[libcxx-commits] [libcxx] [libcxxabi] Adding Separate OpenMP Offloading Backend to `libcxx/include/__algorithm/pstl_backends` (PR #66968)

Louis Dionne via libcxx-commits libcxx-commits at lists.llvm.org
Wed Mar 27 11:01:05 PDT 2024


================
@@ -0,0 +1,195 @@
+//===----------------------------------------------------------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef _LIBCPP___ALGORITHM_PSTL_BACKENDS_OPENMP_BACKEND_TRANSFORM_REDUCE_H
+#define _LIBCPP___ALGORITHM_PSTL_BACKENDS_OPENMP_BACKEND_TRANSFORM_REDUCE_H
+
+#include <__algorithm/pstl_backends/cpu_backends/backend.h>
+#include <__algorithm/pstl_backends/openmp/backend.h>
+#include <__algorithm/unwrap_iter.h>
+#include <__config>
+#include <__functional/operations.h>
+#include <__iterator/wrap_iter.h>
+#include <__numeric/transform_reduce.h>
+#include <__type_traits/is_arithmetic.h>
+#include <__type_traits/is_execution_policy.h>
+#include <__type_traits/is_trivially_copyable.h>
+#include <__type_traits/operation_traits.h>
+#include <__type_traits/remove_pointer.h>
+#include <__utility/move.h>
+#include <optional>
+
+#if !defined(_LIBCPP_HAS_NO_PRAGMA_SYSTEM_HEADER)
+#  pragma GCC system_header
+#endif
+
+_LIBCPP_PUSH_MACROS
+#include <__undef_macros>
+
+#if !defined(_LIBCPP_HAS_NO_INCOMPLETE_PSTL) && _LIBCPP_STD_VER >= 17
+
+_LIBCPP_BEGIN_NAMESPACE_STD
+
+//===----------------------------------------------------------------------===//
+// Templates for predefined reductions
+//===----------------------------------------------------------------------===//
+
+#  define __PSTL_OMP_SIMD_1_REDUCTION(omp_op, std_op)                                                                  \
+    template <class _Iterator,                                                                                         \
+              class _DifferenceType,                                                                                   \
+              typename _Tp,                                                                                            \
+              typename _BinaryOperationType,                                                                           \
+              typename _UnaryOperation>                                                                                \
+    _LIBCPP_HIDE_FROM_ABI _Tp __omp_transform_reduce(                                                                  \
+        _Iterator __first,                                                                                             \
+        _DifferenceType __n,                                                                                           \
+        _Tp __init,                                                                                                    \
+        std_op<_BinaryOperationType> __reduce,                                                                         \
+        _UnaryOperation __transform) noexcept {                                                                        \
+      __par_backend::__omp_map_to(__first, __n);                                                                       \
+_PSTL_PRAGMA(omp target teams distribute parallel for reduction(omp_op:__init))                                        \
+      for (_DifferenceType __i = 0; __i < __n; ++__i)                                                                  \
+        __init = __reduce(__init, __transform(*(__first + __i)));                                                      \
+      __par_backend::__omp_map_release(__first, __n);                                                                  \
+      return __init;                                                                                                   \
+    }
+
+#  define __PSTL_OMP_SIMD_2_REDUCTION(omp_op, std_op)                                                                  \
+    template <class _Iterator1,                                                                                        \
+              class _Iterator2,                                                                                        \
+              class _DifferenceType,                                                                                   \
+              typename _Tp,                                                                                            \
+              typename _BinaryOperationType,                                                                           \
+              typename _UnaryOperation >                                                                               \
+    _LIBCPP_HIDE_FROM_ABI _Tp __omp_transform_reduce(                                                                  \
+        _Iterator1 __first1,                                                                                           \
+        _Iterator2 __first2,                                                                                           \
+        _DifferenceType __n,                                                                                           \
+        _Tp __init,                                                                                                    \
+        std_op<_BinaryOperationType> __reduce,                                                                         \
+        _UnaryOperation __transform) noexcept {                                                                        \
+      __par_backend::__omp_map_to(__first1, __n);                                                                      \
+      __par_backend::__omp_map_to(__first2, __n);                                                                      \
+_PSTL_PRAGMA(omp target teams distribute parallel for reduction(omp_op:__init))                                        \
+      for (_DifferenceType __i = 0; __i < __n; ++__i)                                                                  \
+        __init = __reduce(__init, __transform(*(__first1 + __i), *(__first2 + __i)));                                  \
+      __par_backend::__omp_map_release(__first1, __n);                                                                 \
+      __par_backend::__omp_map_release(__first2, __n);                                                                 \
+      return __init;                                                                                                   \
+    }
+
+#  define __PSTL_OMP_SIMD_REDUCTION(omp_op, std_op)                                                                    \
+    __PSTL_OMP_SIMD_1_REDUCTION(omp_op, std_op)                                                                        \
+    __PSTL_OMP_SIMD_2_REDUCTION(omp_op, std_op)
+
+// Addition
+__PSTL_OMP_SIMD_REDUCTION(+, std::plus)
+
+// Subtraction
+__PSTL_OMP_SIMD_REDUCTION(-, std::minus)
+
+// Multiplication
+__PSTL_OMP_SIMD_REDUCTION(*, std::multiplies)
+
+// Logical and
+__PSTL_OMP_SIMD_REDUCTION(&&, std::logical_and)
+
+// Logical or
+__PSTL_OMP_SIMD_REDUCTION(||, std::logical_or)
+
+// Bitwise and
+__PSTL_OMP_SIMD_REDUCTION(&, std::bit_and)
+
+// Bitwise or
+__PSTL_OMP_SIMD_REDUCTION(|, std::bit_or)
+
+// Bitwise xor
+__PSTL_OMP_SIMD_REDUCTION(^, std::bit_xor)
+
+//===----------------------------------------------------------------------===//
+// The following struct is used to determine whether a reduction is supported by
+// the OpenMP backend.
+//===----------------------------------------------------------------------===//
+
+template <class _T1, class _T2, class _T3>
+struct __is_supported_reduction : std::false_type {};
+
+#  define __PSTL_IS_SUPPORTED_REDUCTION(funname)                                                                       \
----------------
ldionne wrote:

```suggestion
#  define _LIBCPP_PSTL_IS_SUPPORTED_REDUCTION(funname)                                                                       \
```

https://github.com/llvm/llvm-project/pull/66968


More information about the libcxx-commits mailing list