nvmath-python 1.0.0__cp314-cp314t-win_amd64.whl

nvmath/bindings/cublasMp.pxd

@@ -0,0 +1,85 @@
+# Copyright (c) 2025-2026, NVIDIA CORPORATION & AFFILIATES. ALL RIGHTS RESERVED.
+#
+# SPDX-License-Identifier: Apache-2.0
+#
+# This code was automatically generated with version 0.8.0, generator version 0.3.1.dev1303+g031f1197f. Do not modify it directly.
+
+from libc.stdint cimport intptr_t
+
+from .cycublasMp cimport *
+
+
+###############################################################################
+# Types
+###############################################################################
+
+ctypedef ncclComm_t ncclComm
+ctypedef cublasMpHandle_t Handle
+ctypedef cublasMpGrid_t Grid
+ctypedef cublasMpMatrixDescriptor_t MatrixDescriptor
+ctypedef cublasMpMatmulDescriptor_t MatmulDescriptor
+ctypedef cublasMpLoggerCallback_t LoggerCallback
+
+ctypedef cudaStream_t Stream
+ctypedef cudaDataType DataType
+ctypedef libraryPropertyType_t LibraryPropertyType
+
+
+###############################################################################
+# Enum
+###############################################################################
+
+ctypedef cublasMpStatus_t _Status
+ctypedef cublasMpGridLayout_t _GridLayout
+ctypedef cublasMpMatmulDescriptorAttribute_t _MatmulDescriptorAttribute
+ctypedef cublasMpMatmulAlgoType_t _MatmulAlgoType
+ctypedef cublasMpMatmulEpilogue_t _MatmulEpilogue
+ctypedef cublasMpMatmulMatrixScale_t _MatmulMatrixScale
+ctypedef cublasMpEmulationStrategy_t _EmulationStrategy
+ctypedef cublasMpResultScheme_t _ResultScheme
+
+
+###############################################################################
+# Functions
+###############################################################################
+
+cpdef intptr_t create(intptr_t stream) except? 0
+cpdef destroy(intptr_t handle)
+cpdef set_stream(intptr_t handle, intptr_t stream)
+cpdef intptr_t get_stream(intptr_t handle) except? 0
+cpdef set_emulation_strategy(intptr_t handle, int emulation_strategy)
+cpdef int get_emulation_strategy(intptr_t handle) except? -1
+cpdef int get_version() except? 0
+cpdef intptr_t grid_create(int64_t nprow, int64_t npcol, int layout, intptr_t comm) except? 0
+cpdef grid_destroy(intptr_t grid)
+cpdef buffer_register(intptr_t grid, intptr_t ptr, size_t size)
+cpdef buffer_deregister(intptr_t grid, intptr_t ptr)
+cpdef intptr_t matrix_descriptor_create(int64_t m, int64_t n, int64_t mb, int64_t nb, int64_t rsrc, int64_t csrc, int64_t lld, int type, intptr_t grid) except? 0
+cpdef matrix_descriptor_destroy(intptr_t desc)
+cpdef matrix_descriptor_init(int64_t m, int64_t n, int64_t mb, int64_t nb, int64_t rsrc, int64_t csrc, int64_t lld, int type, intptr_t grid, intptr_t desc)
+cpdef intptr_t matmul_descriptor_create(int compute_type) except? 0
+cpdef matmul_descriptor_destroy(intptr_t matmul_desc)
+cpdef matmul_descriptor_init(intptr_t matmul_desc, int compute_type)
+cpdef get_matmul_descriptor_attribute_dtype(int attr)
+cpdef matmul_descriptor_set_attribute(intptr_t matmul_desc, int attr, intptr_t buf, size_t size_in_bytes)
+cpdef matmul_descriptor_get_attribute(intptr_t matmul_desc, int attr, intptr_t buf, size_t size_in_bytes, intptr_t size_written)
