R-tile Avalon® Streaming Intel® FPGA IP for PCI Express* Design Example User Guide

ID 683544
Date 12/13/2021
Public

A newer version of this document is available. Customers should click here to go to the newest version.

Document Table of Contents

1. Design Example Description

Updated for:
Intel® Quartus® Prime Design Suite 21.4
IP Version 4.0.0

The Programmed Input/Output (PIO) design example performs memory transfers from a host processor to a target device. In this example, the host processor requests single-dword Memory Read (MemRd) and Memory Write (MemWr) Transaction Layer Packets (TLPs).

The PIO design example automatically creates the files necessary to simulate and compile in the Intel® Quartus® Prime software. The design example covers a wide range of parameters. However, it does not cover all possible parameterizations of the R-tile Hard IP for PCIe.

This design example supports the following configurations:
  • Gen5 x16 Endpoint, with a 1024-bit data width.
  • Gen4 x16 Endpoint, with a 1024-bit data width.
  • Gen3 x16 Endpoint, with a 1024-bit data width.
  • Gen5 2x8 Endpoint, with a 512-bit data width.
  • Gen4 2x8 Endpoint, with a 512-bit data width.
  • Gen3 2x8 Endpoint, with a 512-bit data width.

The clock is provided from the coreclkout_hip output of the IP and runs at 500 MHz.

Note: In the 21.4 release of Intel® Quartus® Prime, this design example only supports the default settings in the Parameter Editor of the R-tile Avalon® Streaming IP for PCIe.
This design example includes the following components:
  • The generated R-tile Avalon® Streaming (Avalon-ST) Hard IP Endpoint variant (DUT) with the parameters you specified. This component drives TLP data received to the PIO application. It translates the PCIe serial data received from the link to the Avalon® -ST data format.
  • The PIO Application (APPS) component, which performs the necessary translation between the PCI Express TLPs and simple Avalon® Memory-mapped ( Avalon® -MM) writes and reads to the on-chip memory.
    Note: The current APPS component supports only single-cycle data transfers. Data transfers longer than one clock cycle are not supported.
  • An on-chip memory (MEM) component (one 32 KB memory for the x16 design example, and two 32 KB memories for the 2x8 design example).
  • Reset Release IP: This IP holds the control circuit in reset until the device has fully entered user mode. The FPGA asserts the nINIT_DONE output to signal that the device is in user mode. The nINIT_DONE signal is high until the entire device enters user mode. After nINIT_DONE deasserts (low), all logic is in user mode and operates normally.

In the Gen5 x16 design example, a PIO component with a 1024-bit data path is instantiated to interface with the 1024-bit DUT. Also, only one MEM device is instantiated as shown in the figure below.

Figure 1. Gen5 x16 Design Example Block Diagram

In the Gen5 2x8 design example, two PIO components with 512-bit data paths are instantiated to interface with the 2x512-bit DUT. Also, two MEM devices are instantiated as shown in the figure below.

Figure 2. Gen5 2x8 Design Example Block Diagram

For simulation purposes, this design example also generates a testbench that instantiates the PIO design example and a Root Port BFM to interface with the target Endpoint.

Note: The simulation testbench for the PCIe 2x8 PIO design example is configured for a single PCIe x8 link although the actual design implements two PCIe x8 links.
Figure 3. Block Diagram for the Platform Designer PIO Design Example Simulation Testbench

The test program writes to and reads back data from the same location in the on-chip memory. It compares the data read to the expected result. The test reports, "Simulation stopped due to successful completion" if no errors occur.

Figure 4.  Platform Designer System Contents for the R-tile Avalon® -ST PCI Express Gen5 x16 PIO Design Example
Figure 5.  Platform Designer System Contents for the R-tile Avalon® -ST PCI Express Gen5 2x8 PIO Design Example