Allow registers to hold both data and addresses dynamically

 Increase the number of general-purpose registers

Improve cache efficiency by reducing memory access latency

Resolve dependencies caused by register reuse in instruction-level parallelism