pyopencl 2025.2.5__cp312-cp312-manylinux_2_27_x86_64.manylinux_2_28_x86_64.whl

pyopencl/_cl.pyi

@@ -0,0 +1,2006 @@
+__copyright__ = "Copyright (C) 2025 University of Illinois Board of Trustees"
+
+__license__ = """
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in
+all copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
+THE SOFTWARE.
+"""
+
+from collections.abc import Callable, Sequence
+from enum import IntEnum, auto
+from typing import TYPE_CHECKING, Generic, Literal, overload
+
+import numpy as np
+from numpy.typing import NDArray
+from typing_extensions import override
+
+import pyopencl._monkeypatch
+from pyopencl.typing import DTypeT, HasBufferInterface, KernelArg, SVMInnerT, WaitList
+
+if TYPE_CHECKING:
+    from pyopencl import Program
+
+class _ErrorRecord:
+    def __init__(self, msg: str, code: int, routine: str) -> None: ...
+    def routine(self) -> str: ...
+    def code(self) -> int: ...
+    def what(self) -> str: ...
+    def is_out_of_memory(self) -> bool: ...
+
+class Error(Exception):
+    code: int
+    routine: str
+    what: _ErrorRecord
+
+    __str__ = pyopencl._monkeypatch.error_str
+
+class MemoryError(Error):
+    pass
+
+class LogicError(Error):
+    pass
+
+class RuntimeError(Error):
+    pass
+
+class status_code(IntEnum):  # noqa: N801
+    SUCCESS = auto()
+    DEVICE_NOT_FOUND = auto()
+    DEVICE_NOT_AVAILABLE = auto()
+    COMPILER_NOT_AVAILABLE = auto()
+    MEM_OBJECT_ALLOCATION_FAILURE = auto()
+    OUT_OF_RESOURCES = auto()
+    OUT_OF_HOST_MEMORY = auto()
+    PROFILING_INFO_NOT_AVAILABLE = auto()
+    MEM_COPY_OVERLAP = auto()
+    IMAGE_FORMAT_MISMATCH = auto()
+    IMAGE_FORMAT_NOT_SUPPORTED = auto()
+    BUILD_PROGRAM_FAILURE = auto()
+    MAP_FAILURE = auto()
+    INVALID_VALUE = auto()
+    INVALID_DEVICE_TYPE = auto()
+    INVALID_PLATFORM = auto()
+    INVALID_DEVICE = auto()
+    INVALID_CONTEXT = auto()
+    INVALID_QUEUE_PROPERTIES = auto()
+    INVALID_COMMAND_QUEUE = auto()
+    INVALID_HOST_PTR = auto()
+    INVALID_MEM_OBJECT = auto()
+    INVALID_IMAGE_FORMAT_DESCRIPTOR = auto()
+    INVALID_IMAGE_SIZE = auto()
+    INVALID_SAMPLER = auto()
+    INVALID_BINARY = auto()
+    INVALID_BUILD_OPTIONS = auto()
+    INVALID_PROGRAM = auto()
+    INVALID_PROGRAM_EXECUTABLE = auto()
+    INVALID_KERNEL_NAME = auto()
+    INVALID_KERNEL_DEFINITION = auto()
+    INVALID_KERNEL = auto()
+    INVALID_ARG_INDEX = auto()
+    INVALID_ARG_VALUE = auto()
+    INVALID_ARG_SIZE = auto()
+    INVALID_KERNEL_ARGS = auto()
+    INVALID_WORK_DIMENSION = auto()
+    INVALID_WORK_GROUP_SIZE = auto()
+    INVALID_WORK_ITEM_SIZE = auto()
+    INVALID_GLOBAL_OFFSET = auto()
+    INVALID_EVENT_WAIT_LIST = auto()
+    INVALID_EVENT = auto()
+    INVALID_OPERATION = auto()
+    INVALID_GL_OBJECT = auto()
+    INVALID_BUFFER_SIZE = auto()
+    INVALID_MIP_LEVEL = auto()
+    PLATFORM_NOT_FOUND_KHR = auto()
+    MISALIGNED_SUB_BUFFER_OFFSET = auto()
+    EXEC_STATUS_ERROR_FOR_EVENTS_IN_WAIT_LIST = auto()
+    INVALID_GLOBAL_WORK_SIZE = auto()
+    COMPILE_PROGRAM_FAILURE = auto()
+    LINKER_NOT_AVAILABLE = auto()
+    LINK_PROGRAM_FAILURE = auto()
+    DEVICE_PARTITION_FAILED = auto()
+    KERNEL_ARG_INFO_NOT_AVAILABLE = auto()
+    INVALID_IMAGE_DESCRIPTOR = auto()
+    INVALID_COMPILER_OPTIONS = auto()
+    INVALID_LINKER_OPTIONS = auto()
+    INVALID_DEVICE_PARTITION_COUNT = auto()
+    INVALID_PIPE_SIZE = auto()
+    INVALID_DEVICE_QUEUE = auto()
+    INVALID_SPEC_ID = auto()
+    MAX_SIZE_RESTRICTION_EXCEEDED = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class platform_info(IntEnum):  # noqa: N801
+    PROFILE = auto()
+    VERSION = auto()
+    NAME = auto()
+    VENDOR = auto()
+    EXTENSIONS = auto()
+    HOST_TIMER_RESOLUTION = auto()
+    NUMERIC_VERSION = auto()
+    EXTENSIONS_WITH_VERSION = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class device_type(IntEnum):  # noqa: N801
+    DEFAULT = auto()
+    CPU = auto()
+    GPU = auto()
+    ACCELERATOR = auto()
+    CUSTOM = auto()
+    ALL = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class device_info(IntEnum):  # noqa: N801
+    TYPE = auto()
+    VENDOR_ID = auto()
+    MAX_COMPUTE_UNITS = auto()
+    MAX_WORK_ITEM_DIMENSIONS = auto()
+    MAX_WORK_GROUP_SIZE = auto()
+    MAX_WORK_ITEM_SIZES = auto()
+    PREFERRED_VECTOR_WIDTH_CHAR = auto()
+    PREFERRED_VECTOR_WIDTH_SHORT = auto()
+    PREFERRED_VECTOR_WIDTH_INT = auto()
+    PREFERRED_VECTOR_WIDTH_LONG = auto()
+    PREFERRED_VECTOR_WIDTH_FLOAT = auto()
+    PREFERRED_VECTOR_WIDTH_DOUBLE = auto()
+    MAX_CLOCK_FREQUENCY = auto()
+    ADDRESS_BITS = auto()
+    MAX_READ_IMAGE_ARGS = auto()
+    MAX_WRITE_IMAGE_ARGS = auto()
+    MAX_MEM_ALLOC_SIZE = auto()
+    IMAGE2D_MAX_WIDTH = auto()
+    IMAGE2D_MAX_HEIGHT = auto()
+    IMAGE3D_MAX_WIDTH = auto()
+    IMAGE3D_MAX_HEIGHT = auto()
+    IMAGE3D_MAX_DEPTH = auto()
+    IMAGE_SUPPORT = auto()
+    MAX_PARAMETER_SIZE = auto()
+    MAX_SAMPLERS = auto()
+    MEM_BASE_ADDR_ALIGN = auto()
+    MIN_DATA_TYPE_ALIGN_SIZE = auto()
+    SINGLE_FP_CONFIG = auto()
+    DOUBLE_FP_CONFIG = auto()
+    HALF_FP_CONFIG = auto()
+    GLOBAL_MEM_CACHE_TYPE = auto()
+    GLOBAL_MEM_CACHELINE_SIZE = auto()
+    GLOBAL_MEM_CACHE_SIZE = auto()
+    GLOBAL_MEM_SIZE = auto()
+    MAX_CONSTANT_BUFFER_SIZE = auto()
+    MAX_CONSTANT_ARGS = auto()
+    LOCAL_MEM_TYPE = auto()
+    LOCAL_MEM_SIZE = auto()
+    ERROR_CORRECTION_SUPPORT = auto()
+    PROFILING_TIMER_RESOLUTION = auto()
+    ENDIAN_LITTLE = auto()
+    AVAILABLE = auto()
+    COMPILER_AVAILABLE = auto()
+    EXECUTION_CAPABILITIES = auto()
+    QUEUE_PROPERTIES = auto()
+    QUEUE_ON_HOST_PROPERTIES = auto()
+    NAME = auto()
+    VENDOR = auto()
+    DRIVER_VERSION = auto()
+    VERSION = auto()
+    PROFILE = auto()
+    EXTENSIONS = auto()
+    PLATFORM = auto()
+    PREFERRED_VECTOR_WIDTH_HALF = auto()
+    HOST_UNIFIED_MEMORY = auto()
+    NATIVE_VECTOR_WIDTH_CHAR = auto()
+    NATIVE_VECTOR_WIDTH_SHORT = auto()
