Line-Rate Post-Quantum Byzantine Consensus via L1D Balanced Ternary

Justin Grimm published a reference implementation for a 64-byte L1D cache-resident frame that compresses a 128-node consensus vote bitmask to 26 bytes using balanced ternary (3^5 = 243 ≤ 256), letting the full synchronous descriptor fit in one 64-byte cache line. The accompanying eBPF/XDP driver was benchmarked at 100GbE line-rate (99.2 Mpps aggregate across 8 queues, 12.4 Mpps/core) with p50 = 40.0 ns and p99.9 = 60.0 ns, below the 80.5 ns frame budget, while ML-DSA-44 (NIST FIPS 204) signature verification is offloaded over 2MB hugepage lock-free SPSC rings. The repository, released under the MIT License and targeting SOSP/OSDI/ISCA/ASPLOS, includes a Triton kernel that decompresses 5-trit packed bytes into FP16 ternary weights with zero shared-memory bank conflicts.

This repository contains the reference eBPF/XDP drivers, Triton GPU lookup kernels, and microbenchmarking suites for the paper: "Radix Economy and Balanced Ternary Microarchitectures: Resolving the Memory Wall in Line-Rate Post-Quantum Consensus and Nanoscale Computing" Target Venues: SOSP / OSDI / ISCA / ASPLOS 1. 64-Byte L1D Cache-Resident Frame : Compresses a 128-node consensus vote bitmask to 26 bytes $3^5 = 243 \le 256$ , enabling the entire synchronous descriptor epoch + BLAKE3 accumulator + status flags to fit within exactly one 64-byte L1D cache line. 2. Deterministic Wire Latency : Evaluated via eBPF XDP at 100GbE line-rate 99.2 Mpps aggregate across 8 queues, 12.4 Mpps/core with$p50 = 40.0\text{ ns}$ and$p99.9 = 60.0\text{ ns}$ , safely below the$80.5\text{ ns}$ frame budget. 3. State-Crypt Separation : Decouples the 64-byte synchronous consensus frame from asynchronous ML-DSA-44 NIST FIPS 204 signature verification offloaded over 2MB hugepage lock-free SPSC rings. 4. Conflict-Free GPU Decompression : Triton kernel maps 5-trit packed bytes into FP16 ternary weights with zero shared-memory bank conflicts using single-cycle hardware broadcast addressing. - xdp ternary filter.c - Production eBPF XDP C driver for line-rate packet parsing, SipHash-2-4 pre-authentication, monotonic epoch tracking, and fast-path quorum accumulation. - triton lut kernel.py - Triton GPU kernel for high-throughput 5-trit decompression on Tensor Cores. - benchmark harness.py - Microarchitectural evaluation reproducing the latency and throughput ablations across 64-byte ternary and 96-byte binary frames. - LICENSE - MIT License. Requires clang and libbpf clang -O2 -target bpf -c xdp ternary filter.c -o xdp ternary filter.o Attach to your 100GbE network interface e.g. eth0 in native XDP mode ip link set dev eth0 xdpgeneric obj xdp ternary filter.o sec xdp Requires PyTorch and Triton python3 triton lut kernel.py python3 benchmark harness.py @article{grimm2026radix, title={Radix Economy and Balanced Ternary Microarchitectures: Resolving the Memory Wall in Line-Rate Post-Quantum Consensus and Nanoscale Computing}, author={Grimm, Justin}, year={2026} } MIT License - Copyright c 2026 Justin Grimm.