(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_sinh(_Tp __x)
{ return std::sinh(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_tanh(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_tanh(_Tp __x)
{ return std::tanh(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_atan2(_Tp __x, _Tp __y)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_atan2(_Tp __x, _Tp __y)
{ return std::atan2(__x, __y); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_exp(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_exp(_Tp __x)
{ return std::exp(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_exp2(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_exp2(_Tp __x)
{ return std::exp2(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_expm1(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_expm1(_Tp __x)
{ return std::expm1(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_log(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_log(_Tp __x)
{ return std::log(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_log10(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_log10(_Tp __x)
{ return std::log10(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_log1p(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_log1p(_Tp __x)
{ return std::log1p(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_log2(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_log2(_Tp __x)
{ return std::log2(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_logb(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_logb(_Tp __x)
{ return std::logb(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _ST<int> _S_ilogb(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _ST<int>
+ _S_ilogb(_Tp __x)
{ return {std::ilogb(__x)}; }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_pow(_Tp __x, _Tp __y)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_pow(_Tp __x, _Tp __y)
{ return std::pow(__x, __y); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_abs(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_abs(_Tp __x)
{ return std::abs(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_fabs(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_fabs(_Tp __x)
{ return std::fabs(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_sqrt(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_sqrt(_Tp __x)
{ return std::sqrt(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_cbrt(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_cbrt(_Tp __x)
{ return std::cbrt(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_erf(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_erf(_Tp __x)
{ return std::erf(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_erfc(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_erfc(_Tp __x)
{ return std::erfc(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_lgamma(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_lgamma(_Tp __x)
{ return std::lgamma(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_tgamma(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_tgamma(_Tp __x)
{ return std::tgamma(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_trunc(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_trunc(_Tp __x)
{ return std::trunc(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_floor(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_floor(_Tp __x)
{ return std::floor(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_ceil(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_ceil(_Tp __x)
{ return std::ceil(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_nearbyint(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_nearbyint(_Tp __x)
{ return std::nearbyint(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_rint(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_rint(_Tp __x)
{ return std::rint(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _ST<long> _S_lrint(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _ST<long>
+ _S_lrint(_Tp __x)
{ return {std::lrint(__x)}; }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _ST<long long> _S_llrint(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _ST<long long>
+ _S_llrint(_Tp __x)
{ return {std::llrint(__x)}; }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_round(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_round(_Tp __x)
{ return std::round(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _ST<long> _S_lround(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _ST<long>
+ _S_lround(_Tp __x)
{ return {std::lround(__x)}; }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _ST<long long> _S_llround(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static _ST<long long>
+ _S_llround(_Tp __x)
{ return {std::llround(__x)}; }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_ldexp(_Tp __x, _ST<int> __y)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_ldexp(_Tp __x, _ST<int> __y)
{ return std::ldexp(__x, __y.first); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_scalbn(_Tp __x, _ST<int> __y)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_scalbn(_Tp __x, _ST<int> __y)
{ return std::scalbn(__x, __y.first); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_scalbln(_Tp __x, _ST<long> __y)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_scalbln(_Tp __x, _ST<long> __y)
{ return std::scalbln(__x, __y.first); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_fmod(_Tp __x, _Tp __y)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_fmod(_Tp __x, _Tp __y)
