# Copyright (C) 2007-2020 Anders Logg and Michal Habera
#
# This file is part of FFCx.(https://www.fenicsproject.org)
#
# SPDX-License-Identifier: LGPL-3.0-or-later
"""Main interface for compilation of forms.
Breaks the compilation into several sequential stages.
The output of each stage is the input of the next stage.
Compiler stages
---------------
0. Language, parsing
- Input: Python code or .ufl file
- Output: UFL form
This stage consists of parsing and expressing a form in the UFL form
language. This stage is handled by UFL.
1. Analysis
- Input: UFL form
- Output: Preprocessed UFL form and FormData (metadata)
This stage preprocesses the UFL form and extracts form metadata. It
may also perform simplifications on the form.
2. Code representation
- Input: Preprocessed UFL form and FormData (metadata)
- Output: Intermediate Representation (IR)
This stage examines the input and generates all data needed for code
generation. This includes generation of finite element basis
functions, extraction of data for mapping of degrees of freedom and
possible precomputation of integrals. Most of the complexity of
compilation is handled in this stage.
The IR is stored as a dictionary, mapping names of UFCx functions to
data needed for generation of the corresponding code.
3. Code generation
- Input: Intermediate Representation (IR)
- Output: C code
This stage examines the IR and generates the actual C code for the
body of each UFCx function.
The code is stored as a dictionary, mapping names of UFCx functions to
strings containing the C code of the body of each function.
4. Code formatting
- Input: C code
- Output: C code files
This stage examines the generated C++ code and formats it according
to the UFCx format, generating as output one or more .h/.c files
conforming to the UFCx format.
"""
from __future__ import annotations
import logging
import typing
from time import time
import numpy.typing as npt
from ffcx.analysis import analyze_ufl_objects
from ffcx.codegeneration.codegeneration import generate_code
from ffcx.formatting import format_code
from ffcx.ir.representation import compute_ir
logger = logging.getLogger("ffcx")
def _print_timing(stage: int, timing: float) -> None:
logger.info(f"Compiler stage {stage} finished in {timing:.4f} seconds.")
[docs]
def compile_ufl_objects(
ufl_objects: list[typing.Any],
options: dict[str, int | float | npt.DTypeLike],
object_names: dict[int, str] | None = None,
namespace: str | None = None,
visualise: bool = False,
) -> tuple[str, str]:
"""Generate UFCx code for a given UFL objects.
Args:
ufl_objects: Objects to be compiled. Accepts elements, forms,
integrals or coordinate mappings.
object_names: Map from object Python id to object name
namespace: Convenience namespace/prefix for generated code.
options: Options
visualise: Toggle visualisation
Returns: tuple of declaration and implementation.
"""
_object_names = object_names if object_names is not None else {}
_namespace = namespace if namespace is not None else ""
# Stage 1: analysis
cpu_time = time()
analysis = analyze_ufl_objects(ufl_objects, options["scalar_type"]) # type: ignore
_print_timing(1, time() - cpu_time)
# Stage 2: intermediate representation
cpu_time = time()
ir = compute_ir(analysis, _object_names, _namespace, options, visualise)
_print_timing(2, time() - cpu_time)
# Stage 3: code generation
cpu_time = time()
code = generate_code(ir, options)
_print_timing(3, time() - cpu_time)
# Stage 4: format code
cpu_time = time()
code_h, code_c = format_code(code)
_print_timing(4, time() - cpu_time)
return code_h, code_c