Hard Processor System Technical Reference Manual: Agilex™ 5 SoCs

ID 814346
Date 4/01/2024
Public
Document Table of Contents

11.2.1. FPGA-to-HPS Bridge

F2H bridge provides a way for initiators (IPs, accelerators) in the fabric to access HPS peripherals, which makes the peripherals extensions of the HPS system. The accelerator coordinates with the HPS MPU via mailboxes or semaphores in HPS memory, or via various interrupts and GPIOs exposed by the HPS to the fabric. Typical use cases involve the HPS MPU preparing space in memory for fabric accelerators to use, then allowing the accelerator to move and process large amounts of data in HPS memory. The HPS MPU can perform control functions such as inspecting headers in large streams of data to determine the next action and coordinating data movement among multiple fabric accelerators. This bridge places fabric initiators in the same hierarchy as the MPU in the HPS subsystem. F2H bridge supports IO cache coherency with the HPS MPU caches; fabric transactions can snoop the MPU caches, but the MPU caches cannot snoop activity in the fabric. In addition, using ACE-lite, the F2H bridge goes through the system memory management unit (SMMU). This allows fabric initiators to use the same virtual memory view as the MPU.

F2H bypass provides a way for initiators (IPs, accelerators) on the fabric to work with HPS through interface. A typical use case involves HPS MPU preparing the space in memory for fabric accelerators to use, then setting the accelerator free to move and process large amounts of data in HPS memory. More information can be found in the Fabric Bypass section of the MPFE and MPFE-lite Use Cases and also the Fabric Bypass section of the MPFE and MPFE-lite Functional Description.