+    NATIVE_VECTOR_WIDTH_INT = auto()
+    NATIVE_VECTOR_WIDTH_LONG = auto()
+    NATIVE_VECTOR_WIDTH_FLOAT = auto()
+    NATIVE_VECTOR_WIDTH_DOUBLE = auto()
+    NATIVE_VECTOR_WIDTH_HALF = auto()
+    OPENCL_C_VERSION = auto()
+    COMPUTE_CAPABILITY_MAJOR_NV = auto()
+    COMPUTE_CAPABILITY_MINOR_NV = auto()
+    REGISTERS_PER_BLOCK_NV = auto()
+    WARP_SIZE_NV = auto()
+    GPU_OVERLAP_NV = auto()
+    KERNEL_EXEC_TIMEOUT_NV = auto()
+    INTEGRATED_MEMORY_NV = auto()
+    ATTRIBUTE_ASYNC_ENGINE_COUNT_NV = auto()
+    PCI_BUS_ID_NV = auto()
+    PCI_SLOT_ID_NV = auto()
+    PCI_DOMAIN_ID_NV = auto()
+    PROFILING_TIMER_OFFSET_AMD = auto()
+    TOPOLOGY_AMD = auto()
+    BOARD_NAME_AMD = auto()
+    GLOBAL_FREE_MEMORY_AMD = auto()
+    SIMD_PER_COMPUTE_UNIT_AMD = auto()
+    SIMD_WIDTH_AMD = auto()
+    SIMD_INSTRUCTION_WIDTH_AMD = auto()
+    WAVEFRONT_WIDTH_AMD = auto()
+    GLOBAL_MEM_CHANNELS_AMD = auto()
+    GLOBAL_MEM_CHANNEL_BANKS_AMD = auto()
+    GLOBAL_MEM_CHANNEL_BANK_WIDTH_AMD = auto()
+    LOCAL_MEM_SIZE_PER_COMPUTE_UNIT_AMD = auto()
+    LOCAL_MEM_BANKS_AMD = auto()
+    THREAD_TRACE_SUPPORTED_AMD = auto()
+    GFXIP_MAJOR_AMD = auto()
+    GFXIP_MINOR_AMD = auto()
+    AVAILABLE_ASYNC_QUEUES_AMD = auto()
+    PREFERRED_WORK_GROUP_SIZE_AMD = auto()
+    MAX_WORK_GROUP_SIZE_AMD = auto()
+    PREFERRED_CONSTANT_BUFFER_SIZE_AMD = auto()
+    PCIE_ID_AMD = auto()
+    MAX_ATOMIC_COUNTERS_EXT = auto()
+    LINKER_AVAILABLE = auto()
+    BUILT_IN_KERNELS = auto()
+    IMAGE_MAX_BUFFER_SIZE = auto()
+    IMAGE_MAX_ARRAY_SIZE = auto()
+    PARENT_DEVICE = auto()
+    PARTITION_MAX_SUB_DEVICES = auto()
+    PARTITION_PROPERTIES = auto()
+    PARTITION_AFFINITY_DOMAIN = auto()
+    PARTITION_TYPE = auto()
+    REFERENCE_COUNT = auto()
+    PREFERRED_INTEROP_USER_SYNC = auto()
+    PRINTF_BUFFER_SIZE = auto()
+    IMAGE_PITCH_ALIGNMENT = auto()
+    IMAGE_BASE_ADDRESS_ALIGNMENT = auto()
+    MAX_READ_WRITE_IMAGE_ARGS = auto()
+    MAX_GLOBAL_VARIABLE_SIZE = auto()
+    QUEUE_ON_DEVICE_PROPERTIES = auto()
+    QUEUE_ON_DEVICE_PREFERRED_SIZE = auto()
+    QUEUE_ON_DEVICE_MAX_SIZE = auto()
+    MAX_ON_DEVICE_QUEUES = auto()
+    MAX_ON_DEVICE_EVENTS = auto()
+    SVM_CAPABILITIES = auto()
+    GLOBAL_VARIABLE_PREFERRED_TOTAL_SIZE = auto()
+    MAX_PIPE_ARGS = auto()
+    PIPE_MAX_ACTIVE_RESERVATIONS = auto()
+    PIPE_MAX_PACKET_SIZE = auto()
+    PREFERRED_PLATFORM_ATOMIC_ALIGNMENT = auto()
+    PREFERRED_GLOBAL_ATOMIC_ALIGNMENT = auto()
+    PREFERRED_LOCAL_ATOMIC_ALIGNMENT = auto()
+    IL_VERSION = auto()
+    MAX_NUM_SUB_GROUPS = auto()
+    SUB_GROUP_INDEPENDENT_FORWARD_PROGRESS = auto()
+    NUMERIC_VERSION = auto()
+    EXTENSIONS_WITH_VERSION = auto()
+    ILS_WITH_VERSION = auto()
+    BUILT_IN_KERNELS_WITH_VERSION = auto()
+    ATOMIC_MEMORY_CAPABILITIES = auto()
+    ATOMIC_FENCE_CAPABILITIES = auto()
+    NON_UNIFORM_WORK_GROUP_SUPPORT = auto()
+    OPENCL_C_ALL_VERSIONS = auto()
+    PREFERRED_WORK_GROUP_SIZE_MULTIPLE = auto()
+    WORK_GROUP_COLLECTIVE_FUNCTIONS_SUPPORT = auto()
+    GENERIC_ADDRESS_SPACE_SUPPORT = auto()
+    OPENCL_C_FEATURES = auto()
+    DEVICE_ENQUEUE_CAPABILITIES = auto()
+    PIPE_SUPPORT = auto()
+    ME_VERSION_INTEL = auto()
+    EXT_MEM_PADDING_IN_BYTES_QCOM = auto()
+    PAGE_SIZE_QCOM = auto()
+    SPIR_VERSIONS = auto()
+    SIMULTANEOUS_INTEROPS_INTEL = auto()
+    NUM_SIMULTANEOUS_INTEROPS_INTEL = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class device_topology_type_amd(IntEnum):  # noqa: N801
+    PCIE = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class device_fp_config(IntEnum):  # noqa: N801
+    DENORM = auto()
+    INF_NAN = auto()
+    ROUND_TO_NEAREST = auto()
+    ROUND_TO_ZERO = auto()
+    ROUND_TO_INF = auto()
+    FMA = auto()
+    SOFT_FLOAT = auto()
+    CORRECTLY_ROUNDED_DIVIDE_SQRT = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class device_mem_cache_type(IntEnum):  # noqa: N801
+    NONE = auto()
+    READ_ONLY_CACHE = auto()
+    READ_WRITE_CACHE = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class device_local_mem_type(IntEnum):  # noqa: N801
+    LOCAL = auto()
+    GLOBAL = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class device_exec_capabilities(IntEnum):  # noqa: N801
+    KERNEL = auto()
+    NATIVE_KERNEL = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class device_svm_capabilities(IntEnum):  # noqa: N801
+    COARSE_GRAIN_BUFFER = auto()
+    FINE_GRAIN_BUFFER = auto()
+    FINE_GRAIN_SYSTEM = auto()
+    ATOMICS = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class command_queue_properties(IntEnum):  # noqa: N801
+    _zero = 0
+    OUT_OF_ORDER_EXEC_MODE_ENABLE = auto()
+    PROFILING_ENABLE = auto()
+    ON_DEVICE = auto()
+    ON_DEVICE_DEFAULT = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class context_info(IntEnum):  # noqa: N801
+    REFERENCE_COUNT = auto()
+    DEVICES = auto()
+    PROPERTIES = auto()
+    NUM_DEVICES = auto()
+    INTEROP_USER_SYNC = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class gl_context_info(IntEnum):  # noqa: N801
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class context_properties(IntEnum):  # noqa: N801
+    PLATFORM = auto()
+    OFFLINE_DEVICES_AMD = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class command_queue_info(IntEnum):  # noqa: N801
+    CONTEXT = auto()
+    DEVICE = auto()
+    REFERENCE_COUNT = auto()
+    PROPERTIES = auto()
+    PROPERTIES_ARRAY = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class queue_properties(IntEnum):  # noqa: N801
+    PROPERTIES = auto()
+    SIZE = auto()
+    DEVICE_DEFAULT = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class mem_flags(IntEnum):  # noqa: N801
+    READ_WRITE = auto()
+    WRITE_ONLY = auto()
+    READ_ONLY = auto()
+    USE_HOST_PTR = auto()
+    ALLOC_HOST_PTR = auto()
+    COPY_HOST_PTR = auto()
+    USE_PERSISTENT_MEM_AMD = auto()
+    HOST_WRITE_ONLY = auto()
+    HOST_READ_ONLY = auto()
+    HOST_NO_ACCESS = auto()