+cpdef tuple trsm_buffer_size(intptr_t handle, int side, int uplo, int trans, int diag, int64_t m, int64_t n, intptr_t alpha, intptr_t a, int64_t ia, int64_t ja, intptr_t desc_a, intptr_t b, int64_t ib, int64_t jb, intptr_t desc_b, int compute_type)
+cpdef trsm(intptr_t handle, int side, int uplo, int trans, int diag, int64_t m, int64_t n, intptr_t alpha, intptr_t a, int64_t ia, int64_t ja, intptr_t desc_a, intptr_t b, int64_t ib, int64_t jb, intptr_t desc_b, int compute_type, intptr_t d_work, size_t workspace_size_in_bytes_on_device, intptr_t h_work, size_t workspace_size_in_bytes_on_host)
+cpdef tuple gemm_buffer_size(intptr_t handle, int trans_a, int trans_b, int64_t m, int64_t n, int64_t k, intptr_t alpha, intptr_t a, int64_t ia, int64_t ja, intptr_t desc_a, intptr_t b, int64_t ib, int64_t jb, intptr_t desc_b, intptr_t beta, intptr_t c, int64_t ic, int64_t jc, intptr_t desc_c, int compute_type)
+cpdef gemm(intptr_t handle, int trans_a, int trans_b, int64_t m, int64_t n, int64_t k, intptr_t alpha, intptr_t a, int64_t ia, int64_t ja, intptr_t desc_a, intptr_t b, int64_t ib, int64_t jb, intptr_t desc_b, intptr_t beta, intptr_t c, int64_t ic, int64_t jc, intptr_t desc_c, int compute_type, intptr_t d_work, size_t workspace_size_in_bytes_on_device, intptr_t h_work, size_t workspace_size_in_bytes_on_host)
+cpdef tuple matmul_buffer_size(intptr_t handle, intptr_t matmul_desc, int64_t m, int64_t n, int64_t k, intptr_t alpha, intptr_t a, int64_t ia, int64_t ja, intptr_t desc_a, intptr_t b, int64_t ib, int64_t jb, intptr_t desc_b, intptr_t beta, intptr_t c, int64_t ic, int64_t jc, intptr_t desc_c, intptr_t d, int64_t id, int64_t jd, intptr_t desc_d)
+cpdef matmul(intptr_t handle, intptr_t matmul_desc, int64_t m, int64_t n, int64_t k, intptr_t alpha, intptr_t a, int64_t ia, int64_t ja, intptr_t desc_a, intptr_t b, int64_t ib, int64_t jb, intptr_t desc_b, intptr_t beta, intptr_t c, int64_t ic, int64_t jc, intptr_t desc_c, intptr_t d, int64_t id, int64_t jd, intptr_t desc_d, intptr_t d_work, size_t workspace_size_in_bytes_on_device, intptr_t h_work, size_t workspace_size_in_bytes_on_host)
+cpdef tuple syrk_buffer_size(intptr_t handle, int uplo, int trans, int64_t n, int64_t k, intptr_t alpha, intptr_t a, int64_t ia, int64_t ja, intptr_t desc_a, intptr_t beta, intptr_t c, int64_t ic, int64_t jc, intptr_t desc_c, int compute_type)
+cpdef syrk(intptr_t handle, int uplo, int trans, int64_t n, int64_t k, intptr_t alpha, intptr_t a, int64_t ia, int64_t ja, intptr_t desc_a, intptr_t beta, intptr_t c, int64_t ic, int64_t jc, intptr_t desc_c, int compute_type, intptr_t d_work, size_t workspace_size_in_bytes_on_device, intptr_t h_work, size_t workspace_size_in_bytes_on_host)
+cpdef int64_t numroc(int64_t n, int64_t nb, uint32_t iproc, uint32_t isrcproc, uint32_t nprocs) except? -1
+cpdef str get_status_string(int status)
+cpdef tuple gemr2d_buffer_size(intptr_t handle, int64_t m, int64_t n, intptr_t a, int64_t ia, int64_t ja, intptr_t desc_a, intptr_t b, int64_t ib, int64_t jb, intptr_t desc_b, intptr_t global_comm)
+cpdef gemr2d(intptr_t handle, int64_t m, int64_t n, intptr_t a, int64_t ia, int64_t ja, intptr_t desc_a, intptr_t b, int64_t ib, int64_t jb, intptr_t desc_b, intptr_t d_work, size_t workspace_size_in_bytes_on_device, intptr_t h_work, size_t workspace_size_in_bytes_on_host, intptr_t global_comm)
