is generally defined as the number of cycles the processor can wait to release the System interface to slave state for an external request. When no processor requests are in progress, internal activity--such as refilling the primary cache from the secondary cache--can cause the processor to wait some number of cycles before releasing the System interface. Release latency is therefore more specifically defined as the number of cycles that occur between the assertion of ExtRqst* and the assertion of Release*.
There are three categories of release latency:
- Category 1: when the external request signal is asserted two cycles before the last cycle of a processor request, or two cycles before the last cycle of the last request in a cluster.
- Category 2: when the external request signal is not asserted during a processor request or cluster, or is asserted during the last cycle of a processor request or cluster.
- Category 3: when the processor makes an uncompelled change to slave state.
