cuda-cccl 0.3.2__cp313-cp313-manylinux_2_24_aarch64.whl → 0.3.4__cp313-cp313-manylinux_2_24_aarch64.whl

cuda/cccl/headers/include/thrust/system/tbb/memory.h

@@ -1,18 +1,5 @@
-/*
- *  Copyright 2008-2018 NVIDIA Corporation
- *
- *  Licensed under the Apache License, Version 2.0 (the "License");
- *  you may not use this file except in ctbbliance with the License.
- *  You may obtain a copy of the License at
- *
- *      http://www.apache.org/licenses/LICENSE-2.0
- *
- *  Unless required by applicable law or agreed to in writing, software
- *  distributed under the License is distributed on an "AS IS" BASIS,
- *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- *  See the License for the specific language governing permissions and
- *  limitations under the License.
- */
+// SPDX-FileCopyrightText: Copyright (c) 2008-2018, NVIDIA Corporation. All rights reserved.
+// SPDX-License-Identifier: Apache-2.0
 
 /*! \file thrust/system/tbb/memory.h
  *  \brief Managing memory associated with Thrust's TBB system.
@@ -32,15 +19,39 @@
 #include <thrust/detail/type_traits.h>
 #include <thrust/memory.h>
 #include <thrust/mr/allocator.h>
+#include <thrust/system/cpp/detail/execution_policy.h>
+#include <thrust/system/cpp/memory.h>
 #include <thrust/system/tbb/memory_resource.h>
 
+#include <cuda/std/limits>
+
 #include <ostream>
 
 THRUST_NAMESPACE_BEGIN
-namespace system
+namespace system::tbb
 {
-namespace tbb
+namespace detail
+{
+// XXX circular #inclusion problems cause the compiler to believe that cpp::malloc
+//     is not defined
+//     WAR the problem by using adl to call cpp::malloc, which requires it to depend
+//     on a template parameter
+template <typename Tag>
+pointer<void> malloc_workaround(Tag t, std::size_t n)
+{
+  return pointer<void>(malloc(t, n));
+} // end malloc_workaround()
+
+// XXX circular #inclusion problems cause the compiler to believe that cpp::free
+//     is not defined
+//     WAR the problem by using adl to call cpp::free, which requires it to depend
+//     on a template parameter
+template <typename Tag>
+void free_workaround(Tag t, pointer<void> ptr)
 {
+  free(t, ptr.get());
+} // end free_workaround()
+} // namespace detail
 
 /*! Allocates an area of memory available to Thrust's <tt>tbb</tt> system.
  *  \param n Number of bytes to allocate.
@@ -52,7 +63,15 @@ namespace tbb
  *  \see tbb::free
  *  \see std::malloc
  */
-inline pointer<void> malloc(std::size_t n);
+inline pointer<void> malloc(std::size_t n)
+{
+  // XXX this is how we'd like to implement this function,
+  //     if not for circular #inclusion problems:
+  //
+  // return pointer<void>(thrust::system::cpp::malloc(n))
+  //
+  return detail::malloc_workaround(cpp::tag(), n);
+} // end malloc()
 
 /*! Allocates a typed area of memory available to Thrust's <tt>tbb</tt> system.
  *  \param n Number of elements to allocate.
@@ -65,7 +84,11 @@ inline pointer<void> malloc(std::size_t n);
  *  \see std::malloc
  */
 template <typename T>
-inline pointer<T> malloc(std::size_t n);
+inline pointer<T> malloc(std::size_t n)
+{
+  pointer<void> raw_ptr = thrust::system::tbb::malloc(sizeof(T) * n);
+  return pointer<T>(reinterpret_cast<T*>(raw_ptr.get()));
+} // end malloc()
 
 /*! Deallocates an area of memory previously allocated by <tt>tbb::malloc</tt>.
  *  \param ptr A <tt>tbb::pointer<void></tt> pointing to the beginning of an area
@@ -73,7 +96,15 @@ inline pointer<T> malloc(std::size_t n);
  *  \see tbb::malloc
  *  \see std::free
  */
-inline void free(pointer<void> ptr);
+inline void free(pointer<void> ptr)
+{
+  // XXX this is how we'd like to implement this function,
+  //     if not for circular #inclusion problems:
+  //
+  // thrust::system::cpp::free(ptr)
+  //
+  detail::free_workaround(cpp::tag(), ptr);
+} // end free()
 
