Metadata-Version: 2.4
Name: accurpy
Version: 0.1.6
Summary: Ultra-accurate and fast double-double FM_new approximation with ≤1 ULP guarantees
Author: AccurPy Authors
License-Expression: MIT
Project-URL: Homepage, https://github.com/yourname/accurpy
Project-URL: Issues, https://github.com/yourname/accurpy/issues
Project-URL: Source, https://github.com/yourname/accurpy
Keywords: numerics,double-double,special functions,accuracy
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: C
Classifier: Programming Language :: Python :: Implementation :: CPython
Classifier: Operating System :: OS Independent
Classifier: Topic :: Scientific/Engineering :: Mathematics
Requires-Python: >=3.8
Description-Content-Type: text/markdown
License-File: LICENSE
Requires-Dist: numpy>=1.20
Dynamic: license-file

# accurpy

Ultra-accurate and fast **double-double** (106-bit) approximation for the new FM function. The C extension ships a single FMA-enabled pipeline validated to stay within 1 ULP versus high-precision references, and a pure-Python double-double fallback is provided when the extension is unavailable.

## Install
```bash
pip install accurpy
```

## Usage
```python
import numpy as np
from accurpy import syncF

y = syncF(10.0, skip_exp=False)

x = np.geomspace(1e-12, 300.0, 1_000_000)
y_vec = syncF(x, skip_exp=True)
```

## Notes
- Scalar calls and NumPy arrays both dispatch to the fast C implementation when available.
- When the extension cannot be imported, the package falls back to the same algorithm implemented in pure Python double-double arithmetic (slower, but numerically identical).

## License
MIT
