NervaPy

Framework for exploring advanced program synthesis and transformation techniques

NervaPy is an experimental Python framework based on PeachPy that explores advanced program synthesis and transformation techniques. It provides a foundation for research into automated code generation, optimization, and program analysis.

About

This project builds upon the solid foundation provided by PeachPy's assembly generation capabilities to explore:

• Advanced program synthesis techniques
• Code transformation and optimization strategies
• Meta-programming approaches for high-performance code generation
• Research into automated assembly kernel generation

Origin

NervaPy is based on PeachPy by Marat Dukhan. PeachPy was originally designed as a Python framework for writing high-performance assembly kernels with features like:

• Universal assembly syntax for multiple platforms
• Automatic register allocation
• Cross-platform ABI compatibility
• Support for x86-64 instructions up to AVX-512

Installation

pip install pynerva

Usage

The basic API remains similar to PeachPy while providing additional experimental features:

from nervapy import *
from nervapy.x86_64 import *

# Example usage similar to original PeachPy
x = Argument(ptr(const_float_), name="x")
y = Argument(ptr(const_float_))

with Function("DotProduct", (x, y), float_,
              target=uarch.default + isa.sse4_2):

    reg_x, reg_y = GeneralPurposeRegister64(), GeneralPurposeRegister64()

    LOAD.ARGUMENT(reg_x, x)
    LOAD.ARGUMENT(reg_y, y)

    xmm_x = XMMRegister()
    MOVAPS(xmm_x, [reg_x])
    MOVAPS(xmm2, [reg_y])

    DPPS(xmm_x, xmm2, 0xF1)

    RETURN(xmm_x)

License

This project maintains the same Simplified BSD License as the original PeachPy.

Continuous Integration

This project uses GitHub Actions for continuous integration:

• Test Suite: Runs on Python 3.8-3.12 across Ubuntu, Windows, and macOS
• Code Quality: Security scanning, type checking, formatting, and complexity analysis
• Release Builds: Automated wheel building and PyPI publishing

All tests run automatically on every push and pull request to ensure code quality and compatibility.

Acknowledgements

This work is based on PeachPy by Marat Dukhan and the research conducted at the HPC Garage lab in the Georgia Institute of Technology.