+    KERNEL_READ_AND_WRITE = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class svm_mem_flags(IntEnum):  # noqa: N801
+    READ_WRITE = auto()
+    WRITE_ONLY = auto()
+    READ_ONLY = auto()
+    SVM_FINE_GRAIN_BUFFER = auto()
+    SVM_ATOMICS = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class channel_order(IntEnum):  # noqa: N801
+    R = auto()
+    A = auto()
+    RG = auto()
+    RA = auto()
+    RGB = auto()
+    RGBA = auto()
+    BGRA = auto()
+    INTENSITY = auto()
+    LUMINANCE = auto()
+    Rx = auto()
+    RGx = auto()
+    RGBx = auto()
+    sRGB = auto()  # noqa: N815
+    sRGBx = auto()  # noqa: N815
+    sRGBA = auto()  # noqa: N815
+    sBGRA = auto()  # noqa: N815
+    ABGR = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class channel_type(IntEnum):  # noqa: N801
+    SNORM_INT8 = auto()
+    SNORM_INT16 = auto()
+    UNORM_INT8 = auto()
+    UNORM_INT16 = auto()
+    UNORM_SHORT_565 = auto()
+    UNORM_SHORT_555 = auto()
+    UNORM_INT_101010 = auto()
+    SIGNED_INT8 = auto()
+    SIGNED_INT16 = auto()
+    SIGNED_INT32 = auto()
+    UNSIGNED_INT8 = auto()
+    UNSIGNED_INT16 = auto()
+    UNSIGNED_INT32 = auto()
+    HALF_FLOAT = auto()
+    FLOAT = auto()
+    UNORM_INT_101010_2 = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class mem_object_type(IntEnum):  # noqa: N801
+    BUFFER = auto()
+    IMAGE2D = auto()
+    IMAGE3D = auto()
+    IMAGE2D_ARRAY = auto()
+    IMAGE1D = auto()
+    IMAGE1D_ARRAY = auto()
+    IMAGE1D_BUFFER = auto()
+    PIPE = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class mem_info(IntEnum):  # noqa: N801
+    TYPE = auto()
+    FLAGS = auto()
+    SIZE = auto()
+    HOST_PTR = auto()
+    MAP_COUNT = auto()
+    REFERENCE_COUNT = auto()
+    CONTEXT = auto()
+    ASSOCIATED_MEMOBJECT = auto()
+    OFFSET = auto()
+    USES_SVM_POINTER = auto()
+    PROPERTIES = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class image_info(IntEnum):  # noqa: N801
+    FORMAT = auto()
+    ELEMENT_SIZE = auto()
+    ROW_PITCH = auto()
+    SLICE_PITCH = auto()
+    WIDTH = auto()
+    HEIGHT = auto()
+    DEPTH = auto()
+    ARRAY_SIZE = auto()
+    BUFFER = auto()
+    NUM_MIP_LEVELS = auto()
+    NUM_SAMPLES = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class pipe_info(IntEnum):  # noqa: N801
+    PACKET_SIZE = auto()
+    MAX_PACKETS = auto()
+    PROPERTIES = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class pipe_properties(IntEnum):  # noqa: N801
+    PACKET_SIZE = auto()
+    MAX_PACKETS = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class addressing_mode(IntEnum):  # noqa: N801
+    NONE = auto()
+    CLAMP_TO_EDGE = auto()
+    CLAMP = auto()
+    REPEAT = auto()
+    MIRRORED_REPEAT = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class filter_mode(IntEnum):  # noqa: N801
+    NEAREST = auto()
+    LINEAR = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class sampler_info(IntEnum):  # noqa: N801
+    REFERENCE_COUNT = auto()
+    CONTEXT = auto()
+    NORMALIZED_COORDS = auto()
+    ADDRESSING_MODE = auto()
+    FILTER_MODE = auto()
+    MIP_FILTER_MODE = auto()
+    LOD_MIN = auto()
+    LOD_MAX = auto()
+    PROPERTIES = auto()
+    MIP_FILTER_MODE_KHR = auto()
+    LOD_MIN_KHR = auto()
+    LOD_MAX_KHR = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class sampler_properties(IntEnum):  # noqa: N801
+    NORMALIZED_COORDS = auto()
+    ADDRESSING_MODE = auto()
+    FILTER_MODE = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class map_flags(IntEnum):  # noqa: N801
+    READ = auto()
+    WRITE = auto()
+    WRITE_INVALIDATE_REGION = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class program_info(IntEnum):  # noqa: N801
+    REFERENCE_COUNT = auto()
+    CONTEXT = auto()
+    NUM_DEVICES = auto()
+    DEVICES = auto()
+    SOURCE = auto()
+    BINARY_SIZES = auto()
+    BINARIES = auto()
+    NUM_KERNELS = auto()
+    KERNEL_NAMES = auto()
+    IL = auto()
+    SCOPE_GLOBAL_CTORS_PRESENT = auto()
+    SCOPE_GLOBAL_DTORS_PRESENT = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class program_build_info(IntEnum):  # noqa: N801
+    STATUS = auto()
+    OPTIONS = auto()
+    LOG = auto()
+    BINARY_TYPE = auto()
+    GLOBAL_VARIABLE_TOTAL_SIZE = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class program_binary_type(IntEnum):  # noqa: N801
+    NONE = auto()
+    COMPILED_OBJECT = auto()
+    LIBRARY = auto()
+    EXECUTABLE = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class kernel_info(IntEnum):  # noqa: N801
+    FUNCTION_NAME = auto()
+    NUM_ARGS = auto()
+    REFERENCE_COUNT = auto()
+    CONTEXT = auto()
+    PROGRAM = auto()
+    ATTRIBUTES = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class kernel_arg_info(IntEnum):  # noqa: N801
+    ADDRESS_QUALIFIER = auto()
+    ACCESS_QUALIFIER = auto()
+    TYPE_NAME = auto()
+    TYPE_QUALIFIER = auto()
+    NAME = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class kernel_arg_address_qualifier(IntEnum):  # noqa: N801
+    GLOBAL = auto()
+    LOCAL = auto()
+    CONSTANT = auto()
+    PRIVATE = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class kernel_arg_access_qualifier(IntEnum):  # noqa: N801
+    READ_ONLY = auto()
+    WRITE_ONLY = auto()
+    READ_WRITE = auto()
+    NONE = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class kernel_arg_type_qualifier(IntEnum):  # noqa: N801
+    NONE = auto()
+    CONST = auto()
+    RESTRICT = auto()
+    VOLATILE = auto()
+    PIPE = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class kernel_work_group_info(IntEnum):  # noqa: N801
+    WORK_GROUP_SIZE = auto()
+    COMPILE_WORK_GROUP_SIZE = auto()
+    LOCAL_MEM_SIZE = auto()
+    PREFERRED_WORK_GROUP_SIZE_MULTIPLE = auto()
+    PRIVATE_MEM_SIZE = auto()
+    GLOBAL_WORK_SIZE = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class kernel_sub_group_info(IntEnum):  # noqa: N801
+    MAX_SUB_GROUP_SIZE_FOR_NDRANGE = auto()
+    SUB_GROUP_COUNT_FOR_NDRANGE = auto()
