source/cuda/bindings/nvrtc.pyx

# SPDX-FileCopyrightText: Copyright (c) 2021-2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
# SPDX-License-Identifier: LicenseRef-NVIDIA-SOFTWARE-LICENSE

# This code was automatically generated with version 12.9.0. Do not modify it directly.
from typing import Any, Optional
from enum import IntEnum
import cython
import ctypes
from libc.stdlib cimport calloc, malloc, free
from libc cimport string
from libc.stdint cimport int32_t, uint32_t, int64_t, uint64_t, uintptr_t
from libc.stddef cimport wchar_t
from libc.limits cimport CHAR_MIN
from libcpp.vector cimport vector
from cpython.buffer cimport PyObject_CheckBuffer, PyObject_GetBuffer, PyBuffer_Release, PyBUF_SIMPLE, PyBUF_ANY_CONTIGUOUS
from cpython.bytes cimport PyBytes_FromStringAndSize

import cuda.bindings.driver as _driver
_driver = _driver.__dict__
include "_lib/utils.pxi"

ctypedef unsigned long long signed_char_ptr
ctypedef unsigned long long unsigned_char_ptr
ctypedef unsigned long long char_ptr
ctypedef unsigned long long short_ptr
ctypedef unsigned long long unsigned_short_ptr
ctypedef unsigned long long int_ptr
ctypedef unsigned long long long_int_ptr
ctypedef unsigned long long long_long_int_ptr
ctypedef unsigned long long unsigned_int_ptr
ctypedef unsigned long long unsigned_long_int_ptr
ctypedef unsigned long long unsigned_long_long_int_ptr
ctypedef unsigned long long uint32_t_ptr
ctypedef unsigned long long uint64_t_ptr
ctypedef unsigned long long int32_t_ptr
ctypedef unsigned long long int64_t_ptr
ctypedef unsigned long long unsigned_ptr
ctypedef unsigned long long unsigned_long_long_ptr
ctypedef unsigned long long long_long_ptr
ctypedef unsigned long long size_t_ptr
ctypedef unsigned long long long_ptr
ctypedef unsigned long long float_ptr
ctypedef unsigned long long double_ptr
ctypedef unsigned long long void_ptr


class nvrtcResult(IntEnum):
    """
    The enumerated type nvrtcResult defines API call result codes.
    NVRTC API functions return nvrtcResult to indicate the call result.
    """
    NVRTC_SUCCESS = cynvrtc.nvrtcResult.NVRTC_SUCCESS
    NVRTC_ERROR_OUT_OF_MEMORY = cynvrtc.nvrtcResult.NVRTC_ERROR_OUT_OF_MEMORY
    NVRTC_ERROR_PROGRAM_CREATION_FAILURE = cynvrtc.nvrtcResult.NVRTC_ERROR_PROGRAM_CREATION_FAILURE
    NVRTC_ERROR_INVALID_INPUT = cynvrtc.nvrtcResult.NVRTC_ERROR_INVALID_INPUT
    NVRTC_ERROR_INVALID_PROGRAM = cynvrtc.nvrtcResult.NVRTC_ERROR_INVALID_PROGRAM
    NVRTC_ERROR_INVALID_OPTION = cynvrtc.nvrtcResult.NVRTC_ERROR_INVALID_OPTION
    NVRTC_ERROR_COMPILATION = cynvrtc.nvrtcResult.NVRTC_ERROR_COMPILATION
    NVRTC_ERROR_BUILTIN_OPERATION_FAILURE = cynvrtc.nvrtcResult.NVRTC_ERROR_BUILTIN_OPERATION_FAILURE
    NVRTC_ERROR_NO_NAME_EXPRESSIONS_AFTER_COMPILATION = cynvrtc.nvrtcResult.NVRTC_ERROR_NO_NAME_EXPRESSIONS_AFTER_COMPILATION
    NVRTC_ERROR_NO_LOWERED_NAMES_BEFORE_COMPILATION = cynvrtc.nvrtcResult.NVRTC_ERROR_NO_LOWERED_NAMES_BEFORE_COMPILATION
    NVRTC_ERROR_NAME_EXPRESSION_NOT_VALID = cynvrtc.nvrtcResult.NVRTC_ERROR_NAME_EXPRESSION_NOT_VALID
    NVRTC_ERROR_INTERNAL_ERROR = cynvrtc.nvrtcResult.NVRTC_ERROR_INTERNAL_ERROR
    NVRTC_ERROR_TIME_FILE_WRITE_FAILED = cynvrtc.nvrtcResult.NVRTC_ERROR_TIME_FILE_WRITE_FAILED
    NVRTC_ERROR_NO_PCH_CREATE_ATTEMPTED = cynvrtc.nvrtcResult.NVRTC_ERROR_NO_PCH_CREATE_ATTEMPTED
    NVRTC_ERROR_PCH_CREATE_HEAP_EXHAUSTED = cynvrtc.nvrtcResult.NVRTC_ERROR_PCH_CREATE_HEAP_EXHAUSTED
    NVRTC_ERROR_PCH_CREATE = cynvrtc.nvrtcResult.NVRTC_ERROR_PCH_CREATE
    NVRTC_ERROR_CANCELLED = cynvrtc.nvrtcResult.NVRTC_ERROR_CANCELLED

_dict_nvrtcResult = dict(((int(v), v) for k, v in nvrtcResult.__members__.items()))

cdef class nvrtcProgram:
    """ nvrtcProgram is the unit of compilation, and an opaque handle for a program.

