Shared last-level cache management for GPGPUs with hybrid main memory | IEEE Conference Publication | IEEE Xplore