+cpdef tuple trmr2d_buffer_size(intptr_t handle, int uplo, int diag, int64_t m, int64_t n, intptr_t a, int64_t ia, int64_t ja, intptr_t desc_a, intptr_t b, int64_t ib, int64_t jb, intptr_t desc_b, intptr_t global_comm)
+cpdef trmr2d(intptr_t handle, int uplo, int diag, int64_t m, int64_t n, intptr_t a, int64_t ia, int64_t ja, intptr_t desc_a, intptr_t b, int64_t ib, int64_t jb, intptr_t desc_b, intptr_t d_work, size_t workspace_size_in_bytes_on_device, intptr_t h_work, size_t workspace_size_in_bytes_on_host, intptr_t global_comm)
+cpdef tuple geadd_buffer_size(intptr_t handle, int trans, int64_t m, int64_t n, intptr_t alpha, intptr_t a, int64_t ia, int64_t ja, intptr_t desc_a, intptr_t beta, intptr_t c, int64_t ic, int64_t jc, intptr_t desc_c)
+cpdef geadd(intptr_t handle, int trans, int64_t m, int64_t n, intptr_t alpha, intptr_t a, int64_t ia, int64_t ja, intptr_t desc_a, intptr_t beta, intptr_t c, int64_t ic, int64_t jc, intptr_t desc_c, intptr_t d_work, size_t workspace_size_in_bytes_on_device, intptr_t h_work, size_t workspace_size_in_bytes_on_host)
+cpdef tuple tradd_buffer_size(intptr_t handle, int uplo, int trans, int64_t m, int64_t n, intptr_t alpha, intptr_t a, int64_t ia, int64_t ja, intptr_t desc_a, intptr_t beta, intptr_t c, int64_t ic, int64_t jc, intptr_t desc_c)
+cpdef tradd(intptr_t handle, int uplo, int trans, int64_t m, int64_t n, intptr_t alpha, intptr_t a, int64_t ia, int64_t ja, intptr_t desc_a, intptr_t beta, intptr_t c, int64_t ic, int64_t jc, intptr_t desc_c, intptr_t d_work, size_t workspace_size_in_bytes_on_device, intptr_t h_work, size_t workspace_size_in_bytes_on_host)
+cpdef intptr_t malloc(intptr_t grid, size_t size) except? 0
+cpdef free(intptr_t grid, intptr_t ptr)

nvmath/bindings/cublasMp.pyi

@@ -0,0 +1,267 @@
+# Copyright (c) 2026, NVIDIA CORPORATION & AFFILIATES. ALL RIGHTS RESERVED.
+#
+# SPDX-License-Identifier: Apache-2.0
+
+import _cython_3_2_5
+import enum
+from typing import Any, Callable, ClassVar
+
+__pyx_capi__: dict
+__test__: dict
+buffer_deregister: _cython_3_2_5.cython_function_or_method
+buffer_register: _cython_3_2_5.cython_function_or_method
+check_status: _cython_3_2_5.cython_function_or_method
+create: _cython_3_2_5.cython_function_or_method
+destroy: _cython_3_2_5.cython_function_or_method
+free: _cython_3_2_5.cython_function_or_method
+geadd: _cython_3_2_5.cython_function_or_method
+geadd_buffer_size: _cython_3_2_5.cython_function_or_method
+gemm: _cython_3_2_5.cython_function_or_method
+gemm_buffer_size: _cython_3_2_5.cython_function_or_method
+gemr2d: _cython_3_2_5.cython_function_or_method
+gemr2d_buffer_size: _cython_3_2_5.cython_function_or_method
+get_emulation_strategy: _cython_3_2_5.cython_function_or_method
+get_matmul_descriptor_attribute_dtype: _cython_3_2_5.cython_function_or_method
+get_status_string: _cython_3_2_5.cython_function_or_method
+get_stream: _cython_3_2_5.cython_function_or_method
+get_version: _cython_3_2_5.cython_function_or_method
+grid_create: _cython_3_2_5.cython_function_or_method
+grid_destroy: _cython_3_2_5.cython_function_or_method
+malloc: _cython_3_2_5.cython_function_or_method
+matmul: _cython_3_2_5.cython_function_or_method
+matmul_buffer_size: _cython_3_2_5.cython_function_or_method
+matmul_descriptor_create: _cython_3_2_5.cython_function_or_method
+matmul_descriptor_destroy: _cython_3_2_5.cython_function_or_method
