cutcutcodec.core.compilation.sympy_to_torch.printer

Static C compilation of atomic sympy expression.

This is enable to write a C code, compile it with gcc, and import the file as a python function.

It is slowler to initialise than the dynamic version but it is faster to evaluate. This static evaluation do not support broadcasting.

Implemented functions:

• sympy.Abs

• sympy.Add +

• sympy.atan

• sympy.cbrt

• sympy.cos

• sympy.Eq

• sympy.exp

• sympy.GreaterThan

• sympy.LessThan

• sympy.log

• sympy.Max

• sympy.Min

• sympy.Mul *

• sympy.Or

• sympy.Piecewise

• sympy.Pow / and **

• sympy.sin

• sympy.sqrt

• sympy.StrictGreaterThan

• sympy.StrictLessThan

Not implemented functions:

• sympy.acos

• sympy.acosh

• sympy.Add + and -

• sympy.And

• sympy.arg

• sympy.asin

• sympy.asinh

• sympy.atan2

• sympy.atanh

• sympy.ceiling

• sympy.cosh

• sympy.Determinant

• sympy.erf

• sympy.floor

• sympy.HadamardProduct

• sympy.im

• sympy.ITE

• sympy.loggamma

• sympy.MatAdd

• sympy.Mod %

• sympy.Ne

• sympy.Not

• sympy.re

• sympy.sign

• sympy.sinh

• sympy.tan

• sympy.tanh

• sympy.Trace

• sympy.Tuple

Functions

c_piecewise(out, indexing, c_type, *parts)

C .

Details

cutcutcodec.core.compilation.sympy_to_torch.printer.c_piecewise(out: Symbol, indexing: defaultdict[Symbol, str], c_type: str, *parts: Atom) → tuple[set[str], set[Symbol], list[str]]

C … if … else … operation.

Examples

>>> from sympy.abc import x
>>> from sympy.functions.elementary.piecewise import Piecewise
>>> import numpy as np
>>> from cutcutcodec.core.compilation.sympy_to_torch.lambdify import _lambdify_c
>>> from cutcutcodec.core.compilation.sympy_to_torch.printer import _print_atomic, _printer
>>> func = _lambdify_c(_printer([(x, Piecewise((0, x < 0), (x, True)))], {}, {x}))
>>> func(np.array([np.nan, -np.inf, -1.0, 0.0, 1.0, np.inf]))
array([nan,  0.,  0.,  0.,  1., inf])
>>>