    To compile a CUDA program string, an instance of nvrtcProgram must be created first with nvrtcCreateProgram, then compiled with nvrtcCompileProgram.

    Methods
    -------
    getPtr()
        Get memory address of class instance

    """
    def __cinit__(self, void_ptr init_value = 0, void_ptr _ptr = 0):
        if _ptr == 0:
            self._pvt_ptr = &self._pvt_val
            self._pvt_ptr[0] = <cynvrtc.nvrtcProgram>init_value
        else:
            self._pvt_ptr = <cynvrtc.nvrtcProgram *>_ptr
    def __init__(self, *args, **kwargs):
        pass
    def __repr__(self):
        return '<nvrtcProgram ' + str(hex(self.__int__())) + '>'
    def __index__(self):
        return self.__int__()
    def __eq__(self, other):
        if not isinstance(other, nvrtcProgram):
            return False
        return self._pvt_ptr[0] == (<nvrtcProgram>other)._pvt_ptr[0]
    def __hash__(self):
        return hash(<uintptr_t><void*>(self._pvt_ptr[0]))
    def __int__(self):
        return <void_ptr>self._pvt_ptr[0]
    def getPtr(self):
        return <void_ptr>self._pvt_ptr

@cython.embedsignature(True)
def nvrtcGetErrorString(result not None : nvrtcResult):
    """ nvrtcGetErrorString is a helper function that returns a string describing the given nvrtcResult code, e.g., NVRTC_SUCCESS to `"NVRTC_SUCCESS"`. For unrecognized enumeration values, it returns `"NVRTC_ERROR unknown"`.

    Parameters
    ----------
    result : :py:obj:`~.nvrtcResult`
        CUDA Runtime Compilation API result code.

    Returns
    -------
    nvrtcResult.NVRTC_SUCCESS
        nvrtcResult.NVRTC_SUCCESS
    bytes
        Message string for the given :py:obj:`~.nvrtcResult` code.
    """
    cdef cynvrtc.nvrtcResult cyresult = result.value
    with nogil:
        err = cynvrtc.nvrtcGetErrorString(cyresult)
    return (nvrtcResult.NVRTC_SUCCESS, err)

@cython.embedsignature(True)
def nvrtcVersion():
    """ nvrtcVersion sets the output parameters `major` and `minor` with the CUDA Runtime Compilation version number.

    Returns
    -------
    nvrtcResult
        - :py:obj:`~.NVRTC_SUCCESS`
        - :py:obj:`~.NVRTC_ERROR_INVALID_INPUT`
    major : int
        CUDA Runtime Compilation major version number.
    minor : int
        CUDA Runtime Compilation minor version number.
    """
    cdef int major = 0
    cdef int minor = 0
    with nogil:
        err = cynvrtc.nvrtcVersion(&major, &minor)
    if err != cynvrtc.NVRTC_SUCCESS:
        return (_dict_nvrtcResult[err], None, None)
    return (_dict_nvrtcResult[err], major, minor)

@cython.embedsignature(True)
def nvrtcGetNumSupportedArchs():
    """ nvrtcGetNumSupportedArchs sets the output parameter `numArchs` with the number of architectures supported by NVRTC. This can then be used to pass an array to :py:obj:`~.nvrtcGetSupportedArchs` to get the supported architectures.

    see :py:obj:`~.nvrtcGetSupportedArchs`

    Returns
    -------
    nvrtcResult
        - :py:obj:`~.NVRTC_SUCCESS`
        - :py:obj:`~.NVRTC_ERROR_INVALID_INPUT`
    numArchs : int
        number of supported architectures.
    """
    cdef int numArchs = 0
    with nogil:
        err = cynvrtc.nvrtcGetNumSupportedArchs(&numArchs)
    if err != cynvrtc.NVRTC_SUCCESS:
        return (_dict_nvrtcResult[err], None)
    return (_dict_nvrtcResult[err], numArchs)

@cython.embedsignature(True)
def nvrtcGetSupportedArchs():
    """ nvrtcGetSupportedArchs populates the array passed via the output parameter `supportedArchs` with the architectures supported by NVRTC. The array is sorted in the ascending order. The size of the array to be passed can be determined using :py:obj:`~.nvrtcGetNumSupportedArchs`.

    see :py:obj:`~.nvrtcGetNumSupportedArchs`

    Returns
    -------
    nvrtcResult
        - :py:obj:`~.NVRTC_SUCCESS`
        - :py:obj:`~.NVRTC_ERROR_INVALID_INPUT`
    supportedArchs : list[int]
        sorted array of supported architectures.
    """
    cdef vector[int] supportedArchs
    _, s = nvrtcGetNumSupportedArchs()
    supportedArchs.resize(s)

    with nogil:
        err = cynvrtc.nvrtcGetSupportedArchs(supportedArchs.data())
    if err != cynvrtc.NVRTC_SUCCESS:
        return (_dict_nvrtcResult[err], None)
    return (_dict_nvrtcResult[err], supportedArchs)

@cython.embedsignature(True)
def nvrtcCreateProgram(char* src, char* name, int numHeaders, headers : Optional[tuple[bytes] | list[bytes]], includeNames : Optional[tuple[bytes] | list[bytes]]):
    """ nvrtcCreateProgram creates an instance of nvrtcProgram with the given input parameters, and sets the output parameter `prog` with it.

