HFT strategy coding on hardware

Question

Hardware accelaration and embedded programming has mostly been used so far to parse datafeed and/or to route orders to exchange. Have there been attempts to write simpler HFT strategies such as equity market-making in hardware? Have they been successful? Which companies are doing this and what kind of programming model is used?

Answer 1

FPGAs have been used to do this. You quickly and relatively cheaply accelerate all sorts of algorithms, such as neural networks. If you can do it in C, you can probably do it faster and in parallel on an FGPA.

See AutomatedTrader's article - FPGA's - Parallel Perfection?

Answer 2

Also there are offerings like CUDA, AMD Stream, Tilera. Exegy uses Tilera for application acceleration.