P-Tile Avalon® Streaming Intel® FPGA IP for PCI Express* User Guide

ID 683059
Date 3/28/2022
Public

A newer version of this document is available.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

Document Table of Contents

2.2.3. Transaction Layer Overview

The following figure shows the major blocks in the P-Tile Avalon® -ST IP for PCI Express Transaction Layer:

Figure 7. P-Tile Avalon® -ST IP for PCI Express Transaction Layer Block Diagram

The RAS (Reliability, Availability, and Serviceability) block includes a set of features to maintain the integrity of the link.

For example: Transaction Layer inserts an optional ECRC in the transmit logic and checks it in the receive logic to provide End-to-End data protection.

When the application logic sets the TLP Digest (TD) bit in the Header of the TLP, the P-Tile Avalon® -ST IP for PCIe will append the ECRC automatically.

Note that in TLP Bypass mode, the PCIe Hard IP does not generate/check the ECRC and will not remove it if the received TLP has the ECRC.

The TX block sends out the TLPs that it receives as-is. It also sends the information about non-posted TLPs to the CPL Timeout Block for CPL timeout detection.

The P-Tile Avalon® -ST IP for PCI Express RX block consists of two main blocks:
  • Filtering block: This module checks if the TLP is good or bad and generates the associated error message and completion. It also tracks received completions and updates the completion timeout (CPL timeout) block.
  • RX Buffer Queue: The P-Tile IP for PCIe has separate queues for posted/non-posted transactions and completions. This avoids head-of-queue blocking on the received TLPs and provides flexibility to extract TLPs according to the PCIe ordering rules.
Figure 8. P-Tile Avalon® -ST IP for PCI Express RX Block Overview
Note: The Received CPL Processing block includes the CPL tracking mechanism.
Note: The Avalon-ST interface uses a split-bus architecture. In the x16 and x8 configurations, the 512-bit Avalon-ST data bus consists of two segments of 256-bit data. This is done to improve the bandwidth efficiency of this interface. With this split-bus architecture, two TLP packets can be transmitted or received in a single clock cycle (e.g., if a TLP ends in the lower 256-bit segment,the next TLP can start in the upper 256-bit segment in the same clock cycle). For more details, refer to Avalon-ST Interface.