محمد حسيني نيا

In the past years, embedded systems have been used in many fields. It is expected that this trend will continue in the coming years. Considering the challenges in the energy consumption of embedded systems, energy harvesting has become a popular option to use as a power source for these systems. Due to the non-sustainability of energy harvesting, non-volatile processors play a key role in solving this problem. Non-volatile processors help to back up the contents of the system in the event of a disruption in the environmental power supply, and after the problem is resolved, the system can continue its work by continuing the previous execution instructions. On the other hand, the challenge in the non-volatile processor is that if there are many backups of the system, the efficiency of the system decreases, and the energy consumption increases. Previously, a solution to avoid overhead in non-volatile processor is that if there are many backups of the system, the efficiency of the system decreases, and the energy consumption increases. Previously, a solution to avoid overhead in non-volatile processor backups was to add a threshold voltage to sleep and re-energize. This means that the system has two threshold voltages. It goes to sleep after passing one, and a backup copy is made if it passes the other. In this regard, this research presents an effective method called TVTAC (Triple Voltage Threshold Approximate Cache), which by using it in embedded systems with energy harvesters and non-volatile processors, energy consumption is reduced and more progress is achieved in the execution of tasks. In general, TVTAC is based on reducing energy consumption in non-volatile memory, cache memory management, provision of appropriate hardware, using approximate computing and consideration of three voltage thresholds. The results of the eva‎luations on TVTAC show that this method has registered 28 percent at best and 13 percent progress of the tasks compared to other modern methods. In addition, TVTAC has managed to reduce energy consumption by 43.5 percent in the best case and 28.5 percent on average compared to the mentioned methods.

برداشت انرژي , پردازنده هاي غيرفرار , محاسبات تخميني , سامانه هاي نهفته

energy harvesting , non volatile processors , approximate computing , embedded systems

mohammad hosseininia

amir mahdi hosseini monazzah