{ return std::fmod(__x, __y); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_remainder(_Tp __x, _Tp __y)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_remainder(_Tp __x, _Tp __y)
{ return std::remainder(__x, __y); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_nextafter(_Tp __x, _Tp __y)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_nextafter(_Tp __x, _Tp __y)
{ return std::nextafter(__x, __y); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_fdim(_Tp __x, _Tp __y)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_fdim(_Tp __x, _Tp __y)
{ return std::fdim(__x, __y); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_fmax(_Tp __x, _Tp __y)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_fmax(_Tp __x, _Tp __y)
{ return std::fmax(__x, __y); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_fmin(_Tp __x, _Tp __y)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_fmin(_Tp __x, _Tp __y)
{ return std::fmin(__x, __y); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_fma(_Tp __x, _Tp __y, _Tp __z)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_fma(_Tp __x, _Tp __y, _Tp __z)
{ return std::fma(__x, __y, __z); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_remquo(_Tp __x, _Tp __y, _ST<int>* __z)
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_remquo(_Tp __x, _Tp __y, _ST<int>* __z)
{ return std::remquo(__x, __y, &__z->first); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC constexpr static _ST<int> _S_fpclassify(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC constexpr static _ST<int>
+ _S_fpclassify(_Tp __x)
{ return {std::fpclassify(__x)}; }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC constexpr static bool _S_isfinite(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC constexpr static bool
+ _S_isfinite(_Tp __x)
{ return std::isfinite(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC constexpr static bool _S_isinf(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC constexpr static bool
+ _S_isinf(_Tp __x)
{ return std::isinf(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC constexpr static bool _S_isnan(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC constexpr static bool
+ _S_isnan(_Tp __x)
{ return std::isnan(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC constexpr static bool _S_isnormal(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC constexpr static bool
+ _S_isnormal(_Tp __x)
{ return std::isnormal(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC constexpr static bool _S_signbit(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC constexpr static bool
+ _S_signbit(_Tp __x)
{ return std::signbit(__x); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC constexpr static bool _S_isgreater(_Tp __x, _Tp __y)
+ _GLIBCXX_SIMD_INTRINSIC constexpr static bool
+ _S_isgreater(_Tp __x, _Tp __y)
{ return std::isgreater(__x, __y); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC constexpr static bool _S_isgreaterequal(_Tp __x,
- _Tp __y)
+ _GLIBCXX_SIMD_INTRINSIC constexpr static bool
+ _S_isgreaterequal(_Tp __x, _Tp __y)
{ return std::isgreaterequal(__x, __y); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC constexpr static bool _S_isless(_Tp __x, _Tp __y)
+ _GLIBCXX_SIMD_INTRINSIC constexpr static bool
+ _S_isless(_Tp __x, _Tp __y)
{ return std::isless(__x, __y); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC constexpr static bool _S_islessequal(_Tp __x, _Tp __y)
+ _GLIBCXX_SIMD_INTRINSIC constexpr static bool
+ _S_islessequal(_Tp __x, _Tp __y)
{ return std::islessequal(__x, __y); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC constexpr static bool _S_islessgreater(_Tp __x,
- _Tp __y)
+ _GLIBCXX_SIMD_INTRINSIC constexpr static bool
+ _S_islessgreater(_Tp __x, _Tp __y)
{ return std::islessgreater(__x, __y); }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC constexpr static bool _S_isunordered(_Tp __x,
- _Tp __y)
+ _GLIBCXX_SIMD_INTRINSIC constexpr static bool
+ _S_isunordered(_Tp __x, _Tp __y)
{ return std::isunordered(__x, __y); }
// _S_increment & _S_decrement{{{2
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC
- constexpr static void _S_increment(_Tp& __x)
+ _GLIBCXX_SIMD_INTRINSIC constexpr static void
+ _S_increment(_Tp& __x)
{ ++__x; }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC
- constexpr static void _S_decrement(_Tp& __x)
+ _GLIBCXX_SIMD_INTRINSIC constexpr static void
+ _S_decrement(_Tp& __x)
{ --__x; }
// compares {{{2
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC constexpr static bool _S_equal_to(_Tp __x, _Tp __y)
+ _GLIBCXX_SIMD_INTRINSIC constexpr static bool
+ _S_equal_to(_Tp __x, _Tp __y)
{ return __x == __y; }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC constexpr static bool _S_not_equal_to(_Tp __x,
- _Tp __y)
+ _GLIBCXX_SIMD_INTRINSIC constexpr static bool
+ _S_not_equal_to(_Tp __x, _Tp __y)
{ return __x != __y; }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC constexpr static bool _S_less(_Tp __x, _Tp __y)
+ _GLIBCXX_SIMD_INTRINSIC constexpr static bool
+ _S_less(_Tp __x, _Tp __y)
{ return __x < __y; }
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC constexpr static bool _S_less_equal(_Tp __x,
- _Tp __y)
+ _GLIBCXX_SIMD_INTRINSIC constexpr static bool
+ _S_less_equal(_Tp __x, _Tp __y)
{ return __x <= __y; }
// smart_reference access {{{2
template <typename _Tp, typename _Up>
- _GLIBCXX_SIMD_INTRINSIC
