Adaptive Granularity Memory Systems: A Tradeoff Between Storage Efficiency and Throughput
The authors propose adaptive granularity to combine the best of ne-grained and coarse-grained memory accesses. They augment virtual memory to allow each page to specify its preferred granularity of access based on spatial locality and error-tolerance tradeoffs. They use sector caches and sub-ranked memory systems to implement adaptive granularity. They also show how to incorporate adaptive granularity into memory access scheduling. They evaluate their architecture with and without ECC using memory intensive benchmarks from the SPEC, Olden, PARSEC, SPLASH2, and HPCS benchmark suites and micro-benchmarks.