Optimized port cache coherency support is used to enable data and instruction cache coherency for multiple MicroBlaze cores. It provides reliable exclusive transactions to implement software spinlocks and simplifies multi-processor (MP) systems where data is shared among the processors. Any data that is updated from one MicroBlaze is guaranteed to be seen from its peer processors without any special software interactions other than ordinary single MicroBlaze rules for handling self-modifying code. This data manipulation information exchange is handled by the snooping mechanism provided by the AXI Coherency Extension (ACE) (see the AMBA AXI and ACE Protocol Specification). Distributed Virtual Memory (DVM) messages are also available with ACE to ensure that memory management units (MMU) and branch target caches are updated across the system when related changes are performed by any of the connected processors.
The 16 cache interfaces optimized for MicroBlaze provide support for up to eight cache coherent MicroBlaze processors, as shown in the following figure.