- constexpr static void _S_set(_Tp& __v, [[maybe_unused]] int __i,
- _Up&& __x) noexcept
+ _GLIBCXX_SIMD_INTRINSIC constexpr static void
+ _S_set(_Tp& __v, [[maybe_unused]] int __i, _Up&& __x) noexcept
{
_GLIBCXX_DEBUG_ASSERT(__i == 0);
__v = static_cast<_Up&&>(__x);
@@ -625,8 +687,8 @@ struct _SimdImplScalar
// _S_masked_unary {{{2
template <template <typename> class _Op, typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC constexpr static _Tp _S_masked_unary(const bool __k,
- const _Tp __v)
+ _GLIBCXX_SIMD_INTRINSIC constexpr static _Tp
+ _S_masked_unary(const bool __k, const _Tp __v)
{ return static_cast<_Tp>(__k ? _Op<_Tp>{}(__v) : __v); }
// }}}2
@@ -643,13 +705,15 @@ struct _MaskImplScalar
// }}}
// _S_broadcast {{{
template <typename>
- _GLIBCXX_SIMD_INTRINSIC static constexpr bool _S_broadcast(bool __x)
+ _GLIBCXX_SIMD_INTRINSIC static constexpr bool
+ _S_broadcast(bool __x)
{ return __x; }
// }}}
// _S_load {{{
template <typename>
- _GLIBCXX_SIMD_INTRINSIC static constexpr bool _S_load(const bool* __mem)
+ _GLIBCXX_SIMD_INTRINSIC static constexpr bool
+ _S_load(const bool* __mem)
{ return __mem[0]; }
// }}}
@@ -687,7 +751,8 @@ _S_convert(simd_mask<_Up, _UAbi> __x)
}
// _S_store {{{2
- _GLIBCXX_SIMD_INTRINSIC static void _S_store(bool __v, bool* __mem) noexcept
+ _GLIBCXX_SIMD_INTRINSIC static void
+ _S_store(bool __v, bool* __mem) noexcept
{ __mem[0] = __v; }
// _S_masked_store {{{2
@@ -699,42 +764,41 @@ _S_convert(simd_mask<_Up, _UAbi> __x)
}
// logical and bitwise operators {{{2
- _GLIBCXX_SIMD_INTRINSIC
- static constexpr bool _S_logical_and(bool __x, bool __y)
+ _GLIBCXX_SIMD_INTRINSIC static constexpr bool
+ _S_logical_and(bool __x, bool __y)
{ return __x && __y; }
- _GLIBCXX_SIMD_INTRINSIC
- static constexpr bool _S_logical_or(bool __x, bool __y)
+ _GLIBCXX_SIMD_INTRINSIC static constexpr bool
+ _S_logical_or(bool __x, bool __y)
{ return __x || __y; }
- _GLIBCXX_SIMD_INTRINSIC
- static constexpr bool _S_bit_not(bool __x)
+ _GLIBCXX_SIMD_INTRINSIC static constexpr bool
+ _S_bit_not(bool __x)
{ return !__x; }
- _GLIBCXX_SIMD_INTRINSIC
- static constexpr bool _S_bit_and(bool __x, bool __y)
+ _GLIBCXX_SIMD_INTRINSIC static constexpr bool
+ _S_bit_and(bool __x, bool __y)
{ return __x && __y; }
- _GLIBCXX_SIMD_INTRINSIC
- static constexpr bool _S_bit_or(bool __x, bool __y)
+ _GLIBCXX_SIMD_INTRINSIC static constexpr bool
+ _S_bit_or(bool __x, bool __y)
{ return __x || __y; }
- _GLIBCXX_SIMD_INTRINSIC
- static constexpr bool _S_bit_xor(bool __x, bool __y)
+ _GLIBCXX_SIMD_INTRINSIC static constexpr bool
+ _S_bit_xor(bool __x, bool __y)
{ return __x != __y; }
// smart_reference access {{{2
- _GLIBCXX_SIMD_INTRINSIC
- constexpr static void _S_set(bool& __k, [[maybe_unused]] int __i,
- bool __x) noexcept
+ _GLIBCXX_SIMD_INTRINSIC constexpr static void
+ _S_set(bool& __k, [[maybe_unused]] int __i, bool __x) noexcept
{
_GLIBCXX_DEBUG_ASSERT(__i == 0);
__k = __x;
}
// _S_masked_assign {{{2
- _GLIBCXX_SIMD_INTRINSIC static void _S_masked_assign(bool __k, bool& __lhs,
- bool __rhs)
+ _GLIBCXX_SIMD_INTRINSIC static void
+ _S_masked_assign(bool __k, bool& __lhs, bool __rhs)
{
if (__k)
__lhs = __rhs;
diff --git a/libstdc++-v3/include/experimental/bits/simd_x86.h b/libstdc++-v3/include/experimental/bits/simd_x86.h
index 8872ca301b9..61177fa0b04 100644
--- a/libstdc++-v3/include/experimental/bits/simd_x86.h
+++ b/libstdc++-v3/include/experimental/bits/simd_x86.h
@@ -40,10 +40,7 @@
template <typename _Tp, size_t _Np>
_GLIBCXX_SIMD_INTRINSIC constexpr _SimdWrapper<__int_for_sizeof_t<_Tp>, _Np>
__to_masktype(_SimdWrapper<_Tp, _Np> __x)
- {
- return reinterpret_cast<__vector_type_t<__int_for_sizeof_t<_Tp>, _Np>>(
- __x._M_data);
- }
+ { return reinterpret_cast<__vector_type_t<__int_for_sizeof_t<_Tp>, _Np>>( __x._M_data); }
template <typename _TV,
typename _TVT
@@ -434,7 +431,8 @@ struct _CommonImplX86
#ifdef _GLIBCXX_SIMD_WORKAROUND_PR85048
// _S_converts_via_decomposition {{{
template <typename _From, typename _To, size_t _ToSize>
- static constexpr bool _S_converts_via_decomposition()
+ static constexpr bool
+ _S_converts_via_decomposition()
{
if constexpr (is_integral_v<
_From> && is_integral_v<_To> && sizeof(_From) == 8
@@ -465,8 +463,8 @@ struct _CommonImplX86
using _CommonImplBuiltin::_S_store;
template <typename _Tp, size_t _Np>
- _GLIBCXX_SIMD_INTRINSIC static void _S_store(_SimdWrapper<_Tp, _Np> __x,
- void* __addr)
+ _GLIBCXX_SIMD_INTRINSIC static void
+ _S_store(_SimdWrapper<_Tp, _Np> __x, void* __addr)
{
constexpr size_t _Bytes = _Np * sizeof(_Tp);
@@ -702,8 +700,8 @@ _pdep_u32(
// Requires: _Tp to be an intrinsic type (integers blend per byte) and 16/32
// Bytes wide
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_blend_intrin(_Tp __k, _Tp __a,
- _Tp __b) noexcept
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_blend_intrin(_Tp __k, _Tp __a, _Tp __b) noexcept
{
static_assert(is_same_v<decltype(__to_intrin(__a)), _Tp>);
constexpr struct
@@ -843,6 +841,7 @@ struct _SimdImplX86
= (sizeof(_Tp) >= 4 && __have_avx512f) || __have_avx512bw ? 64
: (is_floating_point_v<_Tp>&& __have_avx) || __have_avx2 ? 32
: 16;
+
using _MaskImpl = typename _Abi::_MaskImpl;
// _S_masked_load {{{
@@ -1033,8 +1032,7 @@ _S_masked_load(_SimdWrapper<_Tp, _Np> __merge, _MaskMember<_Tp> __k,
// _S_masked_store_nocvt {{{
template <typename _Tp, size_t _Np>
_GLIBCXX_SIMD_INTRINSIC static void
- _S_masked_store_nocvt(_SimdWrapper<_Tp, _Np> __v, _Tp* __mem,
- _SimdWrapper<bool, _Np> __k)
+ _S_masked_store_nocvt(_SimdWrapper<_Tp, _Np> __v, _Tp* __mem, _SimdWrapper<bool, _Np> __k)