+    LOCAL_SIZE_FOR_SUB_GROUP_COUNT = auto()
+    MAX_NUM_SUB_GROUPS = auto()
+    COMPILE_NUM_SUB_GROUPS = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class event_info(IntEnum):  # noqa: N801
+    COMMAND_QUEUE = auto()
+    COMMAND_TYPE = auto()
+    REFERENCE_COUNT = auto()
+    COMMAND_EXECUTION_STATUS = auto()
+    CONTEXT = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class command_type(IntEnum):  # noqa: N801
+    NDRANGE_KERNEL = auto()
+    TASK = auto()
+    NATIVE_KERNEL = auto()
+    READ_BUFFER = auto()
+    WRITE_BUFFER = auto()
+    COPY_BUFFER = auto()
+    READ_IMAGE = auto()
+    WRITE_IMAGE = auto()
+    COPY_IMAGE = auto()
+    COPY_IMAGE_TO_BUFFER = auto()
+    COPY_BUFFER_TO_IMAGE = auto()
+    MAP_BUFFER = auto()
+    MAP_IMAGE = auto()
+    UNMAP_MEM_OBJECT = auto()
+    MARKER = auto()
+    ACQUIRE_GL_OBJECTS = auto()
+    RELEASE_GL_OBJECTS = auto()
+    READ_BUFFER_RECT = auto()
+    WRITE_BUFFER_RECT = auto()
+    COPY_BUFFER_RECT = auto()
+    USER = auto()
+    BARRIER = auto()
+    MIGRATE_MEM_OBJECTS = auto()
+    FILL_BUFFER = auto()
+    FILL_IMAGE = auto()
+    SVM_FREE = auto()
+    SVM_MEMCPY = auto()
+    SVM_MEMFILL = auto()
+    SVM_MAP = auto()
+    SVM_UNMAP = auto()
+    SVM_MIGRATE_MEM = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class command_execution_status(IntEnum):  # noqa: N801
+    COMPLETE = auto()
+    RUNNING = auto()
+    SUBMITTED = auto()
+    QUEUED = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class profiling_info(IntEnum):  # noqa: N801
+    QUEUED = auto()
+    SUBMIT = auto()
+    START = auto()
+    END = auto()
+    COMPLETE = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class mem_migration_flags(IntEnum):  # noqa: N801
+    HOST = auto()
+    CONTENT_UNDEFINED = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class device_partition_property(IntEnum):  # noqa: N801
+    EQUALLY = auto()
+    BY_COUNTS = auto()
+    BY_COUNTS_LIST_END = auto()
+    BY_AFFINITY_DOMAIN = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class device_affinity_domain(IntEnum):  # noqa: N801
+    NUMA = auto()
+    L4_CACHE = auto()
+    L3_CACHE = auto()
+    L2_CACHE = auto()
+    L1_CACHE = auto()
+    NEXT_PARTITIONABLE = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class device_atomic_capabilities(IntEnum):  # noqa: N801
+    ORDER_RELAXED = auto()
+    ORDER_ACQ_REL = auto()
+    ORDER_SEQ_CST = auto()
+    SCOPE_WORK_ITEM = auto()
+    SCOPE_WORK_GROUP = auto()
+    SCOPE_DEVICE = auto()
+    SCOPE_ALL_DEVICES = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class device_device_enqueue_capabilities(IntEnum):  # noqa: N801
+    SUPPORTED = auto()
+    REPLACEABLE_DEFAULT = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class version_bits(IntEnum):  # noqa: N801
+    MAJOR_BITS = auto()
+    MINOR_BITS = auto()
+    PATCH_BITS = auto()
+    MAJOR_MASK = auto()
+    MINOR_MASK = auto()
+    PATCH_MASK = auto()
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class khronos_vendor_id(IntEnum):  # noqa: N801
+    CODEPLAY = auto()
+
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+class gl_object_type(IntEnum):  # noqa: N801
+    BUFFER = auto()
+    TEXTURE2D = auto()
+    TEXTURE3D = auto()
+    RENDERBUFFER = auto()
+
+    to_string = pyopencl._monkeypatch.to_string
+
+class gl_texture_info(IntEnum):  # noqa: N801
+    TEXTURE_TARGET = auto()
+    MIPMAP_LEVEL = auto()
+
+    to_string = pyopencl._monkeypatch.to_string
+
+class NameVersion:
+    def __init__(self, version: int = 0, name: str = "") -> None: ...
+
+    @property
+    def version(self) -> int: ...
+
+    @version.setter
+    def version(self, arg: int, /) -> None: ...
+
+    @property
+    def name(self) -> str: ...
+
+    @name.setter
+    def name(self, arg: str, /) -> None: ...
+
+class DeviceTopologyAmd:
+    def __init__(self, bus: int = 0, device: int = 0, function: int = 0) -> None: ...
+
+    @property
+    def type(self) -> int: ...
+
+    @type.setter
+    def type(self, arg: int, /) -> None: ...
+
+    @property
+    def bus(self) -> int: ...
+
+    @bus.setter
+    def bus(self, arg: int, /) -> None: ...
+
+    @property
+    def device(self) -> int: ...
+
+    @device.setter
+    def device(self, arg: int, /) -> None: ...
+
+    @property
+    def function(self) -> int: ...
+
+    @function.setter
+    def function(self, arg: int, /) -> None: ...
+
+def get_cl_header_version() -> tuple[int, int]: ...
+
+def get_platforms() -> list[Platform]: ...
+
+class Platform:
+    def get_info(self, arg: platform_info, /) -> object: ...
+
+    def get_devices(self, device_type: int = 4294967295) -> list[Device]: ...
+
+    @override
+    def __hash__(self) -> int: ...
+
+    @overload
+    def __eq__(self, arg: Platform, /) -> bool: ...
+
+    @overload
+    def __eq__(self, obj: object | None) -> bool: ...
+
+    @staticmethod
+    def from_int_ptr(int_ptr_value: int, retain: bool = True) -> Platform: ...
+
+    @property
+    def int_ptr(self) -> int: ...
+
+    __repr__ = pyopencl._monkeypatch.platform_repr
+
+    _get_cl_version = pyopencl._monkeypatch.generic_get_cl_version
+
+    profile: str
+    version: str
+    name: str
+    vendor: str
+    extensions: str
+    host_timer_resolution: int
+    numeric_version: int
+    extensions_with_version: Sequence[NameVersion]
+
+class Device:
+    def get_info(self, arg: device_info, /) -> object: ...
+
+    @overload
+    def __eq__(self, arg: Device, /) -> bool: ...
+
+    @overload
+    def __eq__(self, obj: object | None) -> bool: ...
+
+    @override
+    def __hash__(self) -> int: ...
+
+    def create_sub_devices(self,
+            arg: Sequence[device_partition_property],
+        /) -> list[Device]: ...
+
+    @staticmethod
+    def from_int_ptr(int_ptr_value: int, retain: bool = True) -> Device: ...
+
+    @property
+    def int_ptr(self) -> int: ...
+
+    def device_and_host_timer(self) -> tuple[int, int]: ...
+
+    def host_timer(self) -> int: ...
