Chapter 18. HT Power Management

The Previous Chapter

The previous chapter described the features of the optional HyperTransport double-hosted chain topologies. Topics include the reasons behind sharing and non-sharing chains, PCI configuration space registers used to initialize the fabric for multiple hosts , and tunnel support for upstream and downstream packets moving in both directions.

This Chapter

HT provides a variety of mechanisms to support power management. These mechanisms include LDTSTOP#, LDTREQ#, STPCLK messages, and STOP_GRANT messages. While these mechanisms are optional for HT devices, the specification requires this support for x86-based platforms. Note also that functions other than power management may make use of these signals and messages. This chapter discusses the strategy employed by HT for implementing power management and how a given platform can use these mechanisms to support power management.

The Next Chapter

The next chapter summarizes some of the major additions to the HyperTransport protocol which will be forthcoming in Release 1.05 and Release 1.1 of the specification. Collectively, these additions are referred the HyperTransport Networking Extensions, and target some of the special requirements of communications processing. Key features include a message passing protocol for larger packets, a formal definition for switch devices, a link-level error recovery method, sixteen optional additional posted write virtual channels with defined arbitration and bandwidth allocation, direct peer-peer transfers, an increase in the number of outstanding transactions for host bridges, and a 64-bit addressing option.