See More

from inspect import getfullargspec, getcallargs, isclass, getsource import os import ctypes import platform from dataclasses import dataclass as py_dataclass, is_dataclass as py_is_dataclass from goto import with_goto # TODO: this does not seem to restrict other imports __slots__ = ["i8", "i16", "i32", "i64", "u8", "u16", "u32", "u64", "f32", "f64", "c32", "c64", "CPtr", "overload", "ccall", "TypeVar", "pointer", "c_p_pointer", "Pointer", "p_c_pointer", "vectorize", "inline", "Union", "static", "with_goto", "packed", "Const", "sizeof", "ccallable", "ccallback", "Callable", "Allocatable"] # data-types type_to_convert_func = { "i8": int, "i16": int, "i32": int, "i64": int, "u8": lambda x: x, "u16": lambda x: x, "u32": lambda x: x, "u64": lambda x: x, "f32": float, "f64": float, "c32": complex, "c64": complex, "c_ptr": lambda x: x, "Const": lambda x: x, "Callable": lambda x: x, "Allocatable": lambda x: x, "Pointer": lambda x: x, } class Type: def __init__(self, name): self._name = name self._convert = type_to_convert_func[name] def __getitem__(self, params): return Array(self, params) def __call__(self, arg): return self._convert(arg) def dataclass(arg): def __class_getitem__(key): return Array(arg, key) arg.__class_getitem__ = __class_getitem__ return py_dataclass(arg) def is_ctypes_Structure(obj): return (isclass(obj) and issubclass(obj, ctypes.Structure)) def is_dataclass(obj): return ((isclass(obj) and issubclass(obj, ctypes.Structure)) or py_is_dataclass(obj)) class PointerType(Type): def __getitem__(self, type): if is_dataclass(type): return convert_to_ctypes_Structure(type) return type class ConstType(Type): def __getitem__(self, type): return type class Array: def __init__(self, type, dims): self._type = type self._dims = dims i8 = Type("i8") i16 = Type("i16") i32 = Type("i32") i64 = Type("i64") u8 = Type("u8") u16 = Type("u16") u32 = Type("u32") u64 = Type("u64") f32 = Type("f32") f64 = Type("f64") c32 = Type("c32") c64 = Type("c64") CPtr = Type("c_ptr") Const = ConstType("Const") Callable = Type("Callable") Allocatable = Type("Allocatable") Union = ctypes.Union Pointer = PointerType("Pointer") class Intent: def __init__(self, type): self._type = type def __getitem__(self, params): return params In = Intent("In") Out = Intent("Out") InOut = Intent("InOut") # Generics class TypeVar(): def __init__(self, name): self._name = name def __getitem__(self, params): return Array(self, params) def restriction(func): return func # Overloading support def ltype(x): """ Converts CPython types to LPython types """ if type(x) == int: return i32, i64 elif type(x) == float: return f32, f64 elif type(x) == complex: return c32, c64 elif type(x) == str: return (str, ) elif type(x) == bool: return (bool, ) raise Exception("Unsupported Type: %s" % str(type(x))) class OverloadedFunction: """ A wrapper class for allowing overloading. """ global_map = {} def __init__(self, func): self.func_name = func.__name__ f_list = self.global_map.get(func.__name__, []) f_list.append((func, getfullargspec(func))) self.global_map[func.__name__] = f_list def __call__(self, *args, **kwargs): func_map_list = self.global_map.get(self.func_name, False) if not func_map_list: raise Exception("Function: %s is not defined" % self.func_name) for item in func_map_list: func, key = item try: # This might fail for the cases when arguments don't match ann_dict = getcallargs(func, *args, **kwargs) except TypeError: continue flag = True for k, v in ann_dict.items(): if not key.annotations.get(k, False): flag = False break else: if not (key.annotations.get(k) in ltype(v)): flag = False break if flag: return func(*args, **kwargs) raise Exception(f"Function: {self.func_name} not found with matching " "signature") def overload(f): overloaded_f = OverloadedFunction(f) overloaded_f.