 /*! \p tbb::allocator is the default allocator used by the \p tbb system's
  *  containers such as <tt>tbb::vector</tt> if no user-specified allocator is
@@ -91,8 +122,7 @@ using universal_allocator = thrust::mr::stateless_resource_allocator<T, thrust::
 template <typename T>
 using universal_host_pinned_allocator =
   thrust::mr::stateless_resource_allocator<T, thrust::system::tbb::universal_host_pinned_memory_resource>;
-} // namespace tbb
-} // namespace system
+} // namespace system::tbb
 
 /*! \namespace thrust::tbb
  *  \brief \p thrust::tbb is a top-level alias for thrust::system::tbb.
@@ -107,5 +137,3 @@ using thrust::system::tbb::universal_host_pinned_allocator;
 } // namespace tbb
 
 THRUST_NAMESPACE_END
-
-#include <thrust/system/tbb/detail/memory.inl>

cuda/cccl/headers/include/thrust/system/tbb/memory_resource.h

@@ -1,18 +1,5 @@
-/*
- *  Copyright 2018-2020 NVIDIA Corporation
- *
- *  Licensed under the Apache License, Version 2.0 (the "License");
- *  you may not use this file except in compliance with the License.
- *  You may obtain a copy of the License at
- *
- *      http://www.apache.org/licenses/LICENSE-2.0
- *
- *  Unless required by applicable law or agreed to in writing, software
- *  distributed under the License is distributed on an "AS IS" BASIS,
- *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- *  See the License for the specific language governing permissions and
- *  limitations under the License.
- */
+// SPDX-FileCopyrightText: Copyright (c) 2018-2020, NVIDIA Corporation. All rights reserved.
+// SPDX-License-Identifier: Apache-2.0
 
 /*! \file tbb/memory_resource.h
  *  \brief Memory resources for the TBB system.
@@ -34,11 +21,8 @@
 #include <thrust/system/tbb/pointer.h>
 
 THRUST_NAMESPACE_BEGIN
-namespace system
-{
-namespace tbb
+namespace system::tbb
 {
-
 //! \cond
 namespace detail
 {
@@ -68,8 +52,6 @@ using universal_host_pinned_memory_resource = universal_memory_resource;
 
 /*! \} // memory_resources
  */
-
-} // namespace tbb
-} // namespace system
+} // namespace system::tbb
 
 THRUST_NAMESPACE_END

cuda/cccl/headers/include/thrust/system/tbb/pointer.h

@@ -1,18 +1,5 @@
-/*
- *  Copyright 2008-2020 NVIDIA Corporation
- *
- *  Licensed under the Apache License, Version 2.0 (the "License");
- *  you may not use this file except in compliance with the License.
- *  You may obtain a copy of the License at
- *
- *      http://www.apache.org/licenses/LICENSE-2.0
- *
- *  Unless required by applicable law or agreed to in writing, software
- *  distributed under the License is distributed on an "AS IS" BASIS,
- *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- *  See the License for the specific language governing permissions and
- *  limitations under the License.
- */
+// SPDX-FileCopyrightText: Copyright (c) 2008-2020, NVIDIA Corporation. All rights reserved.
+// SPDX-License-Identifier: Apache-2.0
 
 /*! \file thrust/system/tbb/memory.h
  *  \brief Managing memory associated with Thrust's TBB system.
@@ -36,11 +23,8 @@
 #include <cuda/std/type_traits>
 
 THRUST_NAMESPACE_BEGIN
-namespace system
+namespace system::tbb
 {
-namespace tbb
-{
-
 /*! \p tbb::pointer stores a pointer to an object allocated in memory accessible
  *  by the \p tbb system. This type provides type safety when dispatching
  *  algorithms on ranges resident in \p tbb memory.
@@ -102,9 +86,7 @@ using universal_host_pinned_pointer = universal_pointer<T>;
  */
 template <typename T>
 using reference = thrust::tagged_reference<T, thrust::system::tbb::tag>;
-
-} // namespace tbb
-} // namespace system
+} // namespace system::tbb
 