{
[[maybe_unused]] const auto __vi = __to_intrin(__v);
if constexpr (sizeof(__vi) == 64)
@@ -1301,7 +1299,8 @@ _S_masked_store(const _SimdWrapper<_Tp, _Np> __v, _Up* __mem,
// }}}
// _S_multiplies {{{
template <typename _V, typename _VVT = _VectorTraits<_V>>
- _GLIBCXX_SIMD_INTRINSIC static constexpr _V _S_multiplies(_V __x, _V __y)
+ _GLIBCXX_SIMD_INTRINSIC static constexpr _V
+ _S_multiplies(_V __x, _V __y)
{
using _Tp = typename _VVT::value_type;
if (__builtin_is_constant_evaluated() || __x._M_is_constprop()
@@ -2739,7 +2738,8 @@ _S_round(_SimdWrapper<_Tp, _Np> __x)
// }}}
// _S_nearbyint {{{
template <typename _Tp, typename _TVT = _VectorTraits<_Tp>>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_nearbyint(_Tp __x) noexcept
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_nearbyint(_Tp __x) noexcept
{
if constexpr (_TVT::template _S_is<float, 16>)
return _mm512_roundscale_ps(__x, 0x0c);
@@ -2764,7 +2764,8 @@ _S_round(_SimdWrapper<_Tp, _Np> __x)
// }}}
// _S_rint {{{
template <typename _Tp, typename _TVT = _VectorTraits<_Tp>>
- _GLIBCXX_SIMD_INTRINSIC static _Tp _S_rint(_Tp __x) noexcept
+ _GLIBCXX_SIMD_INTRINSIC static _Tp
+ _S_rint(_Tp __x) noexcept
{
if constexpr (_TVT::template _S_is<float, 16>)
return _mm512_roundscale_ps(__x, 0x04);
@@ -2912,7 +2913,8 @@ _S_signbit(_SimdWrapper<_Tp, _Np> __x)
// _S_isnonzerovalue_mask {{{
// (isnormal | is subnormal == !isinf & !isnan & !is zero)
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static auto _S_isnonzerovalue_mask(_Tp __x)
+ _GLIBCXX_SIMD_INTRINSIC static auto
+ _S_isnonzerovalue_mask(_Tp __x)
{
using _Traits = _VectorTraits<_Tp>;
if constexpr (__have_avx512dq_vl)
@@ -3179,8 +3181,8 @@ _S_isnan(_SimdWrapper<_Tp, _Np> __x)
// }}}
// _S_isgreater {{{
template <typename _Tp, size_t _Np>
- static constexpr _MaskMember<_Tp> _S_isgreater(_SimdWrapper<_Tp, _Np> __x,
- _SimdWrapper<_Tp, _Np> __y)
+ static constexpr _MaskMember<_Tp>
+ _S_isgreater(_SimdWrapper<_Tp, _Np> __x, _SimdWrapper<_Tp, _Np> __y)
{
const auto __xi = __to_intrin(__x);
const auto __yi = __to_intrin(__y);
@@ -3297,8 +3299,8 @@ _S_isgreaterequal(_SimdWrapper<_Tp, _Np> __x, _SimdWrapper<_Tp, _Np> __y)
// }}}
// _S_isless {{{
template <typename _Tp, size_t _Np>
- static constexpr _MaskMember<_Tp> _S_isless(_SimdWrapper<_Tp, _Np> __x,
- _SimdWrapper<_Tp, _Np> __y)
+ static constexpr _MaskMember<_Tp>
+ _S_isless(_SimdWrapper<_Tp, _Np> __x, _SimdWrapper<_Tp, _Np> __y)
{
const auto __xi = __to_intrin(__x);
const auto __yi = __to_intrin(__y);
@@ -3462,11 +3464,9 @@ _S_islessgreater(_SimdWrapper<_Tp, _Np> __x, _SimdWrapper<_Tp, _Np> __y)
}
//}}} }}}
- template <template <typename> class _Op, typename _Tp, typename _K,
- size_t _Np>
+ template <template <typename> class _Op, typename _Tp, typename _K, size_t _Np>
_GLIBCXX_SIMD_INTRINSIC static _SimdWrapper<_Tp, _Np>
- _S_masked_unary(const _SimdWrapper<_K, _Np> __k,
- const _SimdWrapper<_Tp, _Np> __v)
+ _S_masked_unary(const _SimdWrapper<_K, _Np> __k, const _SimdWrapper<_Tp, _Np> __v)
{
if (__k._M_is_constprop_none_of())
return __v;
@@ -3543,8 +3543,8 @@ struct _MaskImplX86Mixin
// _S_to_maskvector(bool) {{{
template <typename _Up, size_t _ToN = 1, typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static constexpr enable_if_t<
- is_same_v<_Tp, bool>, _SimdWrapper<_Up, _ToN>>
+ _GLIBCXX_SIMD_INTRINSIC static constexpr
+ enable_if_t<is_same_v<_Tp, bool>, _SimdWrapper<_Up, _ToN>>
_S_to_maskvector(_Tp __x)
{
static_assert(is_same_v<_Up, __int_for_sizeof_t<_Up>>);
@@ -3554,8 +3554,7 @@ _S_to_maskvector(_Tp __x)
// }}}
// _S_to_maskvector(_SanitizedBitMask) {{{
- template <typename _Up, size_t _UpN = 0, size_t _Np,
- size_t _ToN = _UpN == 0 ? _Np : _UpN>
+ template <typename _Up, size_t _UpN = 0, size_t _Np, size_t _ToN = _UpN == 0 ? _Np : _UpN>
_GLIBCXX_SIMD_INTRINSIC static constexpr _SimdWrapper<_Up, _ToN>
_S_to_maskvector(_SanitizedBitMask<_Np> __x)
{
@@ -4626,8 +4625,8 @@ _mm256_cvtepi8_epi64(
// _S_store {{{2
template <typename _Tp, size_t _Np>
- _GLIBCXX_SIMD_INTRINSIC static void _S_store(_SimdWrapper<_Tp, _Np> __v,
- bool* __mem) noexcept
+ _GLIBCXX_SIMD_INTRINSIC static void
+ _S_store(_SimdWrapper<_Tp, _Np> __v, bool* __mem) noexcept
{
if constexpr (__is_avx512_abi<_Abi>())
{
@@ -4791,8 +4790,7 @@ _S_masked_store(const _SimdWrapper<_Tp, _Np> __v, bool* __mem,
// logical and bitwise operators {{{2
template <typename _Tp, size_t _Np>
_GLIBCXX_SIMD_INTRINSIC static constexpr _SimdWrapper<_Tp, _Np>
- _S_logical_and(const _SimdWrapper<_Tp, _Np>& __x,
- const _SimdWrapper<_Tp, _Np>& __y)
+ _S_logical_and(const _SimdWrapper<_Tp, _Np>& __x, const _SimdWrapper<_Tp, _Np>& __y)
{
if constexpr (is_same_v<_Tp, bool>)
{
@@ -4813,8 +4811,7 @@ _S_logical_and(const _SimdWrapper<_Tp, _Np>& __x,
template <typename _Tp, size_t _Np>
_GLIBCXX_SIMD_INTRINSIC static constexpr _SimdWrapper<_Tp, _Np>
- _S_logical_or(const _SimdWrapper<_Tp, _Np>& __x,
- const _SimdWrapper<_Tp, _Np>& __y)
+ _S_logical_or(const _SimdWrapper<_Tp, _Np>& __x, const _SimdWrapper<_Tp, _Np>& __y)
{
if constexpr (is_same_v<_Tp, bool>)
{
@@ -4860,8 +4857,7 @@ _S_bit_not(const _SimdWrapper<_Tp, _Np>& __x)
template <typename _Tp, size_t _Np>
_GLIBCXX_SIMD_INTRINSIC static constexpr _SimdWrapper<_Tp, _Np>
- _S_bit_and(const _SimdWrapper<_Tp, _Np>& __x,
- const _SimdWrapper<_Tp, _Np>& __y)
+ _S_bit_and(const _SimdWrapper<_Tp, _Np>& __x, const _SimdWrapper<_Tp, _Np>& __y)
{
if constexpr (is_same_v<_Tp, bool>)
{
@@ -4882,8 +4878,7 @@ _S_bit_and(const _SimdWrapper<_Tp, _Np>& __x,