__name__ = f.__name__ overloaded_f.__code__ = f.__code__ overloaded_f.__annotations__ = f.__annotations__ return overloaded_f # To be handled in ASR def vectorize(f): return f # To be handled in backend def inline(f): return f # To be handled in backend def static(f): return f class PackedDataClass: pass def packed(*args, aligned=None): if len(args) == 1: if not is_dataclass(args[0]): raise TypeError("packed can only be applied over a dataclass.") class PackedDataClassLocal(args[0], PackedDataClass): class_to_pack = args[0] return PackedDataClassLocal def _packed(f): if not is_dataclass(f): raise TypeError("packed can only be applied over a dataclass.") class PackedDataClassLocal(f, PackedDataClass): class_to_pack = f return PackedDataClassLocal return _packed def interface(f): def inner_func(): raise Exception("Unexpected to be called by CPython") return inner_func # C interoperation support class c_complex(ctypes.Structure): def __eq__(self, other): if isinstance(other, complex): return self.real == other.real and self.imag == other.imag elif isinstance(other, (int, float)): return self.real == other and self.imag == 0.0 return super().__eq__(other) def __sub__(self, other): import numpy as np if isinstance(other, (complex, np.complex64, np.complex128)): return complex(self.real - other.real, self.imag - other.imag) elif isinstance(other, (int, float)): return complex(self.real - other, self.imag) raise NotImplementedError() class c_float_complex(c_complex): _fields_ = [("real", ctypes.c_float), ("imag", ctypes.c_float)] class c_double_complex(c_complex): _fields_ = [("real", ctypes.c_double), ("imag", ctypes.c_double)] def convert_type_to_ctype(arg): from enum import Enum if arg == f64: return ctypes.c_double elif arg == f32: return ctypes.c_float elif arg == i64: return ctypes.c_int64 elif arg == i32: return ctypes.c_int32 elif arg == i16: return ctypes.c_int16 elif arg == i8: return ctypes.c_int8 elif arg == u64: return ctypes.c_uint64 elif arg == u32: return ctypes.c_uint32 elif arg == u16: return ctypes.c_uint16 elif arg == u8: return ctypes.c_uint8 elif arg == CPtr: return ctypes.c_void_p elif arg == str: return ctypes.c_char_p elif arg == c32: return c_float_complex elif arg == c64: return c_double_complex elif arg == bool: return ctypes.c_bool elif arg is None: raise NotImplementedError("Type cannot be None") elif isinstance(arg, Array): if is_dataclass(arg._type): return arg type = convert_type_to_ctype(arg._type) return ctypes.POINTER(type) elif is_dataclass(arg): return convert_to_ctypes_Structure(arg) elif issubclass(arg, Enum): # TODO: store enum in ctypes.Structure with name and value as fields. return ctypes.c_int64 else: raise NotImplementedError("Type %r not implemented" % arg) def convert_numpy_dtype_to_ctype(arg): import numpy as np if arg == np.float64: return ctypes.c_double elif arg == np.float32: return ctypes.c_float elif arg == np.int64: return ctypes.c_int64 elif arg == np.int32: return ctypes.c_int32 elif arg == np.int16: return ctypes.c_int16 elif arg == np.int8: return ctypes.c_int8 elif arg == np.uint64: return ctypes.c_uint64 elif arg == np.uint32: return ctypes.c_uint32 elif arg == np.uint16: return ctypes.c_uint16 elif arg == np.uint8: return ctypes.c_uint8 elif arg == np.void: return ctypes.c_void_p elif arg is None: raise NotImplementedError("Type cannot be None") else: raise NotImplementedError("Type %r not implemented" % arg) class CTypes: """ A wrapper class for interfacing C via ctypes. """ def __init__(self, f): def get_rtlib_dir(): current_dir = os.path.dirname(os.path.abspath(__file__)) return