High-performance Number Theoretic Transform (NTT) library for FHE operations.

NTT is a critical primitive for efficient polynomial arithmetic in FHE schemes. This library provides optimized implementations with GPU acceleration support.

• SIMD optimizations: AVX2/AVX-512 vectorization
• GPU support: CUDA and Metal backends
• Configurable parameters: Multiple NTT sizes and moduli
• Integration ready: Works with luxfi/fhe and luxfi/lattice

pip install torus-ntt

from torus.ntt import NTT

# Create NTT instance
ntt = NTT(n=1024, modulus=0x1fffffffffe00001)

# Forward and inverse transforms
coeffs = [1, 2, 3, 4] + [0] * 1020
ntt_coeffs = ntt.forward(coeffs)
recovered = ntt.inverse(ntt_coeffs)

• luxfi/fhe
• Lux Network