When an instruction produces a result that is needed by a subsequent instruction, instead of waiting for the result to be written back to the register file, the result is “forwarded” directly from the pipeline stage where it is generated (usually from the ALU output) to the stage where it is needed. This technique enables the subsequent instruction to use the result immediately, without having to wait for the previous instruction to complete all stages of the pipeline.
Consider two instructions:
Instruction 1: R1 = R2 + R3Instruction 2: R4 = R1 * R5Instruction 2 would have to wait until Instruction 1 completes its write-back stage and writes the result into R1.R1 is taken directly from the ALU output of Instruction 1 and provided to Instruction 2 during its execution stage, allowing it to proceed without delay.