    Parameters
    ----------
    src : bytes
        CUDA program source.
    name : bytes
        CUDA program name.  `name` can be `NULL`; `"default_program"` is
        used when `name` is `NULL` or "".
    numHeaders : int
        Number of headers used.  `numHeaders` must be greater than or equal
        to 0.
    headers : list[bytes]
        Sources of the headers.  `headers` can be `NULL` when `numHeaders`
        is 0.
    includeNames : list[bytes]
        Name of each header by which they can be included in the CUDA
        program source.  `includeNames` can be `NULL` when `numHeaders` is
        0. These headers must be included with the exact names specified
        here.

    Returns
    -------
    nvrtcResult
        - :py:obj:`~.NVRTC_SUCCESS`
        - :py:obj:`~.NVRTC_ERROR_OUT_OF_MEMORY`
        - :py:obj:`~.NVRTC_ERROR_PROGRAM_CREATION_FAILURE`
        - :py:obj:`~.NVRTC_ERROR_INVALID_INPUT`
        - :py:obj:`~.NVRTC_ERROR_INVALID_PROGRAM`
    prog : :py:obj:`~.nvrtcProgram`
        CUDA Runtime Compilation program.

    See Also
    --------
    :py:obj:`~.nvrtcDestroyProgram`
    """
    includeNames = [] if includeNames is None else includeNames
    if not all(isinstance(_x, (bytes)) for _x in includeNames):
        raise TypeError("Argument 'includeNames' is not instance of type (expected tuple[bytes] or list[bytes]")
    headers = [] if headers is None else headers
    if not all(isinstance(_x, (bytes)) for _x in headers):
        raise TypeError("Argument 'headers' is not instance of type (expected tuple[bytes] or list[bytes]")
    cdef nvrtcProgram prog = nvrtcProgram()
    if numHeaders > len(headers): raise RuntimeError("List is too small: " + str(len(headers)) + " < " + str(numHeaders))
    if numHeaders > len(includeNames): raise RuntimeError("List is too small: " + str(len(includeNames)) + " < " + str(numHeaders))
    cdef vector[const char*] cyheaders = headers
    cdef vector[const char*] cyincludeNames = includeNames
    with nogil:
        err = cynvrtc.nvrtcCreateProgram(<cynvrtc.nvrtcProgram*>prog._pvt_ptr, src, name, numHeaders, cyheaders.data(), cyincludeNames.data())
    if err != cynvrtc.NVRTC_SUCCESS:
        return (_dict_nvrtcResult[err], None)
    return (_dict_nvrtcResult[err], prog)

@cython.embedsignature(True)
def nvrtcDestroyProgram(prog):
    """ nvrtcDestroyProgram destroys the given program.

    Parameters
    ----------
    prog : :py:obj:`~.nvrtcProgram`
        CUDA Runtime Compilation program.

    Returns
    -------
    nvrtcResult
        - :py:obj:`~.NVRTC_SUCCESS`
        - :py:obj:`~.NVRTC_ERROR_INVALID_PROGRAM`

    See Also
    --------
    :py:obj:`~.nvrtcCreateProgram`
    """
    cdef cynvrtc.nvrtcProgram *cyprog
    if prog is None:
        cyprog = <cynvrtc.nvrtcProgram*><void_ptr>NULL
    elif isinstance(prog, (nvrtcProgram,)):
        pprog = prog.getPtr()
        cyprog = <cynvrtc.nvrtcProgram*><void_ptr>pprog
    elif isinstance(prog, (int)):
        cyprog = <cynvrtc.nvrtcProgram*><void_ptr>prog
    else:
        raise TypeError("Argument 'prog' is not instance of type (expected <class 'int, nvrtc.nvrtcProgram'>, found " + str(type(prog)))
    with nogil:
        err = cynvrtc.nvrtcDestroyProgram(cyprog)
    return (_dict_nvrtcResult[err],)

@cython.embedsignature(True)
def nvrtcCompileProgram(prog, int numOptions, options : Optional[tuple[bytes] | list[bytes]]):
    """ nvrtcCompileProgram compiles the given program.

    It supports compile options listed in :py:obj:`~.Supported Compile
    Options`.

    Parameters
    ----------
    prog : :py:obj:`~.nvrtcProgram`
        CUDA Runtime Compilation program.
    numOptions : int
        Number of compiler options passed.
    options : list[bytes]
        Compiler options in the form of C string array.  `options` can be
        `NULL` when `numOptions` is 0.

    Returns
    -------
    nvrtcResult
        - :py:obj:`~.NVRTC_SUCCESS`
        - :py:obj:`~.NVRTC_ERROR_OUT_OF_MEMORY`
        - :py:obj:`~.NVRTC_ERROR_INVALID_INPUT`
        - :py:obj:`~.NVRTC_ERROR_INVALID_PROGRAM`
        - :py:obj:`~.NVRTC_ERROR_INVALID_OPTION`
        - :py:obj:`~.NVRTC_ERROR_COMPILATION`
        - :py:obj:`~.NVRTC_ERROR_BUILTIN_OPERATION_FAILURE`
        - :py:obj:`~.NVRTC_ERROR_TIME_FILE_WRITE_FAILED`
        - :py:obj:`~.NVRTC_ERROR_CANCELLED`
    """
    options = [] if options is None else options
    if not all(isinstance(_x, (bytes)) for _x in options):
        raise TypeError("Argument 'options' is not instance of type (expected tuple[bytes] or list[bytes]")
    cdef cynvrtc.nvrtcProgram cyprog
    if prog is None:
        pprog = 0
    elif isinstance(prog, (nvrtcProgram,)):
        pprog = int(prog)
    else:
        pprog = int(nvrtcProgram(prog))
    cyprog = <cynvrtc.nvrtcProgram><void_ptr>pprog
    if numOptions > len(options): raise RuntimeError("List is too small: " + str(len(options)) + " < " + str(numOptions))
    cdef vector[const char*] cyoptions = options
    with nogil:
        err = cynvrtc.nvrtcCompileProgram(cyprog, numOptions, cyoptions.data())
    return (_dict_nvrtcResult[err],)

@cython.embedsignature(True)
def nvrtcGetPTXSize(prog):
    """ nvrtcGetPTXSize sets the value of `ptxSizeRet` with the size of the PTX generated by the previous compilation of `prog` (including the trailing `NULL`).

