Methods and apparatus for improving the efficiency of an arithmetic logic unit (ALU) are provided. The ALU of the invention combines the operation of a single-cycle ALU with the processing speed of a pipelined ALU. Arithmetic operations are performed in two stages: a first stage that produces separate sum and carry results in a first cycle, and a second stage that produces a final result in one or more immediately subsequent cycles. While this produces final results in two or more clock cycles, useable partial results are produced each cycle, thus maintaining a one operation per clock cycle throughput.