os.path.join(current_dir, "..") def get_lib_name(name): if platform.system() == "Linux": return "lib" + name + ".so" elif platform.system() == "Darwin": return "lib" + name + ".dylib" elif platform.system() == "Windows": return name + ".dll" else: raise NotImplementedError("Platform not implemented") def get_crtlib_path(): py_mod = os.environ.get("LPYTHON_PY_MOD_NAME", "") if py_mod == "": return os.path.join(get_rtlib_dir(), get_lib_name("lpython_runtime")) else: py_mod_path = os.environ["LPYTHON_PY_MOD_PATH"] return os.path.join(py_mod_path, get_lib_name(py_mod)) self.name = f.__name__ self.args = f.__code__.co_varnames self.annotations = f.__annotations__ if "LPYTHON_PY_MOD_NAME" in os.environ: crtlib = get_crtlib_path() self.library = ctypes.CDLL(crtlib) self.cf = self.library[self.name] else: self.cf = CTypes.emulations[self.name] argtypes = [] for arg in self.args: arg_type = self.annotations[arg] arg_ctype = convert_type_to_ctype(arg_type) argtypes.append(arg_ctype) self.cf.argtypes = argtypes if "return" in self.annotations: res_type = self.annotations["return"] if res_type is not None: self.cf.restype = convert_type_to_ctype(res_type) def __call__(self, *args, **kwargs): if len(kwargs) > 0: raise Exception("kwargs are not supported") new_args = [] for arg in args: import numpy as np if isinstance(arg, str): new_args.append(arg.encode("utf-8")) elif isinstance(arg, np.ndarray): new_args.append(arg.ctypes.data_as(ctypes.POINTER(convert_numpy_dtype_to_ctype(arg.dtype)))) else: new_args.append(arg) return self.cf(*new_args) def convert_to_ctypes_Union(f): fields = [] for name in f.__annotations__: ltype_ = f.__annotations__[name] fields.append((name, convert_type_to_ctype(ltype_))) f._fields_ = fields f.__annotations__ = {} return f def get_fixed_size_of_array(ltype_: Array): if isinstance(ltype_._dims, tuple): size = 1 for dim in ltype_._dims: if not isinstance(dim, int): return None size *= dim elif isinstance(ltype_._dims, int): return ltype_._dims return None def convert_to_ctypes_Structure(f): fields = [] pack_class = issubclass(f, PackedDataClass) if pack_class: f = f.class_to_pack if not issubclass(f, ctypes.Structure): for name in f.__annotations__: ltype_ = f.__annotations__[name] if isinstance(ltype_, Array): array_size = get_fixed_size_of_array(ltype_) if array_size is not None: ltype_ = ltype_._type fields.append((name, convert_type_to_ctype(ltype_) * array_size)) else: fields.append((name, convert_type_to_ctype(ltype_))) else: fields.append((name, convert_type_to_ctype(ltype_))) else: fields = f._fields_ pack_class = pack_class or f._pack_ class ctypes_Structure(ctypes.Structure): _pack_ = int(pack_class) _fields_ = fields def __init__(self, *args): if len(args) != 0 and len(args) != len(self._fields_): super().__init__(*args) for field, arg in zip(self._fields_, args): from enum import Enum member = self.__getattribute__(field[0]) value = arg if isinstance(member, ctypes.Array): import numpy as np if isinstance(value, np.ndarray): if value.dtype == np.complex64: value = value.flatten().tolist() value = [c_float_complex(val.real, val.imag) for val in value] elif value.dtype == np.complex128: value = value.flatten().tolist() value = [c_double_complex(val.real, val.imag) for val in value] value = type(member)(*value) elif isinstance(value, Enum): value = value.value self.__setattr__(field[0], value) ctypes_Structure.__name__ = f.__name__ return ctypes_Structure def ccall(f): if isclass(f) and issubclass(f, Union): return f return CTypes(f) def union(f): fields = [] for name in f.__annotations__: ltype_ = f.__annotations__[name] fields.append((name, convert_type_to_ctype(ltype_))) f._fields_ = fields f.