    Parameters
    ----------
    prog : :py:obj:`~.nvrtcProgram`
        CUDA Runtime Compilation program.

    Returns
    -------
    nvrtcResult
        - :py:obj:`~.NVRTC_SUCCESS`
        - :py:obj:`~.NVRTC_ERROR_INVALID_INPUT`
        - :py:obj:`~.NVRTC_ERROR_INVALID_PROGRAM`
    ptxSizeRet : int
        Size of the generated PTX (including the trailing `NULL`).

    See Also
    --------
    :py:obj:`~.nvrtcGetPTX`
    """
    cdef cynvrtc.nvrtcProgram cyprog
    if prog is None:
        pprog = 0
    elif isinstance(prog, (nvrtcProgram,)):
        pprog = int(prog)
    else:
        pprog = int(nvrtcProgram(prog))
    cyprog = <cynvrtc.nvrtcProgram><void_ptr>pprog
    cdef size_t ptxSizeRet = 0
    with nogil:
        err = cynvrtc.nvrtcGetPTXSize(cyprog, &ptxSizeRet)
    if err != cynvrtc.NVRTC_SUCCESS:
        return (_dict_nvrtcResult[err], None)
    return (_dict_nvrtcResult[err], ptxSizeRet)

@cython.embedsignature(True)
def nvrtcGetPTX(prog, char* ptx):
    """ nvrtcGetPTX stores the PTX generated by the previous compilation of `prog` in the memory pointed by `ptx`.

    Parameters
    ----------
    prog : :py:obj:`~.nvrtcProgram`
        CUDA Runtime Compilation program.
    ptx : bytes
        Compiled result.

    Returns
    -------
    nvrtcResult
        - :py:obj:`~.NVRTC_SUCCESS`
        - :py:obj:`~.NVRTC_ERROR_INVALID_INPUT`
        - :py:obj:`~.NVRTC_ERROR_INVALID_PROGRAM`

    See Also
    --------
    :py:obj:`~.nvrtcGetPTXSize`
    """
    cdef cynvrtc.nvrtcProgram cyprog
    if prog is None:
        pprog = 0
    elif isinstance(prog, (nvrtcProgram,)):
        pprog = int(prog)
    else:
        pprog = int(nvrtcProgram(prog))
    cyprog = <cynvrtc.nvrtcProgram><void_ptr>pprog
    with nogil:
        err = cynvrtc.nvrtcGetPTX(cyprog, ptx)
    return (_dict_nvrtcResult[err],)

@cython.embedsignature(True)
def nvrtcGetCUBINSize(prog):
    """ nvrtcGetCUBINSize sets the value of `cubinSizeRet` with the size of the cubin generated by the previous compilation of `prog`. The value of cubinSizeRet is set to 0 if the value specified to `-arch` is a virtual architecture instead of an actual architecture.

    Parameters
    ----------
    prog : :py:obj:`~.nvrtcProgram`
        CUDA Runtime Compilation program.

    Returns
    -------
    nvrtcResult
        - :py:obj:`~.NVRTC_SUCCESS`
        - :py:obj:`~.NVRTC_ERROR_INVALID_INPUT`
        - :py:obj:`~.NVRTC_ERROR_INVALID_PROGRAM`
    cubinSizeRet : int
        Size of the generated cubin.

    See Also
    --------
    :py:obj:`~.nvrtcGetCUBIN`
    """
    cdef cynvrtc.nvrtcProgram cyprog
    if prog is None:
        pprog = 0
    elif isinstance(prog, (nvrtcProgram,)):
        pprog = int(prog)
    else:
        pprog = int(nvrtcProgram(prog))
    cyprog = <cynvrtc.nvrtcProgram><void_ptr>pprog
    cdef size_t cubinSizeRet = 0
    with nogil:
        err = cynvrtc.nvrtcGetCUBINSize(cyprog, &cubinSizeRet)
    if err != cynvrtc.NVRTC_SUCCESS:
        return (_dict_nvrtcResult[err], None)
    return (_dict_nvrtcResult[err], cubinSizeRet)

@cython.embedsignature(True)
def nvrtcGetCUBIN(prog, char* cubin):
    """ nvrtcGetCUBIN stores the cubin generated by the previous compilation of `prog` in the memory pointed by `cubin`. No cubin is available if the value specified to `-arch` is a virtual architecture instead of an actual architecture.

    Parameters
    ----------
    prog : :py:obj:`~.nvrtcProgram`
        CUDA Runtime Compilation program.
    cubin : bytes
        Compiled and assembled result.

