mirror of
https://github.com/lux-apps/papers.git
synced 2026-10-30 18:15:21 +00:00
zeekay
de41e5d072
Retract the fabricated '581M/597ns / 100M+/s / 434M GPU / planet-scale' headline (a time.Sleep sim + x40 fudge that measured nothing) across the corpus. Flagship lux-lightspeed-dex rewritten (decomplected d-chain engine / dexvm atomic proxy / Go wrapper / 0x9010 CLOB facade; measured GB10 104-126M & M4 Max ~250M fills/s byte-parity-gated; Knuth-D division lever; Block-STM settlement; P3Q->starkfri rollup to Z-Chain). Technical papers (precompile-engineering, gpu-crypto-stack, evmgpu-benchmark, p3q-corona, smart-order-routing, fhe-mpc-hybrid) carry the measured impl + cross-ref the flagship. >1B/s framed honestly as N-node aggregation, never single-GPU. All compile.
107 lines
4.0 KiB
TeX
107 lines
4.0 KiB
TeX
\documentclass[11pt,letterpaper]{article}
|
|
\usepackage{../shared/luxcover}
|
|
\usepackage[utf8]{inputenc}
|
|
\usepackage[T1]{fontenc}
|
|
\usepackage{amsmath,amssymb,amsfonts,amsthm}
|
|
\usepackage{mathtools}
|
|
\usepackage{graphicx}
|
|
\usepackage{booktabs}
|
|
\usepackage{hyperref}
|
|
\usepackage{enumitem}
|
|
\usepackage{algorithm}
|
|
\usepackage{algorithmic}
|
|
\usepackage{xcolor}
|
|
\usepackage{listings}
|
|
\usepackage{tikz}
|
|
\usepackage{float}
|
|
\usepackage{array}
|
|
\usepackage{multirow}
|
|
\input{../shared/lstlang}
|
|
|
|
\hypersetup{colorlinks=true,linkcolor=black,citecolor=blue,urlcolor=blue}
|
|
|
|
\lstset{
|
|
basicstyle=\ttfamily\scriptsize,
|
|
breaklines=true,
|
|
frame=single,
|
|
numbers=left,
|
|
numberstyle=\tiny\color{gray},
|
|
keywordstyle=\color{blue!70!black}\bfseries,
|
|
commentstyle=\color{green!50!black}\itshape,
|
|
stringstyle=\color{red!60!black},
|
|
showstringspaces=false,
|
|
}
|
|
|
|
\newtheorem{theorem}{Theorem}
|
|
\newtheorem{lemma}[theorem]{Lemma}
|
|
\newtheorem{proposition}[theorem]{Proposition}
|
|
\newtheorem{corollary}[theorem]{Corollary}
|
|
\newtheorem{definition}{Definition}
|
|
\newtheorem{remark}{Remark}
|
|
|
|
\title{Lux Lightspeed DEX: High-Frequency Trading at the Speed of Light}
|
|
|
|
\author{
|
|
Zach Kelling\thanks{Corresponding author: [email protected]}\\
|
|
\textit{Lux Industries, Inc.}\\
|
|
\texttt{[email protected]}
|
|
}
|
|
|
|
\date{}
|
|
|
|
\begin{document}
|
|
|
|
\luxcoverpage
|
|
|
|
\begin{abstract}
|
|
We present \textbf{Lux Lightspeed DEX}, a decentralized exchange whose matching
|
|
engine \emph{is} a chain. The DEX is the \textbf{D-Chain}: a single-operator
|
|
high-frequency-trading consensus chain with its own consensus, \texttt{zapdb}
|
|
persistence, and an execution-root state commitment, and the only place orders
|
|
match or DEX state lives. A stateless \textbf{atomic ZAP proxy} (\texttt{chains/dexvm})
|
|
on the public Lux network relays orders to the D-Chain and moves value atomically
|
|
via shared-memory import/export, holding no order book of its own. The design is
|
|
\emph{decomplected}: matching and state in one home, atomic transport in another,
|
|
the client wrapper in a third, and an EVM CLOB facade (precompile \texttt{0x9010})
|
|
in a fourth. Determinism is enforced by running the matcher inside
|
|
\texttt{Block.Verify} against a versioned overlay, committing to \texttt{zapdb} at
|
|
\texttt{Accept}, with the proposer relaying once at block build and carrying the
|
|
fills in the block bytes---no per-validator divergence.
|
|
|
|
We report \textbf{measured}, byte-parity-gated benchmarks on real hardware. A
|
|
byte-exact \texttt{uint256} GPU CLOB matcher sustains \textbf{104--126\,M fills/s}
|
|
(6.9\,ns/match) on an NVIDIA GB10 and \textbf{$\sim$250\,M fills/s} (4.15\,ns/match)
|
|
on an Apple M4 Max, byte-identical to a CPU reference oracle across millions of fuzz
|
|
cases. Settlement via market-sharded Block-STM detects conflicts at 2.2\,M fills/s
|
|
(CPU), byte-identical to the GPU conflict-detection kernel. Per-market trade batches
|
|
roll up to the Z-Chain, which verifies one proof per $K$ trades: P3Q (ML-DSA-65
|
|
attestation) at 4{,}714 verifies/s and \texttt{starkfri} (STARK/FRI validity proof)
|
|
at 986 verifies/s for a toy AIR. We give an honest treatment of throughput beyond
|
|
$10^{9}$/s: a single GPU does \emph{not} clear that for byte-exact \texttt{uint256}
|
|
matching, so it is reached by $N$-node aggregation (per-book arenas are independent)
|
|
or a cheaper fixed-point representation, while settlement \emph{finality} clears
|
|
$10^{9}$/s at fleet scale through the rollup.
|
|
|
|
\textbf{Keywords}: decentralized exchange, high-frequency trading, order matching,
|
|
MEV resistance, GPU acceleration, rollup, Block-STM
|
|
\end{abstract}
|
|
|
|
\input{sections/01-introduction}
|
|
\input{sections/02-architecture}
|
|
\input{sections/03-consensus-determinism}
|
|
\input{sections/04-mev-resistance}
|
|
\input{sections/05-cross-chain-swaps}
|
|
\input{sections/06-precompile-clob-facade}
|
|
\input{sections/07-gpu-matching}
|
|
\input{sections/08-settlement-blockstm}
|
|
\input{sections/09-rollup-zchain}
|
|
\input{sections/10-benchmarks}
|
|
\input{sections/11-security}
|
|
\input{sections/12-economic-model}
|
|
\input{sections/13-related-work}
|
|
\input{sections/14-future-work}
|
|
\input{sections/15-conclusion}
|
|
\input{sections/99-acknowledgments}
|
|
|
|
\end{document}
|