cuda-cccl 0.1.3.2.0.dev438__cp313-cp313-manylinux_2_24_aarch64.whl → 0.3.1__cp313-cp313-manylinux_2_24_aarch64.whl

cuda/compute/__init__.py

@@ -0,0 +1,77 @@
+# Copyright (c) 2024, NVIDIA CORPORATION & AFFILIATES. ALL RIGHTS RESERVED.
+#
+# SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+
+from .algorithms import (
+    DoubleBuffer,
+    SortOrder,
+    binary_transform,
+    exclusive_scan,
+    histogram_even,
+    inclusive_scan,
+    make_binary_transform,
+    make_exclusive_scan,
+    make_histogram_even,
+    make_inclusive_scan,
+    make_merge_sort,
+    make_radix_sort,
+    make_reduce_into,
+    make_segmented_reduce,
+    make_three_way_partition,
+    make_unary_transform,
+    make_unique_by_key,
+    merge_sort,
+    radix_sort,
+    reduce_into,
+    segmented_reduce,
+    three_way_partition,
+    unary_transform,
+    unique_by_key,
+)
+from .iterators import (
+    CacheModifiedInputIterator,
+    ConstantIterator,
+    CountingIterator,
+    ReverseIterator,
+    TransformIterator,
+    TransformOutputIterator,
+    ZipIterator,
+)
+from .op import OpKind
+from .struct import gpu_struct
+
+__all__ = [
+    "binary_transform",
+    "CacheModifiedInputIterator",
+    "ConstantIterator",
+    "CountingIterator",
+    "DoubleBuffer",
+    "exclusive_scan",
+    "gpu_struct",
+    "histogram_even",
+    "inclusive_scan",
+    "make_binary_transform",
+    "make_exclusive_scan",
+    "make_histogram_even",
+    "make_inclusive_scan",
+    "make_merge_sort",
+    "make_radix_sort",
+    "make_reduce_into",
+    "make_segmented_reduce",
+    "make_three_way_partition",
+    "make_unary_transform",
+    "make_unique_by_key",
+    "merge_sort",
+    "OpKind",
+    "radix_sort",
+    "reduce_into",
+    "ReverseIterator",
+    "segmented_reduce",
+    "SortOrder",
+    "TransformIterator",
+    "TransformOutputIterator",
+    "three_way_partition",
+    "unary_transform",
+    "unique_by_key",
+    "ZipIterator",
+]

cuda/{cccl/parallel/experimental → compute}/_bindings.pyi

@@ -390,6 +390,7 @@ class DeviceHistogramBuildResult:
         num_rows: int,
         row_stride_samples: int,
         is_evenly_segmented: bool,
+        info: CommonData,
     ): ...
     def compute_even(
         self,
@@ -403,3 +404,30 @@ class DeviceHistogramBuildResult:
         row_stride_samples: int,
         stream,
     ) -> None: ...
+
+# ---------------------
+# DeviceThreeWayPartition
+# ---------------------
+
+class DeviceThreeWayPartitionBuildResult:
+    def __init__(
+        self,
+        d_in: Iterator,
+        d_first_part_out: Iterator,
+        d_second_part_out: Iterator,
+        d_unselected_out: Iterator,
+        d_num_selected_out: Iterator,
+        select_first_part_op: Op,
+        select_second_part_op: Op,
+        info: CommonData,
+    ): ...
+    def compute(
+        self,
+        d_in: Iterator,
+        d_first_part_out: Iterator,
+        d_second_part_out: Iterator,
+        d_unselected_out: Iterator,
+        d_num_selected_out: Iterator,
+        num_items: int,
+        stream,
+    ) -> int: ...

cuda/{cccl/parallel/experimental → compute}/_bindings_impl.pyx

@@ -4,7 +4,7 @@
 
 # Python signatures are declared in the companion Python stub file _bindings.pyi
 # Make sure to update PYI with change to Python API to ensure that Python
-# static type checker tools like mypy green-lights cuda.cccl.parallel
+# static type checker tools like mypy green-lights cuda.compute
 
