You are correct, I mistakenly thought it was 1/cycle because I had previously remembered IMUL taking 30 cycles (and now it has a 1 cycle throughput). Agner Fog

You are correct, I mistakenly thought it was 1/cycle because I had previously remembered IMUL taking 30 cycles (and now it has a 1 cycle throughput). Agner Fog is reporting anywhere from 4->30 cycles on semi-recent architectures.