Meaning
Method of communication between a central processing unit and peripheral devices treats hardware registers as locations within the physical memory address space. The use of mmio allows the processor to use standard memory instructions, such as load and store, to interact with external hardware like network controllers or graphics chips. This architecture simplifies the instruction set because no specialized input or output commands are required to access device state.
It requires the memory management unit to map specific ranges of the address map to the hardware bus while excluding system ram.
System Mapping
Hardware designers assign unique addresses to each peripheral to avoid conflicts during data transfer. When mmio is active, a write to a specific memory location triggers a command in the hardware controller. This mapping is defined in the system memory map during the early stages of silicon design.
Performance Characteristic
Direct access to device registers through the cache hierarchy can introduce latency if not managed correctly. Software drivers must mark these regions as non-cacheable to ensure that every read or write reaches the hardware immediately. This prevents the cpu from using stale data from the internal buffer.
Software Interface
Programmers interact with these addresses using pointers in low level languages.