+
+    _get_cl_version = pyopencl._monkeypatch.generic_get_cl_version
+
+    __repr__ = pyopencl._monkeypatch.device_repr
+
+    type: device_type
+    vendor_id: int
+    max_compute_units: int
+    max_work_item_dimensions: int
+    max_work_group_size: int
+    max_work_item_sizes: Sequence[int]
+    preferred_vector_width_char: int
+    preferred_vector_width_short: int
+    preferred_vector_width_int: int
+    preferred_vector_width_long: int
+    preferred_vector_width_float: int
+    preferred_vector_width_double: int
+    max_clock_frequency: int
+    address_bits: int
+    max_read_image_args: int
+    max_write_image_args: int
+    max_mem_alloc_size: int
+    image2d_max_width: int
+    image2d_max_height: int
+    image3d_max_width: int
+    image3d_max_height: int
+    image3d_max_depth: int
+    image_support: bool
+    max_parameter_size: int
+    max_samplers: int
+    mem_base_addr_align: int
+    min_data_type_align_size: int
+    single_fp_config: int
+    double_fp_config: int
+    half_fp_config: int
+    global_mem_cache_type: device_mem_cache_type
+    global_mem_cacheline_size: int
+    global_mem_cache_size: int
+    global_mem_size: int
+    max_constant_buffer_size: int
+    max_constant_args: int
+    local_mem_type: device_local_mem_type
+    local_mem_size: int
+    error_correction_support: bool
+    profiling_timer_resolution: int
+    endian_little: bool
+    available: bool
+    compiler_available: bool
+    execution_capabilities: int
+    queue_properties: int
+    queue_on_host_properties: int
+    name: str
+    vendor: str
+    driver_version: str
+    version: str
+    profile: str
+    extensions: str
+    platform: Platform
+    preferred_vector_width_half: int
+    host_unified_memory: bool
+    native_vector_width_char: int
+    native_vector_width_short: int
+    native_vector_width_int: int
+    native_vector_width_long: int
+    native_vector_width_float: int
+    native_vector_width_double: int
+    native_vector_width_half: int
+    opencl_c_version: str
+    compute_capability_major_nv: int
+    compute_capability_minor_nv: int
+    registers_per_block_nv: int
+    warp_size_nv: int
+    gpu_overlap_nv: bool
+    kernel_exec_timeout_nv: bool
+    integrated_memory_nv: bool
+    attribute_async_engine_count_nv: int
+    pci_bus_id_nv: int
+    pci_slot_id_nv: int
+    pci_domain_id_nv: int
+    profiling_timer_offset_amd: int
+    topology_amd: DeviceTopologyAmd
+    board_name_amd: str
+    global_free_memory_amd: int
+    simd_per_compute_unit_amd: int
+    simd_width_amd: int
+    simd_instruction_width_amd: int
+    wavefront_width_amd: int
+    global_mem_channels_amd: int
+    global_mem_channel_banks_amd: int
+    global_mem_channel_bank_width_amd: int
+    local_mem_size_per_compute_unit_amd: int
+    local_mem_banks_amd: int
+    thread_trace_supported_amd: int
+    gfxip_major_amd: int
+    gfxip_minor_amd: int
+    available_async_queues_amd: int
+    # preferred_work_group_size_amd
+    # max_work_group_size_amd
+    # preferred_constant_buffer_size_amd
+    # pcie_id_amd
+    max_atomic_counters_ext: int
+    linker_available: bool
+    built_in_kernels: str
+    image_max_buffer_size: int
+    image_max_array_size: int
+    parent_device: Device
+    partition_max_sub_devices: int
+    partition_properties: int
+    partition_affinity_domain: device_affinity_domain
+    partition_type: device_partition_property
+    reference_count: int
+    preferred_interop_user_sync: bool
+    printf_buffer_size: int
+    # image_pitch_alignment
+    # image_base_address_alignment
+    max_read_write_image_args: int
+    max_global_variable_size: int
+    queue_on_device_properties: command_queue_properties
+    queue_on_device_preferred_size: int
+    queue_on_device_max_size: int
+    max_on_device_queues: int
+    max_on_device_events: int
+    svm_capabilities: device_svm_capabilities
+    global_variable_preferred_total_size: int
+    max_pipe_args: int
+    pipe_max_active_reservations: int
+    pipe_max_packet_size: int
+    preferred_platform_atomic_alignment: int
+    preferred_global_atomic_alignment: int
+    preferred_local_atomic_alignment: int
+    il_version: int
+    max_num_sub_groups: int
+    sub_group_independent_forward_progress: bool
+    numeric_version: int
+    extensions_with_version: Sequence[NameVersion]
+    ils_with_version: Sequence[NameVersion]
+    built_in_kernels_with_version: Sequence[NameVersion]
+    atomic_memory_capabilities: device_atomic_capabilities
+    atomic_fence_capabilities: device_atomic_capabilities
+    non_uniform_work_group_support: bool
+    opencl_c_all_versions: Sequence[NameVersion]
+    preferred_work_group_size_multiple: int
+    work_group_collective_functions_support: bool
+    generic_address_space_support: bool
+    opencl_c_features: Sequence[NameVersion]
+    device_enqueue_capabilities: device_device_enqueue_capabilities
+    pipe_support: bool
+    me_version_intel: int
+    ext_mem_padding_in_bytes_qcom: int
+    page_size_qcom: int
+    spir_versions: str
+    simultaneous_interops_intel: Sequence[Device]
+    num_simultaneous_interops_intel: int
+
+class Context:
+    def __init__(self,
+            devices: Sequence[Device] | None = None,
+            properties: Sequence[context_properties] | None = None,
+            dev_type: device_type | None = None
+        ) -> None: ...
+
+    def get_info(self, arg: context_info, /) -> object: ...
+
+    @overload
+    def __eq__(self, arg: Context, /) -> bool: ...
+
+    @overload
+    def __eq__(self, obj: object | None) -> bool: ...
+
+    @override
+    def __hash__(self) -> int: ...
+
+    @staticmethod
+    def from_int_ptr(int_ptr_value: int, retain: bool = True) -> Context: ...
+
+    @property
+    def int_ptr(self) -> int: ...
+
+    def set_default_device_command_queue(self,
+            device: Device,
+            queue: CommandQueue,
+        /) -> None: ...
+
+    _get_cl_version = pyopencl._monkeypatch.context_get_cl_version
+
+    __repr__ = pyopencl._monkeypatch.context_repr
+
+    reference_count: int
+    devices: Sequence[Device]
+    properties: Sequence[int]
+    num_devices: int
+    # interop_user_sync:
+
+class CommandQueue:
+    def __init__(self,
+             context: Context,
+             device: Device | None = None,
+             properties: command_queue_properties = command_queue_properties._zero
+         ) -> None: ...
+
+    def _finalize(self) -> None: ...
+
+    def get_info(self, arg: command_queue_info, /) -> object: ...
+
+    def flush(self) -> None: ...
+
+    def finish(self) -> None: ...
+
+    @overload
+    def __eq__(self, arg: CommandQueue, /) -> bool: ...
+
+    @overload
+    def __eq__(self, obj: object | None) -> bool: ...
+
+    @override
+    def __hash__(self) -> int: ...
+
+    @staticmethod
+    def from_int_ptr(int_ptr_value: int, retain: bool = True) -> CommandQueue: ...
+
+    @property
+    def int_ptr(self) -> int: ...
+
+    _get_cl_version = pyopencl._monkeypatch.command_queue_get_cl_version
+
+    __enter__ = pyopencl._monkeypatch.command_queue_enter
+
+    __exit__ = pyopencl._monkeypatch.command_queue_exit
+
+    context: Context
+    device: Device
+
+    reference_count: int
+    properties: command_queue_properties
+    properties_array: Sequence[command_queue_properties]
+
+class Event:
+    def get_info(self, arg: event_info, /) -> object: ...
+
+    def get_profiling_info(self, arg: int, /) -> object: ...
+
+    def wait(self) -> None: ...
+
+    @overload
+    def __eq__(self, arg: Event, /) -> bool: ...
+
+    @overload
+    def __eq__(self, obj: object | None) -> bool: ...
+
+    @override
+    def __hash__(self) -> int: ...
+
+    @staticmethod
+    def from_int_ptr(int_ptr_value: int, retain: bool = True) -> Event: ...
+
+    @property
+    def int_ptr(self) -> int: ...
+
+    def set_callback(self,
+             arg0: int,
+             arg1: Callable[[command_execution_status], None],
+         /) -> None: ...
+
+    command_queue: CommandQueue
+    command_type: command_type
+    reference_count: int
+    command_execution_status: command_execution_status
+    profile: pyopencl._monkeypatch.ProfilingInfoGetter
+
+class NannyEvent(Event):
+    def get_ward(self) -> object: ...
+
+def wait_for_events(arg: Sequence[Event], /) -> None: ...
+
+def _enqueue_marker_with_wait_list(
+    queue: CommandQueue,
+    wait_for: WaitList = None) -> Event: ...
+
+def _enqueue_marker(queue: CommandQueue) -> Event: ...
+
+def _enqueue_wait_for_events(
+    queue: CommandQueue,
+    wait_for: WaitList = None) -> None: ...
+
+def _enqueue_barrier_with_wait_list(
+    queue: CommandQueue,
+    wait_for: WaitList = None) -> Event: ...
+
+def _enqueue_barrier(queue: CommandQueue) -> None: ...
+
+class UserEvent(Event):
+    def __init__(self, context: Context) -> None: ...
+
+    def set_status(self, arg: int, /) -> None: ...
+
+class MemoryObjectHolder:
+    def get_info(self, arg: mem_info, /) -> object: ...
+
+    def get_host_array(self,
+            shape: tuple[int, ...],
+            dtype: DTypeT,
+            order: Literal["C"] | Literal["F"] = "C"
+        ) -> np.ndarray[tuple[int, ...], DTypeT]: ...
+
+    @overload
+    def __eq__(self, arg: MemoryObjectHolder, /) -> bool: ...
+
+    @overload
+    def __eq__(self, obj: object | None) -> bool: ...
+
+    @override
+    def __hash__(self) -> int: ...
+
+    @property
+    def int_ptr(self) -> int: ...
+
+    type: int
+    flags: int
+    size: int
+    # FIXME
+    # host_ptr
+    map_count: int
+    reference_count: int
+    context: Context
+    # associated_memobject
+    offset: int
+    uses_svm_pointer: bool
+    properties: int
+
+class MemoryObject(MemoryObjectHolder):
+    def release(self) -> None: ...