 from libc.string cimport memset, memcpy
 from libc.stdint cimport uint8_t, uint32_t, uint64_t, int64_t, uintptr_t
@@ -1982,3 +1982,143 @@ cdef class DeviceHistogramBuildResult:
             <const char*>self.build_data.cubin,
             self.build_data.cubin_size
         )
+
+
+# ----------------------------------
+# DeviceThreeWayPartitionBuildResult
+# ----------------------------------
+cdef extern from "cccl/c/three_way_partition.h":
+    cdef struct cccl_device_three_way_partition_build_result_t 'cccl_device_three_way_partition_build_result_t':
+        const char* cubin
+        size_t cubin_size
+
+    cdef CUresult cccl_device_three_way_partition_build(
+        cccl_device_three_way_partition_build_result_t *build_ptr,
+        cccl_iterator_t d_in,
+        cccl_iterator_t d_first_part_out,
+        cccl_iterator_t d_second_part_out,
+        cccl_iterator_t d_unselected_out,
+        cccl_iterator_t d_num_selected_out,
+        cccl_op_t select_first_part_op,
+        cccl_op_t select_second_part_op,
+        int, int, const char *, const char *, const char *, const char *
+    ) nogil
+
+    CUresult cccl_device_three_way_partition(
+        cccl_device_three_way_partition_build_result_t build,
+        void* d_temp_storage,
+        size_t* temp_storage_bytes,
+        cccl_iterator_t d_in,
+        cccl_iterator_t d_first_part_out,
+        cccl_iterator_t d_second_part_out,
+        cccl_iterator_t d_unselected_out,
+        cccl_iterator_t d_num_selected_out,
+        cccl_op_t select_first_part_op,
+        cccl_op_t select_second_part_op,
+        int64_t num_items,
+        CUstream stream
+    ) nogil
+
+    cdef CUresult cccl_device_three_way_partition_cleanup(
+        cccl_device_three_way_partition_build_result_t *build_ptr
+    ) nogil
+
+
+cdef class DeviceThreeWayPartitionBuildResult:
+    cdef cccl_device_three_way_partition_build_result_t build_data
+
+    def __dealloc__(DeviceThreeWayPartitionBuildResult self):
+        cdef CUresult status = -1
+        with nogil:
+            status = cccl_device_three_way_partition_cleanup(&self.build_data)
+        if (status != 0):
+            print(f"Return code {status} encountered during three_way_partition result cleanup")
+
+
+    def __cinit__(
+        DeviceThreeWayPartitionBuildResult self,
+        Iterator d_in,
+        Iterator d_first_part_out,
+        Iterator d_second_part_out,
+        Iterator d_unselected_out,
+        Iterator d_num_selected_out,
+        Op select_first_part_op,
+        Op select_second_part_op,
+        CommonData common_data
+    ):
+        cdef CUresult status = -1
+        cdef int cc_major = common_data.get_cc_major()
+        cdef int cc_minor = common_data.get_cc_minor()
+        cdef const char *cub_path = common_data.cub_path_get_c_str()
+        cdef const char *thrust_path = common_data.thrust_path_get_c_str()
+        cdef const char *libcudacxx_path = common_data.libcudacxx_path_get_c_str()
+        cdef const char *ctk_path = common_data.ctk_path_get_c_str()
+
+        memset(&self.build_data, 0, sizeof(cccl_device_three_way_partition_build_result_t))
+        with nogil:
+            status = cccl_device_three_way_partition_build(
+                &self.build_data,
+                d_in.iter_data,
+                d_first_part_out.iter_data,
+                d_second_part_out.iter_data,
+                d_unselected_out.iter_data,
+                d_num_selected_out.iter_data,
+                select_first_part_op.op_data,
+                select_second_part_op.op_data,
+                cc_major,
+                cc_minor,
+                cub_path,
+                thrust_path,
+                libcudacxx_path,
+                ctk_path,
+            )
+        if status != 0:
+            raise RuntimeError(
+                f"Failed building three_way_partition, error code: {status}"
+            )
+
+    cpdef int compute(
+        DeviceThreeWayPartitionBuildResult self,
+        temp_storage_ptr,
+        temp_storage_bytes,
+        Iterator d_in,
+        Iterator d_first_part_out,
+        Iterator d_second_part_out,
+        Iterator d_unselected_out,
+        Iterator d_num_selected_out,
+        Op select_first_part_op,
+        Op select_second_part_op,
+        size_t num_items,
+        stream
+    ):
+        cdef CUresult status = -1
+        cdef void *storage_ptr = (<void *><uintptr_t>temp_storage_ptr) if temp_storage_ptr else NULL
+        cdef size_t storage_sz = <size_t>temp_storage_bytes
+        cdef CUstream c_stream = <CUstream><uintptr_t>(stream) if stream else NULL
+
+        with nogil:
+            status = cccl_device_three_way_partition(
+                self.build_data,
+                storage_ptr,
+                &storage_sz,
+                d_in.iter_data,
+                d_first_part_out.iter_data,
+                d_second_part_out.iter_data,
+                d_unselected_out.iter_data,
+                d_num_selected_out.iter_data,
+                select_first_part_op.op_data,
+                select_second_part_op.op_data,
+                <uint64_t>num_items,
+                c_stream
+            )
+        if status != 0:
+            raise RuntimeError(
+                f"Failed executing three_way_partition, error code: {status}"
+            )
+        return storage_sz
+
+    def _get_cubin(self):
+        return PyBytes_FromStringAndSize(
+            <const char*>self.build_data.cubin,
+            self.build_data.cubin_size
+        )

