In terms of memory compression, we focus more or less on the semantics of the Computer Architecture level; instead, this paper proposed a compression framework built on an object-based memory hierarchy with object-aware compression optimization.
About the ziqiwang 3 question. 1. The ISA proposed a store/load pointer instruction tailored to twizzler FOT design. 2/3. I don't think the panelty of a universal forward chunking compared with extra bit stored and do pointer remap is so big for overflow objects. I don't see the implementation so that I should really dig into the metadata bits design and runtime panelty of different operations of memory and ptrs.