+matmul_descriptor_get_attribute: _cython_3_2_5.cython_function_or_method
+matmul_descriptor_init: _cython_3_2_5.cython_function_or_method
+matmul_descriptor_set_attribute: _cython_3_2_5.cython_function_or_method
+matrix_descriptor_create: _cython_3_2_5.cython_function_or_method
+matrix_descriptor_destroy: _cython_3_2_5.cython_function_or_method
+matrix_descriptor_init: _cython_3_2_5.cython_function_or_method
+numroc: _cython_3_2_5.cython_function_or_method
+set_emulation_strategy: _cython_3_2_5.cython_function_or_method
+set_stream: _cython_3_2_5.cython_function_or_method
+syrk: _cython_3_2_5.cython_function_or_method
+syrk_buffer_size: _cython_3_2_5.cython_function_or_method
+tradd: _cython_3_2_5.cython_function_or_method
+tradd_buffer_size: _cython_3_2_5.cython_function_or_method
+trmr2d: _cython_3_2_5.cython_function_or_method
+trmr2d_buffer_size: _cython_3_2_5.cython_function_or_method
+trsm: _cython_3_2_5.cython_function_or_method
+trsm_buffer_size: _cython_3_2_5.cython_function_or_method
+
+class EmulationStrategy(enum.IntEnum):
+    """
+    See `cublasMpEmulationStrategy_t`.
+    """
+    __new__: ClassVar[Callable] = ...
+    DEFAULT: ClassVar[EmulationStrategy] = ...
+    EAGER: ClassVar[EmulationStrategy] = ...
+    PERFORMANT: ClassVar[EmulationStrategy] = ...
+    _generate_next_value_: ClassVar[Callable] = ...
+    _hashable_values_: ClassVar[list] = ...
+    _member_map_: ClassVar[dict] = ...
+    _member_names_: ClassVar[list] = ...
+    _member_type_: ClassVar[type[int]] = ...
+    _unhashable_values_: ClassVar[list] = ...
+    _unhashable_values_map_: ClassVar[dict] = ...
+    _use_args_: ClassVar[bool] = ...
+    _value2member_map_: ClassVar[dict] = ...
+    def __format__(self, *args, **kwargs) -> str:
+        """Convert to a string according to format_spec."""
+
+class GridLayout(enum.IntEnum):
+    """
+    See `cublasMpGridLayout_t`.
+    """
+    __new__: ClassVar[Callable] = ...
+    COL_MAJOR: ClassVar[GridLayout] = ...
+    ROW_MAJOR: ClassVar[GridLayout] = ...
+    _generate_next_value_: ClassVar[Callable] = ...
+    _hashable_values_: ClassVar[list] = ...
+    _member_map_: ClassVar[dict] = ...
+    _member_names_: ClassVar[list] = ...
+    _member_type_: ClassVar[type[int]] = ...
+    _unhashable_values_: ClassVar[list] = ...
+    _unhashable_values_map_: ClassVar[dict] = ...
+    _use_args_: ClassVar[bool] = ...
+    _value2member_map_: ClassVar[dict] = ...
+    def __format__(self, *args, **kwargs) -> str:
+        """Convert to a string according to format_spec."""
+
+class MatmulAlgoType(enum.IntEnum):
+    """
+    See `cublasMpMatmulAlgoType_t`.
+    """
+    __new__: ClassVar[Callable] = ...
+    ATOMIC_MULTICAST: ClassVar[MatmulAlgoType] = ...
+    ATOMIC_P2P: ClassVar[MatmulAlgoType] = ...
+    DEFAULT: ClassVar[MatmulAlgoType] = ...
+    NO_OVERLAP: ClassVar[MatmulAlgoType] = ...
+    SPLIT_MULTICAST: ClassVar[MatmulAlgoType] = ...
+    SPLIT_P2P: ClassVar[MatmulAlgoType] = ...
+    _generate_next_value_: ClassVar[Callable] = ...
+    _hashable_values_: ClassVar[list] = ...
+    _member_map_: ClassVar[dict] = ...
+    _member_names_: ClassVar[list] = ...
+    _member_type_: ClassVar[type[int]] = ...
+    _unhashable_values_: ClassVar[list] = ...
+    _unhashable_values_map_: ClassVar[dict] = ...
