This guide covers the multicast P2P example in ESP 2026.1.0: one dummy accelerator produces data for two dummy-accelerator consumers.
Platform support: the worked example uses Cadence Stratus
HLS and is currently limited to AMD/Xilinx FPGA targets.
What you will learn
- Create a multicast instance of ESP with multiple dummy accelerators
- Enable multicast routing on the DMA NoC planes
- Configure a producer and its consumers with the bare-metal software API
- Run the equivalent three-accelerator application with the ESP Linux API
- Keep every accelerator in the P2P group on a compatible coherence setting
What you will need
- Prerequisites
- Recommended
- Equipment
- A supported AMD/Xilinx FPGA board (see FPGA targets)
- (optional) A network router, to reach the SoC over SSH
What you can read
ESP4ML: Platform-Based Design of Systems-on-Chip for Embedded Machine Learning
(Best Paper Nominee) Design, Automation and Test in Europe Conference (DATE), 2020
Runtime Reconfigurable Memory Hierarchy in Embedded Scalable Platforms
(Invited) Asia and South Pacific Design Automation Conference (ASPDAC), 2019
What you can contribute
The ESP team welcomes external contributions and collaborations, including:
- An improved user-level API for accelerator configuration and scheduling
- Novel algorithms for run-time selection of the cache-coherence model in multi-accelerator applications
See the contributing guidelines.