nvfatbin

Note

The nvfatbin bindings are not supported on nvFatbin installations <12.4. Ensure the installed CUDA toolkit’s nvFatbin version is >=12.4.

The Tile IR API (cuda.bindings.nvfatbin.add_tile_ir()) is only available in CUDA 13.1+.

Functions

NvFatbin defines the following functions for creating and populating fatbinaries.

cuda.bindings.nvfatbin.create(options, size_t options_count) → intptr_t

nvFatbinCreate creates a new handle.

:param options: An array of strings, each containing a single option. It can be:

                - an :class:`int` as the pointer address to the nested sequence, or
                - a Python sequence of :class:`int`\s, each of which is a pointer address
                  to a valid sequence of 'char', or
                - a nested Python sequence of ``str``.
:type options: object
:param options_count: Number of options.
:type options_count: size_t

:returns: Address of nvFatbin handle.
:rtype: intptr_t

.. seealso:: `nvFatbinCreate`

cuda.bindings.nvfatbin.destroy(intptr_t handle)

nvFatbinDestroy frees the memory associated with the given handle.

:param handle: nvFatbin handle.
:type handle: intptr_t

.. seealso:: `nvFatbinDestroy`

cuda.bindings.nvfatbin.add_ptx(

intptr_t handle,

code,

size_t size,

arch,

identifier,

options_cmd_line,

)

nvFatbinAddPTX adds PTX to the fatbinary.

:param handle: nvFatbin handle.
:type handle: intptr_t
:param code: The PTX code.
:type code: bytes
:param size: The size of the PTX code.
:type size: size_t
:param arch: The numerical architecture that this PTX is for (the XX of any sm_XX, lto_XX, or compute_XX).
:type arch: str
:param identifier: Name of the PTX, useful when extracting the fatbin with tools like cuobjdump.
:type identifier: str
:param options_cmd_line: Options used during JIT compilation.
:type options_cmd_line: str

.. seealso:: `nvFatbinAddPTX`

cuda.bindings.nvfatbin.add_cubin(intptr_t handle, code, size_t size, arch, identifier)

nvFatbinAddCubin adds a CUDA binary to the fatbinary.

:param handle: nvFatbin handle.
:type handle: intptr_t
:param code: The cubin.
:type code: bytes
:param size: The size of the cubin.
:type size: size_t
:param arch: The numerical architecture that this cubin is for (the XX of any sm_XX, lto_XX, or compute_XX).
:type arch: str
:param identifier: Name of the cubin, useful when extracting the fatbin with tools like cuobjdump.
:type identifier: str

.. seealso:: `nvFatbinAddCubin`

cuda.bindings.nvfatbin.add_ltoir(

intptr_t handle,

code,

size_t size,

arch,

identifier,

options_cmd_line,

)

nvFatbinAddLTOIR adds LTOIR to the fatbinary.

:param handle: nvFatbin handle.
:type handle: intptr_t
:param code: The LTOIR code.
:type code: bytes
:param size: The size of the LTOIR code.
:type size: size_t
:param arch: The numerical architecture that this LTOIR is for (the XX of any sm_XX, lto_XX, or compute_XX).
:type arch: str
:param identifier: Name of the LTOIR, useful when extracting the fatbin with tools like cuobjdump.
:type identifier: str
:param options_cmd_line: Options used during JIT compilation.
:type options_cmd_line: str

.. seealso:: `nvFatbinAddLTOIR`

cuda.bindings.nvfatbin.add_reloc(intptr_t handle, code, size_t size)

nvFatbinAddReloc adds relocatable PTX entries from a host object to the fatbinary.

:param handle: nvFatbin handle.
:type handle: intptr_t
:param code: The host object image.
:type code: bytes
:param size: The size of the host object image code.
:type size: size_t

.. seealso:: `nvFatbinAddReloc`

cuda.bindings.nvfatbin.add_tile_ir(

intptr_t handle,

code,

size_t size,

identifier,

options_cmd_line,

)

nvFatbinAddTileIR adds Tile IR to the fatbinary.