    Returns
    -------
    nvrtcResult
        - :py:obj:`~.NVRTC_SUCCESS`
        - :py:obj:`~.NVRTC_ERROR_INVALID_INPUT`
        - :py:obj:`~.NVRTC_ERROR_INVALID_PROGRAM`

    See Also
    --------
    :py:obj:`~.nvrtcGetCUBINSize`
    """
    cdef cynvrtc.nvrtcProgram cyprog
    if prog is None:
        pprog = 0
    elif isinstance(prog, (nvrtcProgram,)):
        pprog = int(prog)
    else:
        pprog = int(nvrtcProgram(prog))
    cyprog = <cynvrtc.nvrtcProgram><void_ptr>pprog
    with nogil:
        err = cynvrtc.nvrtcGetCUBIN(cyprog, cubin)
    return (_dict_nvrtcResult[err],)

@cython.embedsignature(True)
def nvrtcGetNVVMSize(prog):
    """ DEPRECATION NOTICE: This function will be removed in a future release. Please use nvrtcGetLTOIRSize (and nvrtcGetLTOIR) instead.

    Parameters
    ----------
    prog : :py:obj:`~.nvrtcProgram`
        None

    Returns
    -------
    nvrtcResult

    nvvmSizeRet : int
        None
    """
    cdef cynvrtc.nvrtcProgram cyprog
    if prog is None:
        pprog = 0
    elif isinstance(prog, (nvrtcProgram,)):
        pprog = int(prog)
    else:
        pprog = int(nvrtcProgram(prog))
    cyprog = <cynvrtc.nvrtcProgram><void_ptr>pprog
    cdef size_t nvvmSizeRet = 0
    with nogil:
        err = cynvrtc.nvrtcGetNVVMSize(cyprog, &nvvmSizeRet)
    if err != cynvrtc.NVRTC_SUCCESS:
        return (_dict_nvrtcResult[err], None)
    return (_dict_nvrtcResult[err], nvvmSizeRet)

@cython.embedsignature(True)
def nvrtcGetNVVM(prog, char* nvvm):
    """ DEPRECATION NOTICE: This function will be removed in a future release. Please use nvrtcGetLTOIR (and nvrtcGetLTOIRSize) instead.

    Parameters
    ----------
    prog : :py:obj:`~.nvrtcProgram`
        None
    nvvm : bytes
        None

    Returns
    -------
    nvrtcResult

    """
    cdef cynvrtc.nvrtcProgram cyprog
    if prog is None:
        pprog = 0
    elif isinstance(prog, (nvrtcProgram,)):
        pprog = int(prog)
    else:
        pprog = int(nvrtcProgram(prog))
    cyprog = <cynvrtc.nvrtcProgram><void_ptr>pprog
    with nogil:
        err = cynvrtc.nvrtcGetNVVM(cyprog, nvvm)
    return (_dict_nvrtcResult[err],)

@cython.embedsignature(True)
def nvrtcGetLTOIRSize(prog):
    """ nvrtcGetLTOIRSize sets the value of `LTOIRSizeRet` with the size of the LTO IR generated by the previous compilation of `prog`. The value of LTOIRSizeRet is set to 0 if the program was not compiled with `-dlto`.

    Parameters
    ----------
    prog : :py:obj:`~.nvrtcProgram`
        CUDA Runtime Compilation program.

    Returns
    -------
    nvrtcResult
        - :py:obj:`~.NVRTC_SUCCESS`
        - :py:obj:`~.NVRTC_ERROR_INVALID_INPUT`
        - :py:obj:`~.NVRTC_ERROR_INVALID_PROGRAM`
    LTOIRSizeRet : int
        Size of the generated LTO IR.

    See Also
    --------
    :py:obj:`~.nvrtcGetLTOIR`
    """
    cdef cynvrtc.nvrtcProgram cyprog
    if prog is None:
        pprog = 0
    elif isinstance(prog, (nvrtcProgram,)):
        pprog = int(prog)
    else:
        pprog = int(nvrtcProgram(prog))
    cyprog = <cynvrtc.nvrtcProgram><void_ptr>pprog
    cdef size_t LTOIRSizeRet = 0
    with nogil:
        err = cynvrtc.nvrtcGetLTOIRSize(cyprog, &LTOIRSizeRet)
    if err != cynvrtc.NVRTC_SUCCESS:
        return (_dict_nvrtcResult[err], None)
    return (_dict_nvrtcResult[err], LTOIRSizeRet)

@cython.embedsignature(True)
def nvrtcGetLTOIR(prog, char* LTOIR):
    """ nvrtcGetLTOIR stores the LTO IR generated by the previous compilation of `prog` in the memory pointed by `LTOIR`. No LTO IR is available if the program was compiled without `-dlto`.

    Parameters
    ----------
    prog : :py:obj:`~.nvrtcProgram`
        CUDA Runtime Compilation program.
    LTOIR : bytes
        Compiled result.

    Returns
    -------
    nvrtcResult
        - :py:obj:`~.NVRTC_SUCCESS`
        - :py:obj:`~.NVRTC_ERROR_INVALID_INPUT`
        - :py:obj:`~.NVRTC_ERROR_INVALID_PROGRAM`

    See Also
    --------
    :py:obj:`~.nvrtcGetLTOIRSize`
    """
    cdef cynvrtc.nvrtcProgram cyprog
    if prog is None:
        pprog = 0
    elif isinstance(prog, (nvrtcProgram,)):
        pprog = int(prog)
    else:
        pprog = int(nvrtcProgram(prog))
    cyprog = <cynvrtc.nvrtcProgram><void_ptr>pprog
    with nogil:
        err = cynvrtc.nvrtcGetLTOIR(cyprog, LTOIR)
    return (_dict_nvrtcResult[err],)

@cython.embedsignature(True)
def nvrtcGetOptiXIRSize(prog):
    """ nvrtcGetOptiXIRSize sets the value of `optixirSizeRet` with the size of the OptiX IR generated by the previous compilation of `prog`. The value of nvrtcGetOptiXIRSize is set to 0 if the program was compiled with options incompatible with OptiX IR generation.