template <typename _Tp, size_t _Np>
_GLIBCXX_SIMD_INTRINSIC static constexpr _SimdWrapper<_Tp, _Np>
- _S_bit_or(const _SimdWrapper<_Tp, _Np>& __x,
- const _SimdWrapper<_Tp, _Np>& __y)
+ _S_bit_or(const _SimdWrapper<_Tp, _Np>& __x, const _SimdWrapper<_Tp, _Np>& __y)
{
if constexpr (is_same_v<_Tp, bool>)
{
@@ -4904,8 +4899,7 @@ _S_bit_or(const _SimdWrapper<_Tp, _Np>& __x,
template <typename _Tp, size_t _Np>
_GLIBCXX_SIMD_INTRINSIC static constexpr _SimdWrapper<_Tp, _Np>
- _S_bit_xor(const _SimdWrapper<_Tp, _Np>& __x,
- const _SimdWrapper<_Tp, _Np>& __y)
+ _S_bit_xor(const _SimdWrapper<_Tp, _Np>& __x, const _SimdWrapper<_Tp, _Np>& __y)
{
if constexpr (is_same_v<_Tp, bool>)
{
@@ -4929,8 +4923,7 @@ _S_bit_xor(const _SimdWrapper<_Tp, _Np>& __x,
template <size_t _Np>
_GLIBCXX_SIMD_INTRINSIC static void
_S_masked_assign(_SimdWrapper<bool, _Np> __k,
- _SimdWrapper<bool, _Np>& __lhs,
- _SimdWrapper<bool, _Np> __rhs)
+ _SimdWrapper<bool, _Np>& __lhs, _SimdWrapper<bool, _Np> __rhs)
{
__lhs._M_data
= (~__k._M_data & __lhs._M_data) | (__k._M_data & __rhs._M_data);
@@ -4952,7 +4945,8 @@ _S_masked_assign(_SimdWrapper<bool, _Np> __k,
//}}}
// _S_all_of {{{
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static bool _S_all_of(simd_mask<_Tp, _Abi> __k)
+ _GLIBCXX_SIMD_INTRINSIC static bool
+ _S_all_of(simd_mask<_Tp, _Abi> __k)
{
if constexpr (__is_sse_abi<_Abi>() || __is_avx_abi<_Abi>())
{
@@ -5008,7 +5002,8 @@ _S_masked_assign(_SimdWrapper<bool, _Np> __k,
// }}}
// _S_any_of {{{
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static bool _S_any_of(simd_mask<_Tp, _Abi> __k)
+ _GLIBCXX_SIMD_INTRINSIC static bool
+ _S_any_of(simd_mask<_Tp, _Abi> __k)
{
if constexpr (__is_sse_abi<_Abi>() || __is_avx_abi<_Abi>())
{
@@ -5043,7 +5038,8 @@ _S_masked_assign(_SimdWrapper<bool, _Np> __k,
// }}}
// _S_none_of {{{
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static bool _S_none_of(simd_mask<_Tp, _Abi> __k)
+ _GLIBCXX_SIMD_INTRINSIC static bool
+ _S_none_of(simd_mask<_Tp, _Abi> __k)
{
if constexpr (__is_sse_abi<_Abi>() || __is_avx_abi<_Abi>())
{
@@ -5078,7 +5074,8 @@ _S_masked_assign(_SimdWrapper<bool, _Np> __k,
// }}}
// _S_some_of {{{
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static bool _S_some_of(simd_mask<_Tp, _Abi> __k)
+ _GLIBCXX_SIMD_INTRINSIC static bool
+ _S_some_of(simd_mask<_Tp, _Abi> __k)
{
if constexpr (__is_sse_abi<_Abi>() || __is_avx_abi<_Abi>())
{
@@ -5119,7 +5116,8 @@ _S_masked_assign(_SimdWrapper<bool, _Np> __k,
// }}}
// _S_popcount {{{
template <typename _Tp>
- _GLIBCXX_SIMD_INTRINSIC static int _S_popcount(simd_mask<_Tp, _Abi> __k)
+ _GLIBCXX_SIMD_INTRINSIC static int
+ _S_popcount(simd_mask<_Tp, _Abi> __k)
{
constexpr size_t _Np = simd_size_v<_Tp, _Abi>;
const auto __kk = _Abi::_S_masked(__k._M_data)._M_data;
^ permalink raw reply [flat|nested] 19+ messages in thread
* [PATCH 7/8] libstdc++: Fix -Wsign-compare issue
2023-02-23 8:48 [PATCH 0/8] std::experimental::simd patchset Matthias Kretz
` (5 preceding siblings ...)
2023-02-23 8:50 ` [PATCH 6/8] libstdc++: Fix formatting Matthias Kretz
@ 2023-02-23 8:50 ` Matthias Kretz
2023-02-23 11:07 ` Jonathan Wakely
2023-02-23 8:50 ` [PATCH 8/8] libstdc++: Test that integral simd reductions are precise Matthias Kretz
7 siblings, 1 reply; 19+ messages in thread
From: Matthias Kretz @ 2023-02-23 8:50 UTC (permalink / raw)
To: gcc-patches, libstdc++
[-- Attachment #1: Type: text/plain, Size: 920 bytes --]
Signed-off-by: Matthias Kretz <m.kretz@gsi.de>
libstdc++-v3/ChangeLog:
* include/experimental/bits/simd_builtin.h (_S_set): Compare as
int. The actual range of these indexes is very small.
---
libstdc++-v3/include/experimental/bits/simd_builtin.h | 2 +-
1 file changed, 1 insertion(+), 1 deletion(-)
--
──────────────────────────────────────────────────────────────────────────
Dr. Matthias Kretz https://mattkretz.github.io
GSI Helmholtz Centre for Heavy Ion Research https://gsi.de
stdₓ::simd
──────────────────────────────────────────────────────────────────────────
[-- Attachment #2: 0007-libstdc-Fix-Wsign-compare-issue.patch --]
[-- Type: text/x-patch, Size: 653 bytes --]
diff --git a/libstdc++-v3/include/experimental/bits/simd_builtin.h b/libstdc++-v3/include/experimental/bits/simd_builtin.h
index 0e75f941288..30bbfa7d478 100644
--- a/libstdc++-v3/include/experimental/bits/simd_builtin.h
+++ b/libstdc++-v3/include/experimental/bits/simd_builtin.h
@@ -2875,7 +2875,7 @@ _S_bit_xor(const _SimdWrapper<_Tp, _Np>& __x, const _SimdWrapper<_Tp, _Np>& __y)
__k = __generate_from_n_evaluations<_Np,
__vector_type_t<_Tp, _Np>>(
[&](auto __j) _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA {
- if (__i == __j)
+ if (__i == static_cast<int>(__j))
return _Tp(-__x);
else
return __k[+__j];
^ permalink raw reply [flat|nested] 19+ messages in thread
* [PATCH 8/8] libstdc++: Test that integral simd reductions are precise
2023-02-23 8:48 [PATCH 0/8] std::experimental::simd patchset Matthias Kretz
` (6 preceding siblings ...)
2023-02-23 8:50 ` [PATCH 7/8] libstdc++: Fix -Wsign-compare issue Matthias Kretz
@ 2023-02-23 8:50 ` Matthias Kretz
2023-02-23 11:08 ` Jonathan Wakely
7 siblings, 1 reply; 19+ messages in thread
From: Matthias Kretz @ 2023-02-23 8:50 UTC (permalink / raw)
To: gcc-patches, libstdc++
[-- Attachment #1: Type: text/plain, Size: 918 bytes --]
Signed-off-by: Matthias Kretz <m.kretz@gsi.de>
libstdc++-v3/ChangeLog:
* testsuite/experimental/simd/tests/reductions.cc: Introduce
max_distance as the type-dependent max error.