:param handle: nvFatbin handle.
:type handle: intptr_t
:param code: The Tile IR.
:type code: bytes
:param size: The size of the Tile IR.
:type size: size_t
:param identifier: Name of the Tile IR, useful when extracting the fatbin with tools like cuobjdump.
:type identifier: str
:param options_cmd_line: Options used during JIT compilation.
:type options_cmd_line: str

.. seealso:: `nvFatbinAddTileIR`

cuda.bindings.nvfatbin.size(intptr_t handle) → size_t

nvFatbinSize returns the fatbinary's size.

:param handle: nvFatbin handle.
:type handle: intptr_t

:returns: The fatbinary's size.
:rtype: size_t

.. seealso:: `nvFatbinSize`

cuda.bindings.nvfatbin.get(intptr_t handle, buffer)

nvFatbinGet returns the completed fatbinary.

:param handle: nvFatbin handle.
:type handle: intptr_t
:param buffer: memory to store fatbinary.
:type buffer: bytes

.. seealso:: `nvFatbinGet`

cuda.bindings.nvfatbin.get_error_string(int result) → str

nvFatbinGetErrorString returns an error description string for each error code.

:param result: error code.
:type result: Result

.. seealso:: `nvFatbinGetErrorString`

cuda.bindings.nvfatbin.version() → tuple

nvFatbinVersion returns the current version of nvFatbin.

:returns:

          - unsigned int: The major version.
          - unsigned int: The minor version.
:rtype: A 2-tuple containing

.. seealso:: `nvFatbinVersion`

Types

class cuda.bindings.nvfatbin.Result(value: int)

The enumerated type nvFatbinResult defines API call result codes.
nvFatbin APIs return nvFatbinResult codes to indicate the result.

See `nvFatbinResult`.

SUCCESS = <Result.SUCCESS: 0>

ERROR_INTERNAL = <Result.ERROR_INTERNAL: 1>

ERROR_ELF_ARCH_MISMATCH = <Result.ERROR_ELF_ARCH_MISMATCH: 2>

ERROR_ELF_SIZE_MISMATCH = <Result.ERROR_ELF_SIZE_MISMATCH: 3>

ERROR_MISSING_PTX_VERSION = <Result.ERROR_MISSING_PTX_VERSION: 4>

ERROR_NULL_POINTER = <Result.ERROR_NULL_POINTER: 5>

ERROR_COMPRESSION_FAILED = <Result.ERROR_COMPRESSION_FAILED: 6>

ERROR_COMPRESSED_SIZE_EXCEEDED = <Result.ERROR_COMPRESSED_SIZE_EXCEEDED: 7>

ERROR_UNRECOGNIZED_OPTION = <Result.ERROR_UNRECOGNIZED_OPTION: 8>

ERROR_INVALID_ARCH = <Result.ERROR_INVALID_ARCH: 9>

ERROR_INVALID_NVVM = <Result.ERROR_INVALID_NVVM: 10>

ERROR_EMPTY_INPUT = <Result.ERROR_EMPTY_INPUT: 11>

ERROR_MISSING_PTX_ARCH = <Result.ERROR_MISSING_PTX_ARCH: 12>

ERROR_PTX_ARCH_MISMATCH = <Result.ERROR_PTX_ARCH_MISMATCH: 13>

ERROR_MISSING_FATBIN = <Result.ERROR_MISSING_FATBIN: 14>

ERROR_INVALID_INDEX = <Result.ERROR_INVALID_INDEX: 15>

ERROR_IDENTIFIER_REUSE = <Result.ERROR_IDENTIFIER_REUSE: 16>

ERROR_INTERNAL_PTX_OPTION = <Result.ERROR_INTERNAL_PTX_OPTION: 17>