Overview

xcore.ai devices and selected xcore-200 devices include an integrated USB transceiver. lib_xud allows the implementation of both full-speed and high-speed USB 2.0 devices on these devices. The library provides an identical API for all devices.

The library performs all of the low-level I/O operations required to meet the USB 2.0 specification. This processing goes up to and includes the transaction level. It removes all low-level timing requirements from the application, allowing quick development of all manner of USB devices.

The XUD library runs in a single thread with endpoint and application tasks communicating with it via a combination of channel communication and shared memory variables.

One channel is required per IN or OUT endpoint. Endpoint 0 (the control endpoint) requires two channels, one for each direction. Please note that throughout this document the USB nomenclature is used: an OUT endpoint is used to transfer data from the host to the device, an IN endpoint is used when the host requests data from the device.

An example task diagram is shown in Fig. 1. Circles represent threads running with arrows depicting communication channels between these threads. In this configuration there is one thread that deals with endpoint 0, which has both the input and output channel for endpoint 0. IN endpoint 1 is dealt with by a second thread, and OUT endpoint 2 and IN endpoint 5 are dealt with by a third thread. Threads must be ready to communicate with the XUD library whenever the host demands it’s attention. If not, the XUD library will NAK.

It is important to note that, for performance reasons, tasks communicate with the XUD library using both XC channels and shared memory communication. Therefore, all tasks using the XUD library must be on the same tile as the library itself.

Fig. 1 XUD Overview