---
libstdc++-v3/testsuite/experimental/simd/tests/reductions.cc | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)
--
──────────────────────────────────────────────────────────────────────────
Dr. Matthias Kretz https://mattkretz.github.io
GSI Helmholtz Centre for Heavy Ion Research https://gsi.de
stdₓ::simd
──────────────────────────────────────────────────────────────────────────
[-- Attachment #2: 0008-libstdc-Test-that-integral-simd-reductions-are-preci.patch --]
[-- Type: text/x-patch, Size: 665 bytes --]
diff --git a/libstdc++-v3/testsuite/experimental/simd/tests/reductions.cc b/libstdc++-v3/testsuite/experimental/simd/tests/reductions.cc
index 0c4c79feb20..fed164314d7 100644
--- a/libstdc++-v3/testsuite/experimental/simd/tests/reductions.cc
+++ b/libstdc++-v3/testsuite/experimental/simd/tests/reductions.cc
@@ -112,6 +112,7 @@ template <typename V>
T acc = x[0];
for (size_t i = 1; i < V::size(); ++i)
acc += x[i];
- ULP_COMPARE(reduce(x), acc, V::size() / 2).on_failure("x = ", x);
+ const T max_distance = std::is_integral_v<T> ? 0 : V::size() / 2;
+ ULP_COMPARE(reduce(x), acc, max_distance).on_failure("x = ", x);
});
}
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: [PATCH 1/8] libstdc++: Simplify three helper functions into one
2023-02-23 8:49 ` [PATCH 1/8] libstdc++: Simplify three helper functions into one Matthias Kretz
@ 2023-02-23 11:05 ` Jonathan Wakely
0 siblings, 0 replies; 19+ messages in thread
From: Jonathan Wakely @ 2023-02-23 11:05 UTC (permalink / raw)
To: Matthias Kretz; +Cc: gcc-patches, libstdc++
On Thu, 23 Feb 2023 at 08:53, Matthias Kretz via Libstdc++
<libstdc++@gcc.gnu.org> wrote:
>
>
>
> Broadcast is a very common function. This should reduce compile-time
> effort.
OK for all branches.
> Signed-off-by: Matthias Kretz <m.kretz@gsi.de>
>
> libstdc++-v3/ChangeLog:
>
> PR libstdc++/108030
> * include/experimental/bits/simd.h (__vector_broadcast):
> Implement via __vector_broadcast_impl instead of
> __call_with_n_evaluations + 2 lambdas.
> (__vector_broadcast_impl): New.
> ---
> libstdc++-v3/include/experimental/bits/simd.h | 12 ++++++------
> 1 file changed, 6 insertions(+), 6 deletions(-)
>
>
> --
> ──────────────────────────────────────────────────────────────────────────
> Dr. Matthias Kretz https://mattkretz.github.io
> GSI Helmholtz Centre for Heavy Ion Research https://gsi.de
> stdₓ::simd
> ──────────────────────────────────────────────────────────────────────────
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: [PATCH 2/8] libstdc++: Fix simd build failure on clang
2023-02-23 8:49 ` [PATCH 2/8] libstdc++: Fix simd build failure on clang Matthias Kretz
@ 2023-02-23 11:06 ` Jonathan Wakely
0 siblings, 0 replies; 19+ messages in thread
From: Jonathan Wakely @ 2023-02-23 11:06 UTC (permalink / raw)
To: Matthias Kretz; +Cc: gcc-patches, libstdc++
On Thu, 23 Feb 2023 at 08:54, Matthias Kretz via Libstdc++
<libstdc++@gcc.gnu.org> wrote:
>
>
>
> Clang does not support __attribute__ on lambdas. Therefore, only set
> _GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA if __clang__ is not defined.
OK for all branches.
> Signed-off-by: Matthias Kretz <m.kretz@gsi.de>
>
> libstdc++-v3/ChangeLog:
>
> PR libstdc++/108030
> * include/experimental/bits/simd_detail.h
> (_GLIBCXX_SIMD_ALWAYS_INLINE_LAMBDA): Define as empty for
> __clang__.
> ---
> libstdc++-v3/include/experimental/bits/simd_detail.h | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
>
> --
> ──────────────────────────────────────────────────────────────────────────
> Dr. Matthias Kretz https://mattkretz.github.io
> GSI Helmholtz Centre for Heavy Ion Research https://gsi.de
> stdₓ::simd
> ──────────────────────────────────────────────────────────────────────────
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: [PATCH 4/8] libstdc++: Add missing constexpr on simd shift implementation
2023-02-23 8:49 ` [PATCH 4/8] libstdc++: Add missing constexpr on simd shift implementation Matthias Kretz
@ 2023-02-23 11:07 ` Jonathan Wakely
2023-02-23 11:33 ` Matthias Kretz
0 siblings, 1 reply; 19+ messages in thread
From: Jonathan Wakely @ 2023-02-23 11:07 UTC (permalink / raw)
To: Matthias Kretz; +Cc: gcc-patches, libstdc++
On Thu, 23 Feb 2023 at 08:55, Matthias Kretz via Libstdc++
<libstdc++@gcc.gnu.org> wrote:
>
>
>
> Resolves -Wtautological-compare warnings about `if
> (__builtin_is_constant_evaluated())` in the implementations of these
> functions.
The 'inline' is redundant now, because these are unconditionally
constexpr which implies inline.
OK for all branches, with or without removing the 'inline'.
>
> Signed-off-by: Matthias Kretz <m.kretz@gsi.de>
>
> libstdc++-v3/ChangeLog:
>
> * include/experimental/bits/simd_x86.h (_S_bit_shift_left)
> (_S_bit_shift_right): Declare constexpr. The implementation was
> already expecting constexpr evaluation.
> ---
> libstdc++-v3/include/experimental/bits/simd_x86.h | 8 ++++----
> 1 file changed, 4 insertions(+), 4 deletions(-)
>
>
> --
> ──────────────────────────────────────────────────────────────────────────
> Dr. Matthias Kretz https://mattkretz.github.io
> GSI Helmholtz Centre for Heavy Ion Research https://gsi.de
> stdₓ::simd
> ──────────────────────────────────────────────────────────────────────────
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: [PATCH 7/8] libstdc++: Fix -Wsign-compare issue
2023-02-23 8:50 ` [PATCH 7/8] libstdc++: Fix -Wsign-compare issue Matthias Kretz
@ 2023-02-23 11:07 ` Jonathan Wakely
0 siblings, 0 replies; 19+ messages in thread
From: Jonathan Wakely @ 2023-02-23 11:07 UTC (permalink / raw)
To: Matthias Kretz; +Cc: gcc-patches, libstdc++
On Thu, 23 Feb 2023 at 08:51, Matthias Kretz via Libstdc++
<libstdc++@gcc.gnu.org> wrote:
>
>
OK for all branches.
> Signed-off-by: Matthias Kretz <m.kretz@gsi.de>
>
> libstdc++-v3/ChangeLog:
>
> * include/experimental/bits/simd_builtin.h (_S_set): Compare as
> int. The actual range of these indexes is very small.