    Parameters
    ----------
    prog : :py:obj:`~.nvrtcProgram`
        CUDA Runtime Compilation program.

    Returns
    -------
    nvrtcResult
        - :py:obj:`~.NVRTC_SUCCESS`
        - :py:obj:`~.NVRTC_ERROR_INVALID_INPUT`
        - :py:obj:`~.NVRTC_ERROR_INVALID_PROGRAM`
    optixirSizeRet : int
        Size of the generated LTO IR.

    See Also
    --------
    :py:obj:`~.nvrtcGetOptiXIR`
    """
    cdef cynvrtc.nvrtcProgram cyprog
    if prog is None:
        pprog = 0
    elif isinstance(prog, (nvrtcProgram,)):
        pprog = int(prog)
    else:
        pprog = int(nvrtcProgram(prog))
    cyprog = <cynvrtc.nvrtcProgram><void_ptr>pprog
    cdef size_t optixirSizeRet = 0
    with nogil:
        err = cynvrtc.nvrtcGetOptiXIRSize(cyprog, &optixirSizeRet)
    if err != cynvrtc.NVRTC_SUCCESS:
        return (_dict_nvrtcResult[err], None)
    return (_dict_nvrtcResult[err], optixirSizeRet)

@cython.embedsignature(True)
def nvrtcGetOptiXIR(prog, char* optixir):
    """ nvrtcGetOptiXIR stores the OptiX IR generated by the previous compilation of `prog` in the memory pointed by `optixir`. No OptiX IR is available if the program was compiled with options incompatible with OptiX IR generation.

    Parameters
    ----------
    prog : :py:obj:`~.nvrtcProgram`
        CUDA Runtime Compilation program.
    optixir : bytes
        Optix IR Compiled result.

    Returns
    -------
    nvrtcResult
        - :py:obj:`~.NVRTC_SUCCESS`
        - :py:obj:`~.NVRTC_ERROR_INVALID_INPUT`
        - :py:obj:`~.NVRTC_ERROR_INVALID_PROGRAM`

    See Also
    --------
    :py:obj:`~.nvrtcGetOptiXIRSize`
    """
    cdef cynvrtc.nvrtcProgram cyprog
    if prog is None:
        pprog = 0
    elif isinstance(prog, (nvrtcProgram,)):
        pprog = int(prog)
    else:
        pprog = int(nvrtcProgram(prog))
    cyprog = <cynvrtc.nvrtcProgram><void_ptr>pprog
    with nogil:
        err = cynvrtc.nvrtcGetOptiXIR(cyprog, optixir)
    return (_dict_nvrtcResult[err],)

@cython.embedsignature(True)
def nvrtcGetProgramLogSize(prog):
    """ nvrtcGetProgramLogSize sets `logSizeRet` with the size of the log generated by the previous compilation of `prog` (including the trailing `NULL`).

    Note that compilation log may be generated with warnings and
    informative messages, even when the compilation of `prog` succeeds.

    Parameters
    ----------
    prog : :py:obj:`~.nvrtcProgram`
        CUDA Runtime Compilation program.

    Returns
    -------
    nvrtcResult
        - :py:obj:`~.NVRTC_SUCCESS`
        - :py:obj:`~.NVRTC_ERROR_INVALID_INPUT`
        - :py:obj:`~.NVRTC_ERROR_INVALID_PROGRAM`
    logSizeRet : int
        Size of the compilation log (including the trailing `NULL`).

    See Also
    --------
    :py:obj:`~.nvrtcGetProgramLog`
    """
    cdef cynvrtc.nvrtcProgram cyprog
    if prog is None:
        pprog = 0
    elif isinstance(prog, (nvrtcProgram,)):
        pprog = int(prog)
    else:
        pprog = int(nvrtcProgram(prog))
    cyprog = <cynvrtc.nvrtcProgram><void_ptr>pprog
    cdef size_t logSizeRet = 0
    with nogil:
        err = cynvrtc.nvrtcGetProgramLogSize(cyprog, &logSizeRet)
    if err != cynvrtc.NVRTC_SUCCESS:
        return (_dict_nvrtcResult[err], None)
    return (_dict_nvrtcResult[err], logSizeRet)

@cython.embedsignature(True)
def nvrtcGetProgramLog(prog, char* log):
    """ nvrtcGetProgramLog stores the log generated by the previous compilation of `prog` in the memory pointed by `log`.

    Parameters
    ----------
    prog : :py:obj:`~.nvrtcProgram`
        CUDA Runtime Compilation program.
    log : bytes
        Compilation log.

    Returns
    -------
    nvrtcResult
        - :py:obj:`~.NVRTC_SUCCESS`
        - :py:obj:`~.NVRTC_ERROR_INVALID_INPUT`
        - :py:obj:`~.NVRTC_ERROR_INVALID_PROGRAM`

    See Also
    --------
    :py:obj:`~.nvrtcGetProgramLogSize`
    """
    cdef cynvrtc.nvrtcProgram cyprog
    if prog is None:
        pprog = 0
    elif isinstance(prog, (nvrtcProgram,)):
        pprog = int(prog)
    else:
        pprog = int(nvrtcProgram(prog))
    cyprog = <cynvrtc.nvrtcProgram><void_ptr>pprog
    with nogil:
        err = cynvrtc.nvrtcGetProgramLog(cyprog, log)
    return (_dict_nvrtcResult[err],)

@cython.embedsignature(True)
def nvrtcAddNameExpression(prog, char* name_expression):
    """ nvrtcAddNameExpression notes the given name expression denoting the address of a global function or device/__constant__ variable.

