System-level power estimation methodology using cycle- and bit-accurate TLM

We propose a new system-level methodology for relative power estimation, which is independent of register transfer level models. Our methodology monitors the number of bit transitions for all input/output gate signals on a bit- and cycle-accurate SystemC virtual platform model. For absolute results and reliable technology-based predictions of system power and speed (e.g. in future 32/22nm technology nodes and variations), relative metrics can be multiplied with bit energy coefficients provided by semiconductor technology datasheets and device models.

[1]  D. Gajski,et al.  Transaction Level Modeling in System Level Design , 2003 .

[2]  Kunle Olukotun,et al.  STAMP: Stanford Transactional Applications for Multi-Processing , 2008, 2008 IEEE International Symposium on Workload Characterization.

[3]  Miltos D. Grammatikakis,et al.  Power-Aware Multicore SoC and NoC Design , 2011, Multiprocessor System-on-Chip.

[4]  Luca Benini,et al.  System-level power optimization: techniques and tools , 1999, ISLPED '99.