eisenstein-cuda — Eisenstein Integer Constraint Math for CUDA/C

One header. Works in CUDA kernels. Works on bare-metal ESP32. Zero dependencies.

Eisenstein integer constraint checking: norm, multiply, conjugate, disk check, constraint check with margin, XOR dual-path anomaly detection.

Why Eisenstein

• Exact integer arithmetic (no float drift)
• Hexagonal symmetry (6-fold rotation)
• 6.8× denser than Pythagorean triples
• Norm N(a,b) = a²-ab+b² is always non-negative
• Closed under multiplication (ring property)

What's Included

• include/eisenstein.cuh — drop-in header, ~200 lines
• test_eisenstein.cu — CPU + GPU test suite
• Batch constraint check kernel (CUDA)
• Batch XOR dual-path kernel (CUDA)

Quick Start

nvcc -O3 -arch=sm_86 -I include test_eisenstein.cu -o test_eisenstein && ./test_eisenstein

Works On

• NVIDIA GPUs (CUDA 11.5+, sm_86/sm_89)
• CPUs (host functions)
• ESP32, STM32, any C99 compiler (use #include "eisenstein.cuh" without CUDA)
• FPGA soft processors

Composable With

• fleet-constraint-kernel: uses this header for constraint math
• guard2mask-gpu: uses this for Eisenstein workspace bounds
• flux-vm-gpu: uses EIS_CHECK/EIS_NORM opcodes from this
• snap-lut-eisenstein: FPGA snap tables using same Eisenstein math
• fleet-proto-rs: Rust constraint types that mirror these C types

License

Apache 2.0