cuda/{cccl/parallel/experimental → compute}/algorithms/__init__.py

@@ -18,6 +18,8 @@ from ._scan import make_exclusive_scan as make_exclusive_scan
 from ._scan import make_inclusive_scan as make_inclusive_scan
 from ._segmented_reduce import make_segmented_reduce as make_segmented_reduce
 from ._segmented_reduce import segmented_reduce
+from ._three_way_partition import make_three_way_partition as make_three_way_partition
+from ._three_way_partition import three_way_partition as three_way_partition
 from ._transform import binary_transform, unary_transform
 from ._transform import make_binary_transform as make_binary_transform
 from ._transform import make_unary_transform as make_unary_transform
@@ -45,6 +47,8 @@ __all__ = [
     "make_segmented_reduce",
     "unique_by_key",
     "make_unique_by_key",
+    "three_way_partition",
+    "make_three_way_partition",
     "DoubleBuffer",
     "SortOrder",
 ]

cuda/{cccl/parallel/experimental → compute}/algorithms/_histogram.py

@@ -148,7 +148,7 @@ def make_histogram_even(
     Example:
         Below, ``make_histogram_even`` is used to create a histogram object that can be reused.
 
-        .. literalinclude:: ../../python/cuda_cccl/tests/parallel/examples/histogram/histogram_object.py
+        .. literalinclude:: ../../python/cuda_cccl/tests/compute/examples/histogram/histogram_object.py
           :language: python
           :start-after: # example-begin
 
@@ -190,7 +190,7 @@ def histogram_even(
     Example:
         Below, ``histogram_even`` is used to compute a histogram with evenly-spaced bins.
 
-        .. literalinclude:: ../../python/cuda_cccl/tests/parallel/examples/histogram/histogram_even_basic.py
+        .. literalinclude:: ../../python/cuda_cccl/tests/compute/examples/histogram/histogram_even_basic.py
             :language: python
             :start-after: # example-begin
             :caption: Basic histogram example.

cuda/{cccl/parallel/experimental → compute}/algorithms/_merge_sort.py

@@ -166,7 +166,7 @@ def make_merge_sort(
     Example:
         Below, ``make_merge_sort`` is used to create a merge sort object that can be reused.
 
-        .. literalinclude:: ../../python/cuda_cccl/tests/parallel/examples/sort/merge_sort_object.py
+        .. literalinclude:: ../../python/cuda_cccl/tests/compute/examples/sort/merge_sort_object.py
           :language: python
           :start-after: # example-begin
 
@@ -201,7 +201,7 @@ def merge_sort(
     Example:
         Below, ``merge_sort`` is used to sort a sequence of keys inplace. It also rearranges the items according to the keys' order.
 
-        .. literalinclude:: ../../python/cuda_cccl/tests/parallel/examples/sort/merge_sort_basic.py
+        .. literalinclude:: ../../python/cuda_cccl/tests/compute/examples/sort/merge_sort_basic.py
             :language: python
             :start-after: # example-begin
 

cuda/{cccl/parallel/experimental → compute}/algorithms/_radix_sort.py

@@ -222,7 +222,7 @@ def make_radix_sort(
     Example:
         Below, ``make_radix_sort`` is used to create a radix sort object that can be reused.
 
-        .. literalinclude:: ../../python/cuda_cccl/tests/parallel/examples/sort/radix_sort_object.py
+        .. literalinclude:: ../../python/cuda_cccl/tests/compute/examples/sort/radix_sort_object.py
           :language: python
           :start-after: # example-begin
 
@@ -259,14 +259,14 @@ def radix_sort(
     Example:
         Below, ``radix_sort`` is used to sort a sequence of keys. It also rearranges the values according to the keys' order.
 
-        .. literalinclude:: ../../python/cuda_cccl/tests/parallel/examples/sort/radix_sort_basic.py
+        .. literalinclude:: ../../python/cuda_cccl/tests/compute/examples/sort/radix_sort_basic.py
             :language: python
             :start-after: # example-begin
 
 
         In the following example, ``radix_sort`` is used to sort a sequence of keys with a ``DoubleBuffer` for reduced temporary storage.
 