> ---
> libstdc++-v3/include/experimental/bits/simd_builtin.h | 2 +-
> 1 file changed, 1 insertion(+), 1 deletion(-)
>
>
> --
> ──────────────────────────────────────────────────────────────────────────
> Dr. Matthias Kretz https://mattkretz.github.io
> GSI Helmholtz Centre for Heavy Ion Research https://gsi.de
> stdₓ::simd
> ──────────────────────────────────────────────────────────────────────────
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: [PATCH 8/8] libstdc++: Test that integral simd reductions are precise
2023-02-23 8:50 ` [PATCH 8/8] libstdc++: Test that integral simd reductions are precise Matthias Kretz
@ 2023-02-23 11:08 ` Jonathan Wakely
0 siblings, 0 replies; 19+ messages in thread
From: Jonathan Wakely @ 2023-02-23 11:08 UTC (permalink / raw)
To: Matthias Kretz; +Cc: gcc-patches, libstdc++
On Thu, 23 Feb 2023 at 08:51, Matthias Kretz via Libstdc++
<libstdc++@gcc.gnu.org> wrote:
>
OK for all branches.
>
> Signed-off-by: Matthias Kretz <m.kretz@gsi.de>
>
> libstdc++-v3/ChangeLog:
>
> * testsuite/experimental/simd/tests/reductions.cc: Introduce
> max_distance as the type-dependent max error.
> ---
> libstdc++-v3/testsuite/experimental/simd/tests/reductions.cc | 3 ++-
> 1 file changed, 2 insertions(+), 1 deletion(-)
>
>
> --
> ──────────────────────────────────────────────────────────────────────────
> Dr. Matthias Kretz https://mattkretz.github.io
> GSI Helmholtz Centre for Heavy Ion Research https://gsi.de
> stdₓ::simd
> ──────────────────────────────────────────────────────────────────────────
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: [PATCH 4/8] libstdc++: Add missing constexpr on simd shift implementation
2023-02-23 11:07 ` Jonathan Wakely
@ 2023-02-23 11:33 ` Matthias Kretz
0 siblings, 0 replies; 19+ messages in thread
From: Matthias Kretz @ 2023-02-23 11:33 UTC (permalink / raw)
To: gcc-patches, libstdc++; +Cc: Jonathan Wakely
On Thursday, 23 February 2023 12:07:11 CET Jonathan Wakely wrote:
> On Thu, 23 Feb 2023 at 08:55, Matthias Kretz via Libstdc++
>
> <libstdc++@gcc.gnu.org> wrote:
> > Resolves -Wtautological-compare warnings about `if
> > (__builtin_is_constant_evaluated())` in the implementations of these
> > functions.
>
> The 'inline' is redundant now, because these are unconditionally
> constexpr which implies inline.
In the simd implementation I always have to make a conscious choice of
always_inline vs. inline. Having the inline keyword there helps documenting
that choice and helps revisiting all not-always_inline functions quickly.
--
──────────────────────────────────────────────────────────────────────────
Dr. Matthias Kretz https://mattkretz.github.io
GSI Helmholtz Centre for Heavy Ion Research https://gsi.de
stdₓ::simd
──────────────────────────────────────────────────────────────────────────
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: [PATCH 5/8] libstdc++: Always-inline most of non-cmath fixed_size implementation
2023-02-23 8:49 ` [PATCH 5/8] libstdc++: Always-inline most of non-cmath fixed_size implementation Matthias Kretz
@ 2023-02-24 17:10 ` Jonathan Wakely
0 siblings, 0 replies; 19+ messages in thread
From: Jonathan Wakely @ 2023-02-24 17:10 UTC (permalink / raw)
To: Matthias Kretz; +Cc: gcc-patches, libstdc++
On Thu, 23 Feb 2023 at 08:54, Matthias Kretz via Libstdc++
<libstdc++@gcc.gnu.org> wrote:
>
>
>
> For simd, the inlining behavior should be similar to builtin types. (No
> operator on buitin types is ever translated into a function call.)
> Therefore, always_inline is the right choice (i.e. inline on -O0 as
> well).
OK for trunk (and OK for backport later if no problems show up from
the extra inlining).
>
> Signed-off-by: Matthias Kretz <m.kretz@gsi.de>
>
> libstdc++-v3/ChangeLog:
>
> PR libstdc++/108030
> * include/experimental/bits/simd_fixed_size.h
> (_SimdImplFixedSize::_S_broadcast): Replace inline with
> _GLIBCXX_SIMD_INTRINSIC.
> (_SimdImplFixedSize::_S_generate): Likewise.
> (_SimdImplFixedSize::_S_load): Likewise.
> (_SimdImplFixedSize::_S_masked_load): Likewise.
> (_SimdImplFixedSize::_S_store): Likewise.
> (_SimdImplFixedSize::_S_masked_store): Likewise.
> (_SimdImplFixedSize::_S_min): Likewise.
> (_SimdImplFixedSize::_S_max): Likewise.
> (_SimdImplFixedSize::_S_complement): Likewise.
> (_SimdImplFixedSize::_S_unary_minus): Likewise.
> (_SimdImplFixedSize::_S_plus): Likewise.
> (_SimdImplFixedSize::_S_minus): Likewise.
> (_SimdImplFixedSize::_S_multiplies): Likewise.
> (_SimdImplFixedSize::_S_divides): Likewise.
> (_SimdImplFixedSize::_S_modulus): Likewise.
> (_SimdImplFixedSize::_S_bit_and): Likewise.
> (_SimdImplFixedSize::_S_bit_or): Likewise.
> (_SimdImplFixedSize::_S_bit_xor): Likewise.
> (_SimdImplFixedSize::_S_bit_shift_left): Likewise.
> (_SimdImplFixedSize::_S_bit_shift_right): Likewise.
> (_SimdImplFixedSize::_S_remquo): Add inline keyword (to be
> explicit about not always-inline, yet).
> (_SimdImplFixedSize::_S_isinf): Likewise.
> (_SimdImplFixedSize::_S_isfinite): Likewise.
> (_SimdImplFixedSize::_S_isnan): Likewise.
> (_SimdImplFixedSize::_S_isnormal): Likewise.
> (_SimdImplFixedSize::_S_signbit): Likewise.
> ---
> .../experimental/bits/simd_fixed_size.h | 60 +++++++++----------
> 1 file changed, 30 insertions(+), 30 deletions(-)
>
>
> --
> ──────────────────────────────────────────────────────────────────────────
> Dr. Matthias Kretz https://mattkretz.github.io
> GSI Helmholtz Centre for Heavy Ion Research https://gsi.de
> stdₓ::simd
> ──────────────────────────────────────────────────────────────────────────
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: [PATCH 3/8] libstdc++: More efficient masked inc-/decrement implementation
2023-02-23 8:49 ` [PATCH 3/8] libstdc++: More efficient masked inc-/decrement implementation Matthias Kretz
@ 2023-02-24 17:12 ` Jonathan Wakely
0 siblings, 0 replies; 19+ messages in thread
From: Jonathan Wakely @ 2023-02-24 17:12 UTC (permalink / raw)
To: Matthias Kretz; +Cc: gcc-patches, libstdc++
On Thu, 23 Feb 2023 at 08:55, Matthias Kretz via Libstdc++
<libstdc++@gcc.gnu.org> wrote:
>
OK for trunk (and maybe backport later if you want to).
