Load/Store Architecture
Developed to simplify CPU design and improve performance
Memory wall: CPUs keep getting faster than memory
Memory accesses slow down CPU, limit compiler optimizations
Change instruction set to make most instructions independent of memory
Data processing instructions can access registers only
Load data into the registers
Process the data
Store results back into memory
More effective when more registers are available
Register/Memory Architecture
Data processing instructions can access memory or registers
Memory wall is not very high at lower CPU speeds (e.g. under 50 MHz)