-        .. literalinclude:: ../../python/cuda_cccl/tests/parallel/examples/sort/radix_sort_buffer.py
+        .. literalinclude:: ../../python/cuda_cccl/tests/compute/examples/sort/radix_sort_buffer.py
             :language: python
             :start-after: # example-begin
 

cuda/{cccl/parallel/experimental → compute}/algorithms/_reduce.py

@@ -3,8 +3,6 @@
 #
 # SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 
-from __future__ import annotations  # TODO: required for Python 3.7 docs env
-
 from typing import Callable, Union
 
 import numba
@@ -132,7 +130,7 @@ def make_reduce_into(
     Example:
         Below, ``make_reduce_into`` is used to create a reduction object that can be reused.
 
-        .. literalinclude:: ../../python/cuda_cccl/tests/parallel/examples/reduction/reduce_object.py
+        .. literalinclude:: ../../python/cuda_cccl/tests/compute/examples/reduction/reduce_object.py
             :language: python
             :start-after: # example-begin
 
@@ -165,7 +163,7 @@ def reduce_into(
     Example:
         Below, ``reduce_into`` is used to compute the sum of a sequence of integers.
 
-        .. literalinclude:: ../../python/cuda_cccl/tests/parallel/examples/reduction/sum_reduction.py
+        .. literalinclude:: ../../python/cuda_cccl/tests/compute/examples/reduction/sum_reduction.py
             :language: python
             :start-after: # example-begin
 

cuda/{cccl/parallel/experimental → compute}/algorithms/_scan.py

@@ -3,8 +3,6 @@
 #
 # SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 
-from __future__ import annotations  # TODO: required for Python 3.7 docs env
-
 from typing import Callable, Union
 
 import numba
@@ -143,7 +141,7 @@ def make_exclusive_scan(
     Example:
         Below, ``make_exclusive_scan`` is used to create an exclusive scan object that can be reused.
 
-        .. literalinclude:: ../../python/cuda_cccl/tests/parallel/examples/scan/exclusive_scan_object.py
+        .. literalinclude:: ../../python/cuda_cccl/tests/compute/examples/scan/exclusive_scan_object.py
           :language: python
           :start-after: # example-begin
 
@@ -176,7 +174,7 @@ def exclusive_scan(
     Example:
         Below, ``exclusive_scan`` is used to compute an exclusive scan with max operation.
 
-        .. literalinclude:: ../../python/cuda_cccl/tests/parallel/examples/scan/exclusive_scan_max.py
+        .. literalinclude:: ../../python/cuda_cccl/tests/compute/examples/scan/exclusive_scan_max.py
             :language: python
             :start-after: # example-begin
 
@@ -209,7 +207,7 @@ def make_inclusive_scan(
     Example:
         Below, ``make_inclusive_scan`` is used to create an inclusive scan object that can be reused.
 
-        .. literalinclude:: ../../python/cuda_cccl/tests/parallel/examples/scan/inclusive_scan_object.py
+        .. literalinclude:: ../../python/cuda_cccl/tests/compute/examples/scan/inclusive_scan_object.py
           :language: python
           :start-after: # example-begin
 
@@ -242,7 +240,7 @@ def inclusive_scan(
     Example:
         Below, ``inclusive_scan`` is used to compute an inclusive scan (prefix sum).
 
-        .. literalinclude:: ../../python/cuda_cccl/tests/parallel/examples/scan/inclusive_scan_custom.py
+        .. literalinclude:: ../../python/cuda_cccl/tests/compute/examples/scan/inclusive_scan_custom.py
             :language: python
             :start-after: # example-begin
 

cuda/{cccl/parallel/experimental → compute}/algorithms/_segmented_reduce.py

@@ -179,7 +179,7 @@ def make_segmented_reduce(
     Example:
         Below, ``make_segmented_reduce`` is used to create a segmented reduction object that can be reused.
 
-        .. literalinclude:: ../../python/cuda_cccl/tests/parallel/examples/segmented/segmented_reduce_object.py
+        .. literalinclude:: ../../python/cuda_cccl/tests/compute/examples/segmented/segmented_reduce_object.py
             :language: python
             :start-after: # example-begin
 
@@ -216,7 +216,7 @@ def segmented_reduce(
     Example:
         Below, ``segmented_reduce`` is used to compute the minimum value of segments in a sequence of integers.
 
-        .. literalinclude:: ../../python/cuda_cccl/tests/parallel/examples/segmented/segmented_reduce_basic.py
+        .. literalinclude:: ../../python/cuda_cccl/tests/compute/examples/segmented/segmented_reduce_basic.py
             :language: python
             :start-after: # example-begin