+    _use_args_: ClassVar[bool] = ...
+    _value2member_map_: ClassVar[dict] = ...
+    def __format__(self, *args, **kwargs) -> str:
+        """Convert to a string according to format_spec."""
+
+class MatmulDescriptorAttribute(enum.IntEnum):
+    """
+    See `cublasMpMatmulDescriptorAttribute_t`.
+    """
+    __new__: ClassVar[Callable] = ...
+    ALGO_TYPE: ClassVar[MatmulDescriptorAttribute] = ...
+    AMAX_D_POINTER: ClassVar[MatmulDescriptorAttribute] = ...
+    A_SCALE_MODE: ClassVar[MatmulDescriptorAttribute] = ...
+    A_SCALE_POINTER: ClassVar[MatmulDescriptorAttribute] = ...
+    BIAS_BATCH_STRIDE: ClassVar[MatmulDescriptorAttribute] = ...
+    BIAS_DATA_TYPE: ClassVar[MatmulDescriptorAttribute] = ...
+    BIAS_POINTER: ClassVar[MatmulDescriptorAttribute] = ...
+    BIAS_RESULT_SCHEME: ClassVar[MatmulDescriptorAttribute] = ...
+    B_SCALE_MODE: ClassVar[MatmulDescriptorAttribute] = ...
+    B_SCALE_POINTER: ClassVar[MatmulDescriptorAttribute] = ...
+    COMMUNICATION_SM_COUNT: ClassVar[MatmulDescriptorAttribute] = ...
+    COMMUNICATION_TYPE: ClassVar[MatmulDescriptorAttribute] = ...
+    COMPUTE_TYPE: ClassVar[MatmulDescriptorAttribute] = ...
+    C_SCALE_MODE: ClassVar[MatmulDescriptorAttribute] = ...
+    C_SCALE_POINTER: ClassVar[MatmulDescriptorAttribute] = ...
+    D_OUT_SCALE_MODE: ClassVar[MatmulDescriptorAttribute] = ...
+    D_OUT_SCALE_POINTER: ClassVar[MatmulDescriptorAttribute] = ...
+    D_SCALE_MODE: ClassVar[MatmulDescriptorAttribute] = ...
+    D_SCALE_POINTER: ClassVar[MatmulDescriptorAttribute] = ...
+    EPILOGUE: ClassVar[MatmulDescriptorAttribute] = ...
+    EPILOGUE_AUX_AMAX_POINTER: ClassVar[MatmulDescriptorAttribute] = ...
+    EPILOGUE_AUX_BATCH_STRIDE: ClassVar[MatmulDescriptorAttribute] = ...
+    EPILOGUE_AUX_DATA_TYPE: ClassVar[MatmulDescriptorAttribute] = ...
+    EPILOGUE_AUX_LD: ClassVar[MatmulDescriptorAttribute] = ...
+    EPILOGUE_AUX_POINTER: ClassVar[MatmulDescriptorAttribute] = ...
+    EPILOGUE_AUX_SCALE_MODE: ClassVar[MatmulDescriptorAttribute] = ...
+    EPILOGUE_AUX_SCALE_POINTER: ClassVar[MatmulDescriptorAttribute] = ...
+    TRANSA: ClassVar[MatmulDescriptorAttribute] = ...
+    TRANSB: ClassVar[MatmulDescriptorAttribute] = ...
+    _generate_next_value_: ClassVar[Callable] = ...
+    _hashable_values_: ClassVar[list] = ...
+    _member_map_: ClassVar[dict] = ...
+    _member_names_: ClassVar[list] = ...
+    _member_type_: ClassVar[type[int]] = ...
+    _unhashable_values_: ClassVar[list] = ...
+    _unhashable_values_map_: ClassVar[dict] = ...
+    _use_args_: ClassVar[bool] = ...
+    _value2member_map_: ClassVar[dict] = ...
+    def __format__(self, *args, **kwargs) -> str:
+        """Convert to a string according to format_spec."""
+
+class MatmulEpilogue(enum.IntEnum):
+    """
+    See `cublasMpMatmulEpilogue_t`.
+    """
+    __new__: ClassVar[Callable] = ...
+    ALLREDUCE: ClassVar[MatmulEpilogue] = ...
+    BGRADA: ClassVar[MatmulEpilogue] = ...