    The identical name expression string must be provided on a subsequent
    call to nvrtcGetLoweredName to extract the lowered name.

    Parameters
    ----------
    prog : :py:obj:`~.nvrtcProgram`
        CUDA Runtime Compilation program.
    name_expression : bytes
        constant expression denoting the address of a global function or
        device/__constant__ variable.

    Returns
    -------
    nvrtcResult
        - :py:obj:`~.NVRTC_SUCCESS`
        - :py:obj:`~.NVRTC_ERROR_INVALID_PROGRAM`
        - :py:obj:`~.NVRTC_ERROR_INVALID_INPUT`
        - :py:obj:`~.NVRTC_ERROR_NO_NAME_EXPRESSIONS_AFTER_COMPILATION`

    See Also
    --------
    :py:obj:`~.nvrtcGetLoweredName`
    """
    cdef cynvrtc.nvrtcProgram cyprog
    if prog is None:
        pprog = 0
    elif isinstance(prog, (nvrtcProgram,)):
        pprog = int(prog)
    else:
        pprog = int(nvrtcProgram(prog))
    cyprog = <cynvrtc.nvrtcProgram><void_ptr>pprog
    with nogil:
        err = cynvrtc.nvrtcAddNameExpression(cyprog, name_expression)
    return (_dict_nvrtcResult[err],)

@cython.embedsignature(True)
def nvrtcGetLoweredName(prog, char* name_expression):
    """ nvrtcGetLoweredName extracts the lowered (mangled) name for a global function or device/__constant__ variable, and updates lowered_name to point to it. The memory containing the name is released when the NVRTC program is destroyed by nvrtcDestroyProgram. The identical name expression must have been previously provided to nvrtcAddNameExpression.

    Parameters
    ----------
    prog : nvrtcProgram
        CUDA Runtime Compilation program.
    name_expression : bytes
        constant expression denoting the address of a global function or
        device/__constant__ variable.

    Returns
    -------
    nvrtcResult
        NVRTC_SUCCESS
        NVRTC_ERROR_NO_LOWERED_NAMES_BEFORE_COMPILATION
        NVRTC_ERROR_NAME_EXPRESSION_NOT_VALID
    lowered_name : bytes
        initialized by the function to point to a C string containing the
        lowered (mangled) name corresponding to the provided name
        expression.

    See Also
    --------
    nvrtcAddNameExpression
    """
    cdef cynvrtc.nvrtcProgram cyprog
    if prog is None:
        pprog = 0
    elif isinstance(prog, (nvrtcProgram,)):
        pprog = int(prog)
    else:
        pprog = int(nvrtcProgram(prog))
    cyprog = <cynvrtc.nvrtcProgram><void_ptr>pprog
    cdef const char* lowered_name = NULL
    with nogil:
        err = cynvrtc.nvrtcGetLoweredName(cyprog, name_expression, &lowered_name)
    if err != cynvrtc.NVRTC_SUCCESS:
        return (_dict_nvrtcResult[err], None)
    return (_dict_nvrtcResult[err], <bytes>lowered_name if lowered_name != NULL else None)

@cython.embedsignature(True)
def nvrtcGetPCHHeapSize():
    """ retrieve the current size of the PCH Heap.

    Returns
    -------
    nvrtcResult
        - :py:obj:`~.NVRTC_SUCCESS`
        - :py:obj:`~.NVRTC_ERROR_INVALID_INPUT`
    ret : int
        pointer to location where the size of the PCH Heap will be stored
    """
    cdef size_t ret = 0
    with nogil:
        err = cynvrtc.nvrtcGetPCHHeapSize(&ret)
    if err != cynvrtc.NVRTC_SUCCESS:
        return (_dict_nvrtcResult[err], None)
    return (_dict_nvrtcResult[err], ret)

@cython.embedsignature(True)
def nvrtcSetPCHHeapSize(size_t size):
    """ set the size of the PCH Heap.

    The requested size may be rounded up to a platform dependent alignment
    (e.g. page size). If the PCH Heap has already been allocated, the heap
    memory will be freed and a new PCH Heap will be allocated.

    Parameters
    ----------
    size : size_t
        requested size of the PCH Heap, in bytes

    Returns
    -------
    nvrtcResult
        - :py:obj:`~.NVRTC_SUCCESS`
    """
    with nogil:
        err = cynvrtc.nvrtcSetPCHHeapSize(size)
    return (_dict_nvrtcResult[err],)

@cython.embedsignature(True)
def nvrtcGetPCHCreateStatus(prog):
    """ returns the PCH creation status.

    NVRTC_SUCCESS indicates that the PCH was successfully created.
    NVRTC_ERROR_NO_PCH_CREATE_ATTEMPTED indicates that no PCH creation was
    attempted, either because PCH functionality was not requested during
    the preceding nvrtcCompileProgram call, or automatic PCH processing was
    requested, and compiler chose not to create a PCH file.
    NVRTC_ERROR_PCH_CREATE_HEAP_EXHAUSTED indicates that a PCH file could
    potentially have been created, but the compiler ran out space in the
    PCH heap. In this scenario, the
    :py:obj:`~.nvrtcGetPCHHeapSizeRequired()` can be used to query the
    required heap size, the heap can be reallocated for this size with
    :py:obj:`~.nvrtcSetPCHHeapSize()` and PCH creation may be reattempted
    again invoking :py:obj:`~.nvrtcCompileProgram()` with a new NVRTC
    program instance. NVRTC_ERROR_PCH_CREATE indicates that an error
    condition prevented the PCH file from being created.

