What is FluxPerp
An overview of the FluxPerp protocol — GPU-accelerated perpetual futures on Solana.
FluxPerp is a perpetual futures exchange built on Solana. The matching engine runs on a GPU cluster that processes order events off-chain and settles net state transitions on-chain in batches. This architecture separates throughput-bound work (matching) from consensus-bound work (finality), which is why FluxPerp can sustain sub-millisecond matching latency while preserving non-custodial settlement guarantees.
What it does
Users deposit collateral into a margin account controlled by a Solana program. From there they can open leveraged long or short positions on perpetual contracts. Positions accrue or pay funding every 8 hours. The protocol never holds user funds in a hot wallet — all withdrawals are processed through the on-chain program with a cryptographic proof from the matching engine.
What it is not
FluxPerp is not a spot exchange. There is no token-to-token swapping, no liquidity pools, and no automated market maker. All price discovery happens through a central limit order book (CLOB) maintained by the matching engine.
FluxPerp is not a copy of dYdX or Drift. Those protocols have different trade-offs around decentralization of the sequencer. FluxPerp's matching engine is operated by the protocol team with a roadmap toward permissionless validation; the current design prioritizes performance and auditability over sequencer decentralization.
Key numbers
| Metric | Value |
|---|---|
| Matching latency (p99) | < 1 ms |
| Supported markets | SOL-PERP, BTC-PERP, ETH-PERP, JUP-PERP |
| Max leverage | 50× |
| Settlement | Solana mainnet-beta |
| Collateral | USDC |
How the pieces fit
The protocol has three main components:
Matching engine — A GPU-resident process that maintains the order book and matches taker orders against resting limit orders. State is snapshotted periodically and a Merkle root is published on-chain.
Settlement program — A Solana program (FluxSettlement...) that processes deposits, withdrawals, and liquidations. It verifies proofs from the matching engine before updating balances.
Oracle network — A set of price feeds (Pyth primary, Switchboard fallback) that the settlement program reads to determine mark price for PnL and liquidation threshold calculations.
The matching engine snapshot is published to Arweave every 256 batches. Anyone can reconstruct the full order history and verify that the on-chain state is consistent with matching engine output.
Next steps
- Quick Start — deposit collateral and place your first order
- Architecture — detailed breakdown of each subsystem
- Order Types — market, limit, stop, and conditional orders