+    BGRADB: ClassVar[MatmulEpilogue] = ...
+    BIAS: ClassVar[MatmulEpilogue] = ...
+    DEFAULT: ClassVar[MatmulEpilogue] = ...
+    DGELU: ClassVar[MatmulEpilogue] = ...
+    DGELU_BGRAD: ClassVar[MatmulEpilogue] = ...
+    DRELU: ClassVar[MatmulEpilogue] = ...
+    DRELU_BGRAD: ClassVar[MatmulEpilogue] = ...
+    GELU: ClassVar[MatmulEpilogue] = ...
+    GELU_AUX: ClassVar[MatmulEpilogue] = ...
+    GELU_AUX_BIAS: ClassVar[MatmulEpilogue] = ...
+    GELU_BIAS: ClassVar[MatmulEpilogue] = ...
+    RELU: ClassVar[MatmulEpilogue] = ...
+    RELU_AUX: ClassVar[MatmulEpilogue] = ...
+    RELU_AUX_BIAS: ClassVar[MatmulEpilogue] = ...
+    RELU_BIAS: ClassVar[MatmulEpilogue] = ...
+    _generate_next_value_: ClassVar[Callable] = ...
+    _hashable_values_: ClassVar[list] = ...
+    _member_map_: ClassVar[dict] = ...
+    _member_names_: ClassVar[list] = ...
+    _member_type_: ClassVar[type[int]] = ...
+    _unhashable_values_: ClassVar[list] = ...
+    _unhashable_values_map_: ClassVar[dict] = ...
+    _use_args_: ClassVar[bool] = ...
+    _value2member_map_: ClassVar[dict] = ...
+    def __format__(self, *args, **kwargs) -> str:
+        """Convert to a string according to format_spec."""
+
+class MatmulMatrixScale(enum.IntEnum):
+    """
+    See `cublasMpMatmulMatrixScale_t`.
+    """
+    __new__: ClassVar[Callable] = ...
+    BLK128x128_FP32: ClassVar[MatmulMatrixScale] = ...
+    OUTER_VEC_FP32: ClassVar[MatmulMatrixScale] = ...
+    SCALAR_FP32: ClassVar[MatmulMatrixScale] = ...
+    VEC128_FP32: ClassVar[MatmulMatrixScale] = ...
+    VEC16_UE4M3: ClassVar[MatmulMatrixScale] = ...
+    VEC32_UE8M0: ClassVar[MatmulMatrixScale] = ...
+    _generate_next_value_: ClassVar[Callable] = ...
+    _hashable_values_: ClassVar[list] = ...
+    _member_map_: ClassVar[dict] = ...
+    _member_names_: ClassVar[list] = ...
+    _member_type_: ClassVar[type[int]] = ...
+    _unhashable_values_: ClassVar[list] = ...
+    _unhashable_values_map_: ClassVar[dict] = ...
+    _use_args_: ClassVar[bool] = ...
+    _value2member_map_: ClassVar[dict] = ...
+    def __format__(self, *args, **kwargs) -> str:
+        """Convert to a string according to format_spec."""
+
+class ResultScheme(enum.IntEnum):
+    """
+    See `cublasMpResultScheme_t`.
+    """
+    __new__: ClassVar[Callable] = ...
+    DEFAULT: ClassVar[ResultScheme] = ...
+    FULL: ClassVar[ResultScheme] = ...
+    PARTIAL: ClassVar[ResultScheme] = ...
+    _generate_next_value_: ClassVar[Callable] = ...
+    _hashable_values_: ClassVar[list] = ...
+    _member_map_: ClassVar[dict] = ...
+    _member_names_: ClassVar[list] = ...
+    _member_type_: ClassVar[type[int]] = ...
+    _unhashable_values_: ClassVar[list] = ...
+    _unhashable_values_map_: ClassVar[dict] = ...
+    _use_args_: ClassVar[bool] = ...
+    _value2member_map_: ClassVar[dict] = ...
+    def __format__(self, *args, **kwargs) -> str:
+        """Convert to a string according to format_spec."""
+
+class Status(enum.IntEnum):
+    """
+    See `cublasMpStatus_t`.
+    """
+    __new__: ClassVar[Callable] = ...
+    ALLOCATION_FAILED: ClassVar[Status] = ...
