In this paper, a new timing generation method is proposed for the performance analysis of embedded software. The time stamp generation of I/O accesses is crucial to performance estimation and architecture exploration in the timed functional simulation, which simulates the whole design at a functional level with timing. A portable compiler is modified to generate time-deltas, which are the estimated cycle counts between two adjacent I/O accesses, by counting the cycles of the intermediate representation (IR) operations and using a machine description that contains information on a target processor. Since the proposed method is based on the machine-independent IR of a compiler, the method can be applied to various processors by changing the machine description. The experimental results show that the proposed method is effective in that the average estimation error is about 2% and the maximum speed-up over the corresponding instruction-set simulators is about 300 times. The proposed meth...