Over the past several decades, memory technologies have exploitedcontinual scaling of CMOS to drastically improve performance andcost. Unfortunately, charge-based memories become unreliable beyond20 nm feature sizes. A promising alternative is Phase-Change-Memory(PCM) which leverages scalable resistive thermal mechanisms. Torealize PCM's potential, a number of challenges, including thelimited wear-endurance and costly writes, need to be addressed. Thisthesis introduces novel methodologies for encoding data on PCM which exploit asymmetries in read/write performance to minimize memory's wear/energy consumption. First, we map the problem to adistance-based graph clustering problem and prove it is NP-hard.Next, we propose two different approaches: an optimal solutionbased on Integer-Linear-Programming, and an approximately-optimal solution based on Dynamic-Programming. Our methods target both single-level and multi-level cell PCM and provide furtheroptimizations for stochastically-distributed data. We devise a lowoverhead hardware architecture for the encoder. Evaluationsdemonstrate significant performance gains of our framework.
展开▼
