Emulating the FMAdd Instruction, Part 1: 32-bit Floats