__annotations__ = {} return f def pointer(x, type_=None): if type_ is None: type_ = type(x) from numpy import ndarray if isinstance(x, ndarray): return x.ctypes.data_as(ctypes.POINTER(convert_numpy_dtype_to_ctype(x.dtype))) else: if type_ == i8: return ctypes.cast(ctypes.pointer(ctypes.c_int8(x)), ctypes.c_void_p) elif type_ == i16: return ctypes.cast(ctypes.pointer(ctypes.c_int16(x)), ctypes.c_void_p) elif type_ == i32: return ctypes.cast(ctypes.pointer(ctypes.c_int32(x)), ctypes.c_void_p) elif type_ == i64: return ctypes.cast(ctypes.pointer(ctypes.c_int64(x)), ctypes.c_void_p) elif type_ == u8: return ctypes.cast(ctypes.pointer(ctypes.c_uint8(x)), ctypes.c_void_p) elif type_ == u16: return ctypes.cast(ctypes.pointer(ctypes.c_uint16(x)), ctypes.c_void_p) elif type_ == u32: return ctypes.cast(ctypes.pointer(ctypes.c_uint32(x)), ctypes.c_void_p) elif type_ == u64: return ctypes.cast(ctypes.pointer(ctypes.c_uint64(x)), ctypes.c_void_p) elif type_ == f32: return ctypes.cast(ctypes.pointer(ctypes.c_float(x)), ctypes.c_void_p) elif type_ == f64: return ctypes.cast(ctypes.pointer(ctypes.c_double(x)), ctypes.c_void_p) elif is_dataclass(type_): if issubclass(type_, ctypes.Structure): return ctypes.cast(ctypes.pointer(x), ctypes.c_void_p) else: return x else: raise Exception("Type not supported in pointer()") class PointerToStruct: def __init__(self, ctypes_ptr_): self.__dict__["ctypes_ptr"] = ctypes_ptr_ def __getattr__(self, name: str): if name == "ctypes_ptr": return self.__dict__[name] value = self.ctypes_ptr.contents.__getattribute__(name) if isinstance(value, (c_float_complex, c_double_complex)): value = complex(value.real, value.imag) return value def __setattr__(self, name: str, value): name_ = self.ctypes_ptr.contents.__getattribute__(name) from enum import Enum if isinstance(name_, c_float_complex): if isinstance(value, complex): value = c_float_complex(value.real, value.imag) else: value = c_float_complex(value.real, 0.0) elif isinstance(name_, c_double_complex): if isinstance(value, complex): value = c_double_complex(value.real, value.imag) else: value = c_double_complex(value.real, 0.0) elif isinstance(name_, ctypes.Array): import numpy as np if isinstance(value, np.ndarray): if value.dtype == np.complex64: value = value.flatten().tolist() value = [c_float_complex(val.real, val.imag) for val in value] elif value.dtype == np.complex128: value = value.flatten().tolist() value = [c_double_complex(val.real, val.imag) for val in value] value = type(name_)(*value) elif isinstance(value, Enum): value = value.value self.ctypes_ptr.contents.__setattr__(name, value) def c_p_pointer(cptr, targettype): targettype_ptr = convert_type_to_ctype(targettype) if isinstance(targettype, Array): if py_is_dataclass(targettype._type): return ctypes.cast(cptr.value, ctypes.py_object).value newa = ctypes.cast(cptr, targettype_ptr) return newa else: if py_is_dataclass(targettype): if cptr.value is None: return None return ctypes.cast(cptr, ctypes.py_object).value targettype_ptr = ctypes.POINTER(targettype_ptr) newa = ctypes.cast(cptr, targettype_ptr) if is_ctypes_Structure(targettype): # return after wrapping newa inside PointerToStruct return PointerToStruct(newa) return newa def p_c_pointer(ptr, cptr): if isinstance(ptr, ctypes.c_void_p): cptr.value = ptr.value else: # assign the address of ptr in memory to cptr.value # the case for numpy arrays converted to a pointer cptr.value = id(ptr) def empty_c_void_p(): class ctypes_c_void_p(ctypes.c_void_p): def __eq__(self, value): return self.value == value.value def __repr__(self): return str(self.value) return ctypes_c_void_p() def sizeof(arg): return ctypes.sizeof(convert_type_to_ctype(arg)) def ccallable(f): if py_is_dataclass(f): return