+
+    @property
+    def hostbuf(self) -> object: ...
+
+    @staticmethod
+    def from_int_ptr(int_ptr_value: int, retain: bool = True) -> object: ...
+
+def enqueue_migrate_mem_objects(
+        queue: CommandQueue,
+        mem_objects: Sequence[MemoryObjectHolder],
+        flags: int = 0,
+        wait_for: WaitList = None
+    ) -> Event: ...
+
+class Buffer(MemoryObject):
+    def __init__(
+             self,
+             context: Context,
+             flags: int,
+             size: int = 0,
+             hostbuf: HasBufferInterface | None = None
+         ) -> None: ...
+
+    def get_sub_region(self, origin: int, size: int, flags: int = 0) -> Buffer: ...
+
+    def __getitem__(self, arg: slice, /) -> Buffer: ...
+
+def _enqueue_read_buffer(
+        queue: CommandQueue,
+        mem: MemoryObjectHolder,
+        hostbuf: HasBufferInterface,
+        src_offset: int = 0,
+        wait_for: WaitList = None,
+        is_blocking: bool = True
+    ) -> Event: ...
+
+def _enqueue_write_buffer(
+        queue: CommandQueue,
+        mem: MemoryObjectHolder,
+        hostbuf: HasBufferInterface,
+        dst_offset: int = 0,
+        wait_for: WaitList = None,
+        is_blocking: bool = True
+    ) -> Event: ...
+
+def _enqueue_copy_buffer(
+        queue: CommandQueue,
+        src: MemoryObjectHolder,
+        dst: MemoryObjectHolder,
+        byte_count: int = -1,
+        src_offset: int = 0,
+        dst_offset: int = 0,
+        wait_for: WaitList = None
+    ) -> Event: ...
+
+def _enqueue_read_buffer_rect(
+        queue: CommandQueue,
+        mem: MemoryObjectHolder,
+        hostbuf: HasBufferInterface,
+        buffer_origin: tuple[int, ...],
+        host_origin: tuple[int, ...],
+        region: tuple[int, ...],
+        buffer_pitches: tuple[int, ...] | None = None,
+        host_pitches: tuple[int, ...] | None = None,
+        wait_for: WaitList = None,
+        is_blocking: bool = True
+    ) -> Event: ...
+
+def _enqueue_write_buffer_rect(
+        queue: CommandQueue,
+        mem: MemoryObjectHolder,
+        hostbuf: HasBufferInterface,
+        buffer_origin: tuple[int, ...],
+        host_origin: tuple[int, ...],
+        region: object,
+        buffer_pitches: tuple[int, ...] | None = None,
+        host_pitches: tuple[int, ...] | None = None,
+        wait_for: WaitList = None,
+        is_blocking: bool = True
+    ) -> Event: ...
+
+def _enqueue_copy_buffer_rect(
+        queue: CommandQueue,
+        src: MemoryObjectHolder,
+        dst: MemoryObjectHolder,
+        src_origin: tuple[int, ...],
+        dst_origin: tuple[int, ...],
+        region: object,
+        src_pitches: tuple[int, ...] | None = None,
+        dst_pitches: tuple[int, ...] | None = None,
+        wait_for: WaitList = None
+    ) -> Event: ...
+
+def _enqueue_fill_buffer(
+        queue: CommandQueue,
+        mem: MemoryObjectHolder,
+        pattern: object,
+        offset: int,
+        size: int,
+        wait_for: WaitList = None
+    ) -> Event: ...
+
+def enqueue_copy_buffer_p2p_amd(
+        platform: Platform,
+        queue: CommandQueue,
+        src: MemoryObjectHolder,
+        dst: MemoryObjectHolder,
+        byte_count: int | None = None,
+        wait_for: WaitList = None
+    ) -> Event: ...
+
+class ImageDescriptor:
+    def __init__(self) -> None: ...
+
+    @property
+    def image_type(self) -> mem_object_type: ...
+
+    @image_type.setter
+    def image_type(self, arg: mem_object_type, /) -> None: ...
+
+    @property
+    def shape(self) -> tuple[int, int, int]: ...
+
+    @shape.setter
+    def shape(self, arg: tuple[int, int, int], /) -> None: ...
+
+    @property
+    def array_size(self) -> int: ...
+
+    @array_size.setter
+    def array_size(self, arg: int, /) -> None: ...
+
+    @property
+    def pitches(self) -> object: ...
+
+    @pitches.setter
+    def pitches(self, arg: tuple[int, int], /) -> None: ...
+
+    @property
+    def num_mip_levels(self) -> int: ...
+
+    @num_mip_levels.setter
+    def num_mip_levels(self, arg: int, /) -> None: ...
+
+    @property
+    def num_samples(self) -> int: ...
+
+    @num_samples.setter
+    def num_samples(self, arg: int, /) -> None: ...
+
+    @property
+    def buffer(self) -> object: ...
+
+    @buffer.setter
+    def buffer(self, buffer: MemoryObject | None) -> None: ...
+
+class Image(MemoryObject):
+    # monkeypatched, but apparently pyright doesn't like monkeypatched __init__
+    def __init__(self,
+                context: Context,
+                flags: mem_flags,
+                format: ImageFormat,
+                shape: tuple[int, ...] | None = None,
+                pitches: tuple[int, ...] | None = None,
+
+                hostbuf: HasBufferInterface | None = None,
+                is_array: bool = False,
+                buffer: Buffer | None = None,
+                *,
+                desc: ImageDescriptor | None = None,
+                _through_create_image: bool = False,
+            ) -> None: ...
+
+    @overload
+    @staticmethod
+    def _custom_init(
+            h: Image,
+            context: Context,
+            flags: int,
+            format: ImageFormat,
+            shape: tuple[int, ...] | None = None,
+            pitches: tuple[int, ...] | None = None,
+            hostbuf: object | None = None
+        ) -> None: ...
+
+    @overload
+    @staticmethod
+    def _custom_init(
+            h: Image,
+            context: Context,
+            flags: int,
+            format: ImageFormat,
+            desc: ImageDescriptor,
+            hostbuf: object | None = None
+        ) -> None: ...
+
+    def get_image_info(self, arg: image_info, /) -> object: ...
+
+    format: ImageFormat
+    element_size: int
+    row_pitch: int
+    slice_pitch: int
+    width: int
+    height: int
+    depth: int
+    array_size: int
+    buffer: Buffer
+    num_mip_levels: int
+    num_samples: int
+
+class ImageFormat:
+    def __init__(self, arg0: int, arg1: int, /) -> None: ...
+
+    @property
+    def channel_order(self) -> int: ...
+
+    @channel_order.setter
+    def channel_order(self, arg: int, /) -> None: ...
+
+    @property
+    def channel_data_type(self) -> int: ...
+
+    @channel_data_type.setter
+    def channel_data_type(self, arg: int, /) -> None: ...
+
+    @property
+    def channel_count(self) -> int: ...
+
+    @property
+    def dtype_size(self) -> int: ...
+
+    @property
+    def itemsize(self) -> int: ...
+
+    __repr__ = pyopencl._monkeypatch.image_format_repr
+
+    __eq__ = pyopencl._monkeypatch.image_format_eq
+
+    __ne__ = pyopencl._monkeypatch.image_format_ne
+
+    __hash__ = pyopencl._monkeypatch.image_format_hash
+
+def get_supported_image_formats(
+        context: Context,
+        arg1: mem_flags,
+        arg2: mem_object_type,
+    /) -> Sequence[ImageFormat]: ...
+
+def _enqueue_read_image(
+        queue: CommandQueue,
+        mem: Image,
+        origin: tuple[int, ...],
+        region: tuple[int, ...],
+        hostbuf: HasBufferInterface,
+        row_pitch: int = 0,
+        slice_pitch: int = 0,
+        wait_for: WaitList = None,
+        is_blocking: bool = True
+    ) -> Event: ...
+
+def _enqueue_write_image(
+        queue: CommandQueue,
+        mem: Image,
+        origin: tuple[int, ...],
+        region: tuple[int, ...],
+        hostbuf: HasBufferInterface,
+        row_pitch: int = 0,
+        slice_pitch: int = 0,
+        wait_for: WaitList = None,
+        is_blocking: bool = True
+    ) -> Event: ...
+
+def _enqueue_copy_image(
+        queue: CommandQueue,
+        src: MemoryObjectHolder,
+        dest: MemoryObjectHolder,
+        src_origin: tuple[int, ...],
+        dest_origin: tuple[int, ...],
+        region: tuple[int, ...],
+        wait_for: WaitList = None
+    ) -> Event: ...