>
> Signed-off-by: Matthias Kretz <m.kretz@gsi.de>
>
> libstdc++-v3/ChangeLog:
>
> PR libstdc++/108856
> * include/experimental/bits/simd_builtin.h
> (_SimdImplBuiltin::_S_masked_unary): More efficient
> implementation of masked inc-/decrement for integers and floats
> without AVX2.
> * include/experimental/bits/simd_x86.h
> (_SimdImplX86::_S_masked_unary): New. Use AVX512 masked subtract
> builtins for masked inc-/decrement.
> ---
> .../include/experimental/bits/simd_builtin.h | 27 +++++++-
> .../include/experimental/bits/simd_x86.h | 68 +++++++++++++++++++
> 2 files changed, 93 insertions(+), 2 deletions(-)
>
>
> --
> ──────────────────────────────────────────────────────────────────────────
> Dr. Matthias Kretz https://mattkretz.github.io
> GSI Helmholtz Centre for Heavy Ion Research https://gsi.de
> stdₓ::simd
> ──────────────────────────────────────────────────────────────────────────
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: [PATCH 6/8] libstdc++: Fix formatting
2023-02-23 8:50 ` [PATCH 6/8] libstdc++: Fix formatting Matthias Kretz
@ 2023-02-24 17:14 ` Jonathan Wakely
2023-02-24 18:44 ` Matthias Kretz
0 siblings, 1 reply; 19+ messages in thread
From: Jonathan Wakely @ 2023-02-24 17:14 UTC (permalink / raw)
To: Matthias Kretz; +Cc: gcc-patches, libstdc++
On Thu, 23 Feb 2023 at 08:54, Matthias Kretz via Libstdc++
<libstdc++@gcc.gnu.org> wrote:
>
>
>
> Whitespace changes only.
Looks like there are a few remaining spaces that could be removed
where you've joined lines, e.g.
+ { return static_cast<_Up*>( __builtin_assume_aligned(__ptr,
_S_alignment<_Tp, _Up>)); }
and
+ { return __vector_type_t<_Tp, _Np>{
static_cast<_Tp>(__gen(_SizeConstant<_I>()))...}; }
OK for trunk anyway (and the branches if you want).
>
> Signed-off-by: Matthias Kretz <m.kretz@gsi.de>
>
> libstdc++-v3/ChangeLog:
>
> * include/experimental/bits/simd.h: Line breaks and indenting
> fixed to follow the libstdc++ standard.
> * include/experimental/bits/simd_builtin.h: Likewise.
> * include/experimental/bits/simd_fixed_size.h: Likewise.
> * include/experimental/bits/simd_neon.h: Likewise.
> * include/experimental/bits/simd_ppc.h: Likewise.
> * include/experimental/bits/simd_scalar.h: Likewise.
> * include/experimental/bits/simd_x86.h: Likewise.
> ---
> libstdc++-v3/include/experimental/bits/simd.h | 473 ++++++------
> .../include/experimental/bits/simd_builtin.h | 692 +++++++++---------
> .../experimental/bits/simd_fixed_size.h | 228 +++---
> .../include/experimental/bits/simd_neon.h | 24 +-
> .../include/experimental/bits/simd_ppc.h | 3 +-
> .../include/experimental/bits/simd_scalar.h | 362 +++++----
> .../include/experimental/bits/simd_x86.h | 90 ++-
> 7 files changed, 942 insertions(+), 930 deletions(-)
>
>
> --
> ──────────────────────────────────────────────────────────────────────────
> Dr. Matthias Kretz https://mattkretz.github.io
> GSI Helmholtz Centre for Heavy Ion Research https://gsi.de
> stdₓ::simd
> ──────────────────────────────────────────────────────────────────────────
^ permalink raw reply [flat|nested] 19+ messages in thread
* Re: [PATCH 6/8] libstdc++: Fix formatting
2023-02-24 17:14 ` Jonathan Wakely
@ 2023-02-24 18:44 ` Matthias Kretz
0 siblings, 0 replies; 19+ messages in thread
From: Matthias Kretz @ 2023-02-24 18:44 UTC (permalink / raw)
To: gcc-patches, libstdc++; +Cc: Jonathan Wakely
On Friday, 24 February 2023 18:14:53 CET Jonathan Wakely wrote:
> Looks like there are a few remaining spaces that could be removed
> where you've joined lines, e.g.
Fixed and pushed.
> OK for trunk anyway (and the branches if you want).
I'll likely backport after I backported all other patches to trunk that came
before this one.
--
──────────────────────────────────────────────────────────────────────────
Dr. Matthias Kretz https://mattkretz.github.io
GSI Helmholtz Centre for Heavy Ion Research https://gsi.de
stdₓ::simd
──────────────────────────────────────────────────────────────────────────
^ permalink raw reply [flat|nested] 19+ messages in thread
end of thread, other threads:[~2023-02-24 18:44 UTC | newest]
Thread overview: 19+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-02-23 8:48 [PATCH 0/8] std::experimental::simd patchset Matthias Kretz
2023-02-23 8:49 ` [PATCH 1/8] libstdc++: Simplify three helper functions into one Matthias Kretz
2023-02-23 11:05 ` Jonathan Wakely
2023-02-23 8:49 ` [PATCH 2/8] libstdc++: Fix simd build failure on clang Matthias Kretz
2023-02-23 11:06 ` Jonathan Wakely
2023-02-23 8:49 ` [PATCH 3/8] libstdc++: More efficient masked inc-/decrement implementation Matthias Kretz
2023-02-24 17:12 ` Jonathan Wakely
2023-02-23 8:49 ` [PATCH 4/8] libstdc++: Add missing constexpr on simd shift implementation Matthias Kretz
2023-02-23 11:07 ` Jonathan Wakely
2023-02-23 11:33 ` Matthias Kretz
2023-02-23 8:49 ` [PATCH 5/8] libstdc++: Always-inline most of non-cmath fixed_size implementation Matthias Kretz
2023-02-24 17:10 ` Jonathan Wakely
2023-02-23 8:50 ` [PATCH 6/8] libstdc++: Fix formatting Matthias Kretz
2023-02-24 17:14 ` Jonathan Wakely
2023-02-24 18:44 ` Matthias Kretz
2023-02-23 8:50 ` [PATCH 7/8] libstdc++: Fix -Wsign-compare issue Matthias Kretz
2023-02-23 11:07 ` Jonathan Wakely
2023-02-23 8:50 ` [PATCH 8/8] libstdc++: Test that integral simd reductions are precise Matthias Kretz
2023-02-23 11:08 ` Jonathan Wakely
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).