convert_to_ctypes_Structure(f) return f def ccallback(f): return f class lpython: """ The @lpython decorator compiles a given function using LPython. The decorator should be used from CPython mode, i.e., when the module is being run using CPython. When possible, it is recommended to use LPython for the main program, and use the @cpython decorator from the LPython mode to access CPython features that are not supported by LPython. """ def __init__(self, function): def get_rtlib_dir(): current_dir = os.path.dirname(os.path.abspath(__file__)) return os.path.join(current_dir, "..") def get_type_info(arg): # return_type -> (`type_format`, `variable type`, `array struct name`) # See: https://docs.python.org/3/c-api/arg.html for more info on type_format if arg == f64: return ('d', "double", 'r64') elif arg == f32: return ('f', "float", 'r32') elif arg == i64: return ('l', "long int", 'i64') elif arg == i32: return ('i', "int", 'i32') elif arg == bool: return ('p', "bool", '') elif isinstance(arg, Array): t = get_type_info(arg._type) if t[2] == '': raise NotImplementedError("Type %r not implemented" % arg) return ('O', ["PyArrayObject *", "struct "+t[2]+" *", t[1]+" *"], '') else: raise NotImplementedError("Type %r not implemented" % arg) def get_data_type(t): if isinstance(t, list): return t[0] else: return t + " " self.fn_name = function.__name__ # Get the source code of the function source_code = getsource(function) source_code = source_code[source_code.find('\n'):] dir_name = "./lpython_decorator_" + self.fn_name if not os.path.exists(dir_name): os.mkdir(dir_name) filename = dir_name + "/" + self.fn_name # Open the file for writing with open(filename + ".py", "w") as file: # Write the Python source code to the file file.write("@ccallable") file.write(source_code) # ---------------------------------------------------------------------- types = function.__annotations__ self.arg_type_formats = "" self.return_type = "" self.return_type_format = "" self.arg_types = {} counter = 1 for t in types.keys(): if t == "return": type = get_type_info(types[t]) self.return_type_format = type[0] self.return_type = type[1] else: type = get_type_info(types[t]) self.arg_type_formats += type[0] self.arg_types[counter] = type[1] counter += 1 # ---------------------------------------------------------------------- # `arg_0`: used as the return variables # arguments are declared as `arg_1`, `arg_2`, ... variables_decl = "" if self.return_type != "": variables_decl = "// Declare return variables and arguments\n" variables_decl += " " + get_data_type(self.return_type) + "arg_" \ + str(0) + ";\n" # ---------------------------------------------------------------------- # `PyArray_AsCArray` is used to convert NumPy Arrays to C Arrays # `fill_array_details` contains arrays operations to be # performed on the arguments # `parse_args` are used to capture the args from CPython # `pass_args` are the args that are passed to the shared library function fill_array_details = "" parse_args = "" pass_args = "" numpy_init = "" for i, t in self.arg_types.items(): if i > 1: parse_args += ", " pass_args += ", " if isinstance(t, list): if numpy_init == "": numpy_init = "// Initialize NumPy\n import_array();\n\n " fill_array_details += f"""\n // fill array details for args[{i-1}] if (PyArray_NDIM(arg_{i}) != 1) {{ PyErr_SetString(PyExc_TypeError, "Only 1 dimension is implemented for now."); return NULL; }} {t[1]}s_array_{i} = malloc(sizeof(struct r64)); {{ {t[2]}array; // Create C arrays from numpy objects: PyArray_Descr *descr = PyArray_DescrFromType(PyArray_TYPE(arg_{i})); npy_intp dims[1]; if (PyArray_AsCArray((PyObject **)&arg_{i}, (void *)&array, dims, 1, descr) < 0) {{ PyErr_SetString(PyExc_TypeError, "error converting to c array"); return NULL; }} s_array_{i}->data = array; s_array_{i}->n_dims = 1; s_array_{i}->dims[0].lower_bound = 0; s_array_{i}->dims[0].length = dims[0]; s_array_{i}->is_allocated = false; }}""" pass_args += "s_array_" + str(i) else: pass_args += "arg_" + str(i) variables_decl += " " + get_data_type(t) + "arg_" + str(i) + ";\n" parse_args += "&arg_" + str(i) if parse_args != "": parse_args = f"""\n // Parse the arguments from Python if (!PyArg_ParseTuple(args, "{self.arg_type_formats}", {parse_args})) {{ return NULL; }}""" # ---------------------------------------------------------------------- # Handle the return variable if any; otherwise, return None fill_return_details = "" if self.return_type != "": fill_return_details = f"""\n\n // Call the C function arg_0 = {self.fn_name}({pass_args}); // Build and return the result as a Python object return Py_BuildValue("{self.return_type_format}", arg_0);""" else: fill_return_details = f"""{self.fn_name}({pass_args}); Py_RETURN_NONE;""" # ---------------------------------------------------------------------- # Python wrapper for the Shared library template = f"""// Python headers #include // NumPy C/API headers #define NPY_NO_DEPRECATED_API NPY_1_7_API_VERSION // remove warnings #include // LPython generated C code #include "{self.fn_name}.h" // Define the Python module and method mappings static PyObject* define_module(PyObject* self, PyObject* args) {{ {numpy_init}{variables_decl}{parse_args}\ {fill_array_details}{fill_return_details} }} // Define the module's method table static PyMethodDef module_methods[] = {{ {{"{self.fn_name}", define_module, METH_VARARGS, "Handle arguments & return variable and call the function"}}, {{NULL, NULL, 0, NULL}} }}; // Define the module initialization function static struct PyModuleDef module_def = {{ PyModuleDef_HEAD_INIT, "lpython_module_{self.fn_name}", "Shared library to use LPython generated functions", -1, module_methods }}; PyMODINIT_FUNC PyInit_lpython_module_{self.fn_name}(void) {{ PyObject* module; // Create the module object module = PyModule_Create(&module_def); if (!module) {{ return NULL; }} return module; }} """ # ---------------------------------------------------------------------- # Write the C source code to the file with open(filename + ".c", "w") as file: file.write(template) # ---------------------------------------------------------------------- # Generate the Shared library # TODO: Use LLVM instead of C backend r = os.system("lpython --show-c --disable-main " + filename + ".py > " + filename + ".h") assert r == 0, "Failed to create C file" gcc_flags = "" if platform.system() == "Linux": gcc_flags = " -shared -fPIC " elif platform.system() == "Darwin": gcc_flags = " -bundle -flat_namespace -undefined suppress " else: raise NotImplementedError("Platform not implemented") from numpy import get_include from distutils.sysconfig import get_python_inc, get_python_lib python_path = "-I" + get_python_inc() + " " numpy_path = "-I" + get_include() + " " rt_path_01 = "-I" + get_rtlib_dir() + "/../libasr/runtime " rt_path_02 = "-L" + get_rtlib_dir() + " -Wl,-rpath " \ + get_rtlib_dir() + " -llpython_runtime " python_lib = "-L" + get_python_lib() + "/../.. -lpython3.10 -lm" r = os.system("gcc -g" + gcc_flags + python_path + numpy_path + filename + ".c -o lpython_module_" + self.fn_name + ".so " + rt_path_01 + rt_path_02 + python_lib) assert r == 0, "Failed to create the shared library" def __call__(self, *args, **kwargs): import sys; sys.path.append('.') # import the symbol from the shared library function = getattr(__import__("lpython_module_" + self.fn_name), self.fn_name) return function(*args, **kwargs)