On the Pentium 4 you can use the event x87 Instructions Retired. On the PIII you can use the event FP Operations Retired. However these will not give you a true meaning of FLOPs since they do not count floating point calculations using the SSE and SSE2 instructions (Pentium 4 only). Generally when calculating FLOPs you looking at you algorithm at a high level and see how many floating point operations there are per iteration. Than you multiply that by the total interations and divide by the execution time.