 /*! \addtogroup system_backends Systems
  *  \ingroup system

cuda/cccl/headers/include/thrust/system/tbb/vector.h

@@ -1,18 +1,5 @@
-/*
- *  Copyright 2008-2013 NVIDIA Corporation
- *
- *  Licensed under the Apache License, Version 2.0 (the "License");
- *  you may not use this file except in compliance with the License.
- *  You may obtain a copy of the License at
- *
- *      http://www.apache.org/licenses/LICENSE-2.0
- *
- *  Unless required by applicable law or agreed to in writing, software
- *  distributed under the License is distributed on an "AS IS" BASIS,
- *  WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- *  See the License for the specific language governing permissions and
- *  limitations under the License.
- */
+// SPDX-FileCopyrightText: Copyright (c) 2008-2013, NVIDIA Corporation. All rights reserved.
+// SPDX-License-Identifier: Apache-2.0
 
 /*! \file thrust/system/tbb/vector.h
  *  \brief A dynamically-sizable array of elements which reside in memory available to
@@ -36,11 +23,8 @@
 #include <vector>
 
 THRUST_NAMESPACE_BEGIN
-namespace system
+namespace system::tbb
 {
-namespace tbb
-{
-
 /*! \p tbb::vector is a container that supports random access to elements,
  *  constant time removal of elements at the end, and linear time insertion
  *  and removal of elements at the beginning or in the middle. The number of
@@ -87,8 +71,7 @@ using universal_vector = thrust::detail::vector_base<T, Allocator>;
 //! \see universal_vector
 template <typename T>
 using universal_host_pinned_vector = thrust::detail::vector_base<T, universal_host_pinned_allocator<T>>;
-} // namespace tbb
-} // namespace system
+} // namespace system::tbb
 
 namespace tbb
 {

cuda/cccl/headers/include/thrust/transform.h

@@ -32,13 +32,24 @@
 
 #include <thrust/detail/execution_policy.h>
 #include <thrust/iterator/iterator_traits.h>
-#include <thrust/system/detail/adl/transform.h>
 #include <thrust/system/detail/generic/select_system.h>
-#include <thrust/system/detail/generic/transform.h>
-#include <thrust/transform.h>
 
 #include <cuda/std/iterator>
 
+// Include all active backend system implementations (generic, sequential, host and device)
+#include <thrust/system/detail/generic/transform.h>
+#include <thrust/system/detail/sequential/transform.h>
+#include __THRUST_HOST_SYSTEM_ALGORITH_DETAIL_HEADER_INCLUDE(transform.h)
+#include __THRUST_DEVICE_SYSTEM_ALGORITH_DETAIL_HEADER_INCLUDE(transform.h)
+
+// Some build systems need a hint to know which files we could include
+#if 0
+#  include <thrust/system/cpp/detail/transform.h>
+#  include <thrust/system/cuda/detail/transform.h>
+#  include <thrust/system/omp/detail/transform.h>
+#  include <thrust/system/tbb/detail/transform.h>
+#endif
+
 THRUST_NAMESPACE_BEGIN
 
 /*! \addtogroup algorithms

cuda/cccl/headers/include/thrust/type_traits/integer_sequence.h

@@ -114,7 +114,6 @@ using make_index_sequence = ::cuda::std::make_index_sequence<N>;
 
 namespace detail
 {
-
 /*! \brief Create a new \c integer_sequence containing the elements of \c
  *  Sequence0 followed by the elements of \c Sequence1. \c Sequence1::size() is
  *  added to each element from \c Sequence0 in the new sequence.
@@ -148,7 +147,6 @@ struct merge_and_renumber_reversed_integer_sequences_impl<integer_sequence<T, Is
 {
   using type = integer_sequence<T, (sizeof...(Is1) + Is0)..., Is1...>;
 };
-
 } // namespace detail
 
 /*! \endcond
@@ -207,7 +205,6 @@ using integer_sequence_push_back = typename detail::integer_sequence_push_back_i
 
 namespace detail
 {
-
 template <typename T, ::cuda::std::size_t N>
 struct make_reversed_integer_sequence_impl
 {
@@ -246,7 +243,6 @@ struct integer_sequence_push_back_impl<T, I0, integer_sequence<T, Is...>>
 };
 
 ///////////////////////////////////////////////////////////////////////////////
-
 } // namespace detail
 
 /*! \endcond

cuda/cccl/headers/include/thrust/type_traits/is_contiguous_iterator.h

@@ -111,7 +111,6 @@ inline constexpr bool is_contiguous_iterator_impl_v =
   ::cuda::std::contiguous_iterator<Iterator> || is_thrust_pointer_v<Iterator> || is_libcxx_wrap_iter_v<Iterator>
   || is_libstdcxx_normal_iterator_v<Iterator> || is_msvc_contiguous_iterator_v<Iterator>
   || proclaim_contiguous_iterator<Iterator>::value;
-
 } // namespace detail
 