+
+def _enqueue_copy_image_to_buffer(
+        queue: CommandQueue,
+        src: MemoryObjectHolder,
+        dest: MemoryObjectHolder,
+        origin: tuple[int, ...],
+        region: tuple[int, ...],
+        offset: int,
+        wait_for: WaitList = None
+    ) -> Event: ...
+
+def _enqueue_copy_buffer_to_image(
+        queue: CommandQueue,
+        src: MemoryObjectHolder,
+        dest: MemoryObjectHolder,
+        offset: int,
+        origin: tuple[int, ...],
+        region: tuple[int, ...],
+        wait_for: WaitList = None
+    ) -> Event: ...
+
+def enqueue_fill_image(
+        queue: CommandQueue,
+        mem: MemoryObjectHolder,
+        color: HasBufferInterface,
+        origin: tuple[int, ...],
+        region: tuple[int, ...],
+        wait_for: WaitList = None
+    ) -> Event: ...
+
+class Pipe(MemoryObject):
+    def __init__(self,
+            context: Context,
+            flags: int,
+            packet_size: int,
+            max_packets: int,
+            properties: Sequence[pipe_properties] = ()
+        ) -> None: ...
+
+    def get_pipe_info(self, arg: pipe_info, /) -> object: ...
+
+    packet_size: int
+    max_packets: int
+    # FIXME
+    # properties:
+
+class MemoryMap:
+    def release(self,
+            queue: CommandQueue | None = None,
+            wait_for: WaitList = None
+        ) -> Event: ...
+
+    __enter__ = pyopencl._monkeypatch.memory_map_enter
+
+    __exit__ = pyopencl._monkeypatch.memory_map_exit
+
+def enqueue_map_buffer(
+        queue: CommandQueue,
+        buf: MemoryObjectHolder,
+        flags: map_flags,
+        offset: int,
+        shape: tuple[int, ...],
+        dtype: DTypeT,
+        order: Literal["C"] | Literal["F"] = "C",
+        strides: tuple[int, ...] | None = None,
+        wait_for: WaitList = None,
+        is_blocking: bool = True
+    ) -> np.ndarray[tuple[int, ...], DTypeT]: ...
+
+def enqueue_map_image(
+        queue: CommandQueue,
+        img: MemoryObjectHolder,
+        flags: int,
+        origin: tuple[int, ...],
+        region: tuple[int, ...],
+        shape: tuple[int, ...],
+        dtype: DTypeT,
+        order: Literal["C"] | Literal["F"] = "C",
+        strides: tuple[int, ...] | None = None,
+        wait_for: WaitList = None,
+        is_blocking: bool = True
+    ) -> np.ndarray[tuple[int, ...], DTypeT]: ...
+
+class SVMPointer:
+    @property
+    def svm_ptr(self) -> int: ...
+
+    @property
+    def size(self) -> int | None: ...
+
+    @property
+    def buf(self) -> NDArray[np.uint8]: ...
+
+    map = pyopencl._monkeypatch.svmptr_map
+
+    map_ro = pyopencl._monkeypatch.svmptr_map_ro
+
+    map_rw = pyopencl._monkeypatch.svmptr_map_rw
+
+    _enqueue_unmap = pyopencl._monkeypatch.svmptr__enqueue_unmap
+
+    as_buffer = pyopencl._monkeypatch.svmptr_as_buffer
+
+class SVM(SVMPointer, Generic[SVMInnerT]):
+    def __init__(self, arg: SVMInnerT, /) -> None: ...
+
+    @property
+    def mem(self) -> SVMInnerT: ...
+
+    map = pyopencl._monkeypatch.svm_map
+
+    map_ro = pyopencl._monkeypatch.svm_map_ro
+
+    map_rw = pyopencl._monkeypatch.svm_map_rw
+
+    _enqueue_unmap = pyopencl._monkeypatch.svm__enqueue_unmap
+
+class SVMAllocation(SVMPointer):
+    def __init__(self,
+             context: Context,
+             size: int,
+             alignment: int,
+             flags: int,
+             queue: CommandQueue | None = None
+         ) -> None: ...
+
+    def release(self) -> None: ...
+
+    def enqueue_release(self,
+            queue: CommandQueue | None = None,
+            wait_for: WaitList = None
+        ) -> Event: ...
+
+    @overload
+    def __eq__(self, arg: SVMAllocation, /) -> bool: ...
+
+    @overload
+    def __eq__(self, obj: object | None) -> bool: ...
+
+    @override
+    def __hash__(self) -> int: ...
+
+    def bind_to_queue(self, queue: CommandQueue) -> None: ...
+
+    def unbind_from_queue(self) -> None: ...
+
+    @property
+    def _queue(self) -> object: ...
+
+def _enqueue_svm_memcpy(
+        queue: CommandQueue,
+        is_blocking: int,
+        dst: SVMPointer,
+        src: SVMPointer,
+        wait_for: WaitList = None,
+        byte_count: object | None = None
+    ) -> Event: ...
+
+def _enqueue_svm_memfill(
+        queue: CommandQueue,
+        dst: SVMPointer,
+        pattern: object,
+        byte_count: object | None = None,
+        wait_for: WaitList = None
+    ) -> Event: ...
+
+def _enqueue_svm_map(
+        queue: CommandQueue,
+        is_blocking: int,
+        flags: int,
+        svm: SVMPointer,
+        wait_for: WaitList = None,
+        size: object | None = None
+    ) -> Event: ...
+
+def _enqueue_svm_unmap(
+        queue: CommandQueue,
+        svm: SVMPointer,
+        wait_for: WaitList = None
+    ) -> Event: ...
+
+def _enqueue_svm_migrate_mem(
+        queue: CommandQueue,
+        svms: Sequence[SVMPointer],
+        flags: int | None = None,
+        wait_for: WaitList = None
+    ) -> Event: ...
+
+class Sampler:
+    @overload
+    def __init__(self, arg0: Context, arg1: Sequence[int], /) -> None: ...
+
+    @overload
+    def __init__(self, arg0: Context, arg1: bool, arg2: int, arg3: int, /) -> None: ...
+
+    def get_info(self, arg: sampler_info, /) -> object: ...
+
+    @overload
+    def __eq__(self, arg: Sampler, /) -> bool: ...
+
+    @overload
+    def __eq__(self, obj: object | None) -> bool: ...
+
+    @override
+    def __hash__(self) -> int: ...
+
+    @staticmethod
+    def from_int_ptr(int_ptr_value: int, retain: bool = True) -> Sampler: ...
+
+    @property
+    def int_ptr(self) -> int: ...
+
+    reference_count: int
+    context: Context
+    normalized_coords: bool
+    addressing_mode: addressing_mode
+    filter_mode: filter_mode
+    mip_filter_mode: filter_mode
+    lod_min: float
+    lod_max: float
+    properties: Sequence[sampler_properties]
+    mip_filter_mode_khr: filter_mode
+    lod_min_khr: float
+    lod_max_khr: float
+
+class program_kind(IntEnum):  # noqa: N801
+    UNKNOWN = auto()
+    SOURCE = auto()
+    BINARY = auto()
+    IL = auto()
+
+    to_string = classmethod(pyopencl._monkeypatch.to_string)
+
+def unload_platform_compiler(arg: Platform, /) -> None: ...
+
+class Kernel:
+    def __init__(self, arg0: _Program | Program, arg1: str, /) -> None: ...
+
+    get_info = pyopencl._monkeypatch.kernel_get_info
+
+    get_work_group_info = pyopencl._monkeypatch.kernel_get_work_group_info
+
+    def clone(self) -> Kernel: ...
+
+    def set_arg(self, arg0: int, arg1: KernelArg, /) -> None: ...
+
+    def get_arg_info(self, index: int, param_name: kernel_arg_info, /) -> object: ...
+
+    @overload
+    def __eq__(self, arg: Kernel, /) -> bool: ...
+
+    @overload
+    def __eq__(self, obj: object | None) -> bool: ...
+
+    @override
+    def __hash__(self) -> int: ...
+
+    @staticmethod
+    def from_int_ptr(int_ptr_value: int, retain: bool = True) -> Kernel: ...
+
+    @property
+    def int_ptr(self) -> int: ...
+
+    def get_sub_group_info(self,
+            device: Device,
+            param: int,
+            input_value: Sequence[int] | int | None = None
+        ) -> object: ...
