shows the DMM macro architecture. The DMM consists of six blocks:
- Two TILERs, each with its own interconnect slave port for converting requests back and forth between the input virtual addressing mode and the output physical tiled addressing mode. The tiling conversions of requests, write data, and responses is performed entirely in the TILER blocks.
- Two reordering buffer and initiator nodes (ROBINs), each with its own interconnect master port to initiate requests to the SDRC and allow tiled data, tiled response, and split response reconstruction. The ROBIN block manages only the reordering buffer and performs data reordering due to the orientation.
- A physical address translator (PAT) for managing the memory fragmentation
- A priority extension generator (PEG) to generate priorities required by the SDRC; these priorities are not used in the DMM.
- A local interconnect and synchronization agent (LISA) to synchronize all DMM subsystems and provide access to their configuration registers
Figure 15-4 is a block diagram of the DMM.
CAUTION: The interconnect must ensure that virtually addressed requests target only a TILER port.