 /*! \endcond

cuda/cccl/headers/include/thrust/type_traits/is_operator_less_or_greater_function_object.h

@@ -50,13 +50,11 @@ THRUST_NAMESPACE_BEGIN
 
 namespace detail
 {
-
 template <typename T>
 struct is_operator_less_function_object_impl;
 
 template <typename T>
 struct is_operator_greater_function_object_impl;
-
 } // namespace detail
 
 /*! \endcond
@@ -147,7 +145,6 @@ constexpr bool is_operator_less_or_greater_function_object_v = is_operator_less_
 
 namespace detail
 {
-
 template <typename T>
 struct is_operator_less_function_object_impl : false_type
 {};
@@ -167,7 +164,6 @@ struct is_operator_greater_function_object_impl<::cuda::std::greater<T>> : true_
 template <typename T>
 struct is_operator_greater_function_object_impl<std::greater<T>> : true_type
 {};
-
 } // namespace detail
 
 /*! \endcond

cuda/cccl/headers/include/thrust/type_traits/is_operator_plus_function_object.h

@@ -50,10 +50,8 @@ THRUST_NAMESPACE_BEGIN
 
 namespace detail
 {
-
 template <typename T>
 struct is_operator_plus_function_object_impl;
-
 } // namespace detail
 
 /*! \endcond
@@ -91,7 +89,6 @@ constexpr bool is_operator_plus_function_object_v = is_operator_plus_function_ob
 
 namespace detail
 {
-
 template <typename T>
 struct is_operator_plus_function_object_impl : false_type
 {};
@@ -101,7 +98,6 @@ struct is_operator_plus_function_object_impl<::cuda::std::plus<T>> : true_type
 template <typename T>
 struct is_operator_plus_function_object_impl<std::plus<T>> : true_type
 {};
-
 } // namespace detail
 
 /*! \endcond

cuda/cccl/headers/include/thrust/type_traits/is_trivially_relocatable.h

@@ -57,10 +57,8 @@ THRUST_NAMESPACE_BEGIN
 
 namespace detail
 {
-
 template <typename T>
 struct is_trivially_relocatable_impl;
-
 } // namespace detail
 
 /*! \endcond
@@ -208,7 +206,6 @@ struct proclaim_trivially_relocatable : false_type
 
 namespace detail
 {
-
 // https://wg21.link/P1144R0#wording-inheritance
 template <typename T>
 struct is_trivially_relocatable_impl
@@ -218,7 +215,6 @@ struct is_trivially_relocatable_impl
 template <typename T, ::cuda::std::size_t N>
 struct is_trivially_relocatable_impl<T[N]> : is_trivially_relocatable_impl<T>
 {};
-
 } // namespace detail
 
 THRUST_NAMESPACE_END

cuda/cccl/headers/include/thrust/universal_allocator.h

@@ -31,6 +31,14 @@
 #  pragma system_header
 #endif // no system header
 
+// Some build systems need a hint to know which files we could include
+#if 0
+#  include <thrust/system/cpp/memory.h>
+#  include <thrust/system/cuda/memory.h>
+#  include <thrust/system/omp/memory.h>
+#  include <thrust/system/tbb/memory.h>
+#endif
+
 // #include the device system's vector header
 #define __THRUST_DEVICE_SYSTEM_MEMORY_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/memory.h>
 #include __THRUST_DEVICE_SYSTEM_MEMORY_HEADER

cuda/cccl/headers/include/thrust/universal_vector.h

@@ -30,8 +30,17 @@
 #elif defined(_CCCL_IMPLICIT_SYSTEM_HEADER_MSVC)
 #  pragma system_header
 #endif // no system header
+
 #include <thrust/universal_allocator.h>
 
+// Some build systems need a hint to know which files we could include
+#if 0
+#  include <thrust/system/cpp/vector.h>
+#  include <thrust/system/cuda/vector.h>
+#  include <thrust/system/omp/vector.h>
+#  include <thrust/system/tbb/vector.h>
+#endif
+
 // #include the device system's vector header
 #define __THRUST_DEVICE_SYSTEM_VECTOR_HEADER <__THRUST_DEVICE_SYSTEM_ROOT/vector.h>
 #include __THRUST_DEVICE_SYSTEM_VECTOR_HEADER

cuda/cccl/headers/include/thrust/zip_function.h

@@ -18,10 +18,9 @@
 #endif // no system header
 
 #include <thrust/detail/type_deduction.h>
-#include <thrust/tuple.h>
-#include <thrust/type_traits/integer_sequence.h>
 