+
+    def __call__(self,
+            queue: CommandQueue,
+            global_work_size: tuple[int, ...],
+            local_work_size: tuple[int, ...] | None,
+            *args: KernelArg,
+            wait_for: WaitList = None,
+            g_times_l: bool = False,
+            allow_empty_ndrange: bool = False,
+            global_offset: tuple[int, ...] | None = None,
+         ) -> Event: ...
+
+    def set_args(self, *args: KernelArg) -> None: ...
+
+    set_scalar_arg_dtypes = pyopencl._monkeypatch.kernel_set_arg_types
+
+    set_arg_types = pyopencl._monkeypatch.kernel_set_arg_types
+
+    capture_call = pyopencl._monkeypatch.kernel_capture_call
+
+    function_name: str
+    num_args: int
+    reference_count: int
+    context: Context
+    program: _Program
+    attributes: str
+
+class LocalMemory:
+    def __init__(self, size: int) -> None: ...
+
+    @property
+    def size(self) -> int: ...
+
+def enqueue_nd_range_kernel(
+        queue: CommandQueue,
+        kernel: Kernel,
+        global_work_size: tuple[int, ...],
+        local_work_size: tuple[int, ...] | None,
+        global_offset: tuple[int, ...] | None = None,
+        wait_for: WaitList = None,
+        *,
+        g_times_l: bool = False,
+        allow_empty_ndrange: bool = False
+    ) -> Event: ...
+
+def have_gl() -> bool: ...
+
+class GLBuffer(MemoryObject):
+    def __init__(self, context: Context, flags: int, bufobj: int) -> None: ...
+
+    def get_gl_object_info(self) -> tuple[gl_object_type, int]: ...
+
+class GLRenderBuffer(MemoryObject):
+    def __init__(self, context: Context, flags: int, bufobj: int) -> None: ...
+
+    def get_gl_object_info(self) -> tuple[gl_object_type, int]: ...
+
+class GLTexture(Image):
+    def __init__(
+             self,
+             context: Context,
+             flags: int,
+             texture_target: int,
+             miplevel: int,
+             texture: int,
+             dims: int
+         ) -> None: ...
+
+    def get_gl_object_info(self) -> tuple[gl_object_type, int]: ...
+
+    def get_gl_texture_info(self, arg: gl_texture_info, /) -> int: ...
+
+def enqueue_acquire_gl_objects(
+    queue: CommandQueue,
+    mem_objects: object,
+    wait_for: WaitList = None) -> Event: ...
+
+def enqueue_release_gl_objects(
+    queue: CommandQueue,
+    mem_objects: object,
+    wait_for: WaitList = None) -> Event: ...
+
+def get_gl_context_info_khr(
+    properties: object,
+    param_name: int,
+    platform: object | None = None) -> object: ...
+
+def bitlog2(arg: int, /) -> int: ...
+
+class AllocatorBase:
+    def __call__(self, size: int) -> Buffer: ...
+
+class DeferredAllocator(AllocatorBase):
+    @overload
+    def __init__(self, arg: Context, /) -> None: ...
+
+    @overload
+    def __init__(self, queue: Context, mem_flags: int) -> None: ...
+
+class ImmediateAllocator(AllocatorBase):
+    @overload
+    def __init__(self, arg: CommandQueue, /) -> None: ...
+
+    @overload
+    def __init__(self, queue: CommandQueue, mem_flags: int) -> None: ...
+
+class PooledBuffer(MemoryObjectHolder):
+    def release(self) -> None: ...
+
+    def bind_to_queue(self, arg: CommandQueue, /) -> None: ...
+
+    def unbind_from_queue(self) -> None: ...
+
+class MemoryPool:
+    def __init__(self,
+             allocator: AllocatorBase,
+             leading_bits_in_bin_id: int = 4
+         ) -> None: ...
+
+    def allocate(self, size: int) -> PooledBuffer: ...
+
+    def __call__(self, size: int) -> PooledBuffer: ...
+
+    @property
+    def held_blocks(self) -> int: ...
+
+    @property
+    def active_blocks(self) -> int: ...
+
+    @property
+    def managed_bytes(self) -> int: ...
+
+    @property
+    def active_bytes(self) -> int: ...
+
+    def bin_number(self, arg: int, /) -> int: ...
+
+    def alloc_size(self, arg: int, /) -> int: ...
+
+    def free_held(self) -> None: ...
+
+    def stop_holding(self) -> None: ...
+
+class SVMAllocator:
+    def __init__(self,
+            context: Context,
+            alignment: int = 0,
+            flags: int = 1,
+            queue: CommandQueue | None = None
+         ) -> None: ...
+
+    def __call__(self, size: int) -> SVMAllocation: ...
+
+class PooledSVM(SVMPointer):
+    def release(self) -> None: ...
+
+    def enqueue_release(self) -> None: ...
+
+    @override
+    def __eq__(self, arg: PooledSVM, /) -> bool: ...
+
+    @override
+    def __hash__(self) -> int: ...
+
+    def bind_to_queue(self, arg: CommandQueue, /) -> None: ...
+
+    def unbind_from_queue(self) -> None: ...
+
+class SVMPool:
+    def __init__(self,
+            allocator: SVMAllocator,
+            leading_bits_in_bin_id: int = 4
+         ) -> None: ...
+
+    def __call__(self, size: int) -> PooledSVM: ...
+
+    @property
+    def held_blocks(self) -> int: ...
+
+    @property
+    def active_blocks(self) -> int: ...
+
+    @property
+    def managed_bytes(self) -> int: ...
+
+    @property
+    def active_bytes(self) -> int: ...
+
+    def bin_number(self, arg: int, /) -> int: ...
+
+    def alloc_size(self, arg: int, /) -> int: ...
+
+    def free_held(self) -> None: ...
+
+    def stop_holding(self) -> None: ...
+
+class _Program:
+    @overload
+    def __init__(self, arg1: Context, arg2: str | bytes) -> None: ...
+
+    @overload
+    def __init__(
+            self,
+            arg1: Context,
+            arg2: Sequence[Device],
+            arg3: Sequence[bytes]
+        ) -> None: ...
+
+    @staticmethod
+    def create_with_built_in_kernels(
+        context: Context,
+        devices: Sequence[Device],
+        kernel_names: str) -> _Program: ...
+
+    def kind(self) -> program_kind: ...
+
+    def get_info(self, arg: program_info, /) -> object: ...
+
+    def get_build_info(self, arg0: Device, arg1: int, /) -> object: ...
+
+    def _build(self,
+            options: bytes = b"",
+            devices: Sequence[Device] | None = None
+        ) -> None: ...
+
+    def compile(self,
+            options: bytes = b"",
+            devices: Sequence[Device] | None = None,
+            headers: object = []
+        ) -> None: ...
+
+    @staticmethod
+    def link(
+            context: Context,
+            programs: object,
+            options: bytes = b"",
+            devices: Sequence[Device] | None = None
+        ) -> _Program: ...
+
+    def set_specialization_constant(self, spec_id: int, buffer: object) -> None: ...
+
+    @overload
+    def __eq__(self, arg: _Program, /) -> bool: ...
+
+    @overload
+    def __eq__(self, obj: object | None) -> bool: ...
+
+    @override
+    def __hash__(self) -> int: ...
+
+    def all_kernels(self) -> Sequence[Kernel]: ...
+
+    @staticmethod
+    def from_int_ptr(int_ptr_value: int, retain: bool = True) -> _Program: ...
+
+    @property
+    def int_ptr(self) -> int: ...
+
+    _get_build_logs = pyopencl._monkeypatch.program_get_build_logs
+
+    build = pyopencl._monkeypatch.program_build
+
+def _create_program_with_il(arg0: Context, arg1: bytes, /) -> _Program: ...
+
+class _TestMemoryPool:
+    def __init__(self, leading_bits_in_bin_id: int = 4) -> None: ...
+
+    def allocate(self, arg: int, /) -> object: ...
+
+    @property
+    def held_blocks(self) -> int: ...
+
+    @property
+    def active_blocks(self) -> int: ...
+
+    @property
+    def managed_bytes(self) -> int: ...
+
+    @property
+    def active_bytes(self) -> int: ...
+
+    def bin_number(self, arg: int, /) -> int: ...
+
+    def alloc_size(self, arg: int, /) -> int: ...
+
+    def free_held(self) -> None: ...
+
+    def stop_holding(self) -> None: ...
+
+    def _set_trace(self, arg: bool, /) -> None: ...
+
+def get_apple_cgl_share_group() -> context_properties: ...