    Parameters
    ----------
    prog : :py:obj:`~.nvrtcProgram`
        CUDA Runtime Compilation program.

    Returns
    -------
    nvrtcResult
        - :py:obj:`~.NVRTC_SUCCESS`
        - :py:obj:`~.NVRTC_ERROR_NO_PCH_CREATE_ATTEMPTED`
        - :py:obj:`~.NVRTC_ERROR_PCH_CREATE`
        - :py:obj:`~.NVRTC_ERROR_PCH_CREATE_HEAP_EXHAUSTED`
        - :py:obj:`~.NVRTC_ERROR_INVALID_PROGRAM`
    """
    cdef cynvrtc.nvrtcProgram cyprog
    if prog is None:
        pprog = 0
    elif isinstance(prog, (nvrtcProgram,)):
        pprog = int(prog)
    else:
        pprog = int(nvrtcProgram(prog))
    cyprog = <cynvrtc.nvrtcProgram><void_ptr>pprog
    with nogil:
        err = cynvrtc.nvrtcGetPCHCreateStatus(cyprog)
    return (_dict_nvrtcResult[err],)

@cython.embedsignature(True)
def nvrtcGetPCHHeapSizeRequired(prog):
    """ retrieve the required size of the PCH heap required to compile the given program.

    Parameters
    ----------
    prog : :py:obj:`~.nvrtcProgram`
        CUDA Runtime Compilation program.

    Returns
    -------
    nvrtcResult
        - :py:obj:`~.NVRTC_SUCCESS`
        - :py:obj:`~.NVRTC_ERROR_INVALID_PROGRAM`
        - :py:obj:`~.NVRTC_ERROR_INVALID_INPUT` The size retrieved using this function is only valid if :py:obj:`~.nvrtcGetPCHCreateStatus()` returned NVRTC_SUCCESS or NVRTC_ERROR_PCH_CREATE_HEAP_EXHAUSTED
    size : int
        pointer to location where the required size of the PCH Heap will be
        stored
    """
    cdef cynvrtc.nvrtcProgram cyprog
    if prog is None:
        pprog = 0
    elif isinstance(prog, (nvrtcProgram,)):
        pprog = int(prog)
    else:
        pprog = int(nvrtcProgram(prog))
    cyprog = <cynvrtc.nvrtcProgram><void_ptr>pprog
    cdef size_t size = 0
    with nogil:
        err = cynvrtc.nvrtcGetPCHHeapSizeRequired(cyprog, &size)
    if err != cynvrtc.NVRTC_SUCCESS:
        return (_dict_nvrtcResult[err], None)
    return (_dict_nvrtcResult[err], size)

@cython.embedsignature(True)
def nvrtcSetFlowCallback(prog, callback, payload):
    """ nvrtcSetFlowCallback registers a callback function that the compiler will invoke at different points during a call to nvrtcCompileProgram, and the callback function can decide whether to cancel compilation by returning specific values.

    The callback function must satisfy the following constraints:

    (1) Its signature should be:

    **View CUDA Toolkit Documentation for a C++ code example**

    When invoking the callback, the compiler will always pass `payload` to
    param1 so that the callback may make decisions based on `payload` .
    It'll always pass NULL to param2 for now which is reserved for future
    extensions.

    (2) It must return 1 to cancel compilation or 0 to continue. Other
    return values are reserved for future use.

    (3) It must return consistent values. Once it returns 1 at one point,
    it must return 1 in all following invocations during the current
    nvrtcCompileProgram call in progress.

    (4) It must be thread-safe.

    (5) It must not invoke any nvrtc/libnvvm/ptx APIs.

    Parameters
    ----------
    prog : :py:obj:`~.nvrtcProgram`
        CUDA Runtime Compilation program.
    callback : Any
        the callback that issues cancellation signal.
    payload : Any
        to be passed as a parameter when invoking the callback.

    Returns
    -------
    nvrtcResult
        - :py:obj:`~.NVRTC_SUCCESS`
        - :py:obj:`~.NVRTC_ERROR_INVALID_PROGRAM`
        - :py:obj:`~.NVRTC_ERROR_INVALID_INPUT`
    """
    cdef cynvrtc.nvrtcProgram cyprog
    if prog is None:
        pprog = 0
    elif isinstance(prog, (nvrtcProgram,)):
        pprog = int(prog)
    else:
        pprog = int(nvrtcProgram(prog))
    cyprog = <cynvrtc.nvrtcProgram><void_ptr>pprog
    cycallback = _HelperInputVoidPtr(callback)
    cdef void* cycallback_ptr = <void*><void_ptr>cycallback.cptr
    cypayload = _HelperInputVoidPtr(payload)
    cdef void* cypayload_ptr = <void*><void_ptr>cypayload.cptr
    with nogil:
        err = cynvrtc.nvrtcSetFlowCallback(cyprog, cycallback_ptr, cypayload_ptr)
    return (_dict_nvrtcResult[err],)

@cython.embedsignature(True)
def sizeof(objType):
    """ Returns the size of provided CUDA Python structure in bytes

    Parameters
    ----------
    objType : Any
        CUDA Python object

    Returns
    -------
    lowered_name : int
        The size of `objType` in bytes
    """
    if objType == nvrtcProgram:
        return sizeof(cynvrtc.nvrtcProgram)
    raise TypeError("Unknown type: " + str(objType))