 #include <cuda/functional>
+#include <cuda/std/tuple>
 
 THRUST_NAMESPACE_BEGIN
 
@@ -34,29 +33,6 @@ THRUST_NAMESPACE_BEGIN
  *  \{
  */
 
-namespace detail
-{
-namespace zip_detail
-{
-
-// Add workaround for decltype(auto) on C++11-only compilers:
-_CCCL_EXEC_CHECK_DISABLE
-template <typename Function, typename Tuple, std::size_t... Is>
-_CCCL_HOST_DEVICE decltype(auto) apply_impl(Function&& func, Tuple&& args, index_sequence<Is...>)
-{
-  return func(thrust::get<Is>(THRUST_FWD(args))...);
-}
-
-template <typename Function, typename Tuple>
-_CCCL_HOST_DEVICE decltype(auto) apply(Function&& func, Tuple&& args)
-{
-  constexpr auto tuple_size = thrust::tuple_size<::cuda::std::decay_t<Tuple>>::value;
-  return apply_impl(THRUST_FWD(func), THRUST_FWD(args), make_index_sequence<tuple_size>{});
-}
-
-} // namespace zip_detail
-} // namespace detail
-
 /*! \p zip_function is a function object that allows the easy use of N-ary
  *  function objects with \p zip_iterators without redefining them to take a
  *  \p tuple instead of N arguments.
@@ -129,12 +105,10 @@ public:
       : func(::cuda::std::move(func))
   {}
 
-  // Add workaround for decltype(auto) on C++11-only compilers:
-
   template <typename Tuple>
   _CCCL_HOST_DEVICE decltype(auto) operator()(Tuple&& args) const
   {
-    return detail::zip_detail::apply(func, THRUST_FWD(args));
+    return ::cuda::std::apply(func, ::cuda::std::forward<Tuple>(args));
   }
 
   //! Returns a reference to the underlying function.

cuda/compute/__init__.py

@@ -17,6 +17,7 @@ from .algorithms import (
     make_radix_sort,
     make_reduce_into,
     make_segmented_reduce,
+    make_segmented_sort,
     make_three_way_partition,
     make_unary_transform,
     make_unique_by_key,
@@ -24,6 +25,7 @@ from .algorithms import (
     radix_sort,
     reduce_into,
     segmented_reduce,
+    segmented_sort,
     three_way_partition,
     unary_transform,
     unique_by_key,
@@ -59,6 +61,7 @@ __all__ = [
     "make_radix_sort",
     "make_reduce_into",
     "make_segmented_reduce",
+    "make_segmented_sort",
     "make_three_way_partition",
     "make_unary_transform",
     "make_unique_by_key",
@@ -69,6 +72,7 @@ __all__ = [
     "reduce_into",
     "ReverseIterator",
     "segmented_reduce",
+    "segmented_sort",
     "SortOrder",
     "TransformIterator",
     "TransformOutputIterator",

cuda/compute/_bindings.pyi

@@ -26,10 +26,10 @@ class OpKind(IntEnum):
     BIT_OR = ...
     BIT_XOR = ...
     BIT_NOT = ...
-    # IDENTITY = ...
+    IDENTITY = ...
     NEGATE = ...
-    # MINIMUM = ...
-    # MAXIMUM = ...
+    MINIMUM = ...
+    MAXIMUM = ...
 
 class TypeEnum(IntEnum):
     _value_: int
@@ -447,6 +447,29 @@ class DeviceHistogramBuildResult:
         stream,
     ) -> None: ...
 
+# -----------------
+# DeviceSegmentedSort
+# -----------------
+
+class DeviceSegmentedSortBuildResult:
+    def __init__(self): ...
+    def compute(
+        self,
+        temp_storage_ptr: int | None,
+        temp_storage_nbytes: int,
+        d_in_keys: Iterator,
+        d_out_keys: Iterator,
+        d_in_values: Iterator,
+        d_out_values: Iterator,
+        num_items: int,
+        num_segments: int,
+        d_begin_offsets: Iterator,
+        d_end_offsets: Iterator,
+        is_overwrite_okay: bool,
+        selector: int,
+        stream,
+    ) -> tuple[int, int]: ...
+
 # ---------------------
 # DeviceThreeWayPartition
 # ---------------------