+    ARCHITECTURE_MISMATCH: ClassVar[Status] = ...
+    EXECUTION_FAILED: ClassVar[Status] = ...
+    INTERNAL_ERROR: ClassVar[Status] = ...
+    INVALID_VALUE: ClassVar[Status] = ...
+    NOT_INITIALIZED: ClassVar[Status] = ...
+    NOT_SUPPORTED: ClassVar[Status] = ...
+    SUCCESS: ClassVar[Status] = ...
+    _generate_next_value_: ClassVar[Callable] = ...
+    _hashable_values_: ClassVar[list] = ...
+    _member_map_: ClassVar[dict] = ...
+    _member_names_: ClassVar[list] = ...
+    _member_type_: ClassVar[type[int]] = ...
+    _unhashable_values_: ClassVar[list] = ...
+    _unhashable_values_map_: ClassVar[dict] = ...
+    _use_args_: ClassVar[bool] = ...
+    _value2member_map_: ClassVar[dict] = ...
+    def __format__(self, *args, **kwargs) -> str:
+        """Convert to a string according to format_spec."""
+
+class cuBLASMpError(Exception):
+    def __init__(self, status) -> Any:
+        """cuBLASMpError.__init__(self, status)"""
+    def __reduce__(self) -> Any:
+        """cuBLASMpError.__reduce__(self)"""

nvmath/bindings/cudss.pxd

@@ -0,0 +1,98 @@
+# Copyright (c) 2025-2026, NVIDIA CORPORATION & AFFILIATES. ALL RIGHTS RESERVED.
+#
+# SPDX-License-Identifier: Apache-2.0
+#
+# This code was automatically generated with version 0.8.0, generator version 0.3.1.dev1725+g07a86fe3d.d20260603. Do not modify it directly.
+
+from libc.stdint cimport intptr_t
+
+from .cycudss cimport *
+
+
+###############################################################################
+# Types
+###############################################################################
+
+ctypedef cudssHandle_t Handle
+ctypedef cudssMatrix_t Matrix
+ctypedef cudssData_t Data
+ctypedef cudssConfig_t Config
+ctypedef cudssDeviceMemHandler_t DeviceMemHandler
+ctypedef cudssDistributedInterface_t DistributedInterface
+ctypedef cudssThreadingInterface_t ThreadingInterface
+ctypedef cudssLoggerCallback_t LoggerCallback
+
+ctypedef cudaStream_t Stream
+ctypedef libraryPropertyType_t LibraryPropertyType
+
+
+###############################################################################
+# Enum
+###############################################################################
+
+ctypedef cudssDataType_t _DataType
+ctypedef cudssConfigParam_t _ConfigParam
+ctypedef cudssDataParam_t _DataParam
+ctypedef cudssPhase_t _Phase
+ctypedef cudssStatus_t _Status
+ctypedef cudssMatrixType_t _MatrixType
+ctypedef cudssMatrixViewType_t _MatrixViewType
+ctypedef cudssIndexBase_t _IndexBase
+ctypedef cudssLayout_t _Layout
+ctypedef cudssReorderingAlg_t _ReorderingAlg
+ctypedef cudssFactorizationAlg_t _FactorizationAlg
+ctypedef cudssPivotEpsilonAlg_t _PivotEpsilonAlg
+ctypedef cudssSolveAlg_t _SolveAlg
+ctypedef cudssMatchingAlg_t _MatchingAlg
+ctypedef cudssPivotType_t _PivotType
+ctypedef cudssMatrixFormat_t _MatrixFormat
+ctypedef cudssOpType_t _OpType
+
+
+###############################################################################
+# Functions
+###############################################################################
+
+cpdef get_config_param_dtype(int attr)
+cpdef config_set(intptr_t config, int param, intptr_t value, size_t size_in_bytes)
+cpdef config_get(intptr_t config, int param, intptr_t value, size_t size_in_bytes, intptr_t size_written)
+cpdef get_data_param_dtype(int attr)
+cpdef data_set(intptr_t handle, intptr_t data, int param, intptr_t value, size_t size_in_bytes)
+cpdef data_get(intptr_t handle, intptr_t data, int param, intptr_t value, size_t size_in_bytes, intptr_t size_written)
+cpdef execute(intptr_t handle, int phase, intptr_t solver_config, intptr_t solver_data, intptr_t input_matrix, intptr_t solution, intptr_t rhs)
+cpdef set_stream(intptr_t handle, intptr_t stream)
+cpdef set_mg_streams(intptr_t handle, streams, int stream_count)
+cpdef set_comm_layer(intptr_t handle, comm_lib_file_name)
+cpdef set_threading_layer(intptr_t handle, thr_lib_file_name)
+cpdef intptr_t config_create() except? 0
+cpdef config_destroy(intptr_t solver_config)
+cpdef intptr_t data_create(intptr_t handle) except? 0
+cpdef data_destroy(intptr_t handle, intptr_t solver_data)
+cpdef intptr_t create() except? 0
+cpdef intptr_t create_mg(int device_count, device_indices) except? 0
+cpdef destroy(intptr_t handle)
+cpdef int get_property(int property_type) except? -1
+cpdef intptr_t matrix_create_dn(int64_t nrows, int64_t ncols, int64_t ld, intptr_t values, int value_type, int layout) except? 0
+cpdef intptr_t matrix_create_csr(int64_t nrows, int64_t ncols, int64_t nnz, intptr_t row_start, intptr_t row_end, intptr_t col_indices, intptr_t values, int offset_type, int index_type, int value_type, int mtype, int mview, int index_base) except? 0
+cpdef intptr_t matrix_create_batch_dn(int64_t batch_count, intptr_t nrows, intptr_t ncols, intptr_t ld, intptr_t values, int integer_type, int value_type, int layout) except? 0
+cpdef intptr_t matrix_create_batch_csr(int64_t batch_count, intptr_t nrows, intptr_t ncols, intptr_t nnz, intptr_t row_start, intptr_t row_end, intptr_t col_indices, intptr_t values, int offset_type, int index_type, int value_type, int mtype, int mview, int index_base) except? 0
+cpdef matrix_destroy(intptr_t matrix)
+cpdef matrix_get_dn(intptr_t matrix, intptr_t nrows, intptr_t ncols, intptr_t ld, intptr_t values, intptr_t type, intptr_t layout)
+cpdef matrix_get_csr(intptr_t matrix, intptr_t nrows, intptr_t ncols, intptr_t nnz, intptr_t row_start, intptr_t row_end, intptr_t col_indices, intptr_t values, intptr_t offset_type, intptr_t index_type, intptr_t value_type, intptr_t mtype, intptr_t mview, intptr_t index_base)
+cpdef matrix_set_values(intptr_t matrix, intptr_t values)
+cpdef matrix_set_csr_pointers(intptr_t matrix, intptr_t row_offsets, intptr_t row_end, intptr_t col_indices, intptr_t values)
+cpdef matrix_get_batch_dn(intptr_t matrix, intptr_t batch_count, intptr_t nrows, intptr_t ncols, intptr_t ld, intptr_t values, intptr_t index_type, intptr_t value_type, intptr_t layout)
+cpdef matrix_get_batch_csr(intptr_t matrix, intptr_t batch_count, intptr_t nrows, intptr_t ncols, intptr_t nnz, intptr_t row_start, intptr_t row_end, intptr_t col_indices, intptr_t values, intptr_t offset_type, intptr_t index_type, intptr_t value_type, intptr_t mtype, intptr_t mview, intptr_t index_base)
+cpdef matrix_set_batch_values(intptr_t matrix, intptr_t values)
+cpdef matrix_set_batch_csr_pointers(intptr_t matrix, intptr_t row_offsets, intptr_t row_end, intptr_t col_indices, intptr_t values)
+cpdef int matrix_get_format(intptr_t matrix) except? -1
+cpdef matrix_set_distribution_row1d(intptr_t matrix, int64_t first_row, int64_t last_row)
+cpdef matrix_get_distribution_row1d(intptr_t matrix, intptr_t first_row, intptr_t last_row)
+cpdef get_device_mem_handler(intptr_t handle, intptr_t handler)
+cpdef set_device_mem_handler(intptr_t handle, intptr_t handler)
+cpdef logger_open_file(log_file)
+cpdef logger_set_level(int level)
+cpdef logger_set_mask(int mask)
+cpdef logger_force_disable()
+
+cpdef bint set_async_workspace_allocator(intptr_t cudss_handle, intptr_t pool_handle) except? -1