连贯因果记忆
Coherent Causal Memory
摘要:
一致性因果记忆(CCM)是一种因果记忆,其中执行的前缀可以以一致的方式映射到全局记忆状态。虽然CCM要求冲突的写入对是全局有序的,但它允许写入相对于读取和非冲突的写入保持无序。然而,它支持断言,基于状态的程序推理使用广义Owicki-Gries证明大纲(断言可以附加到任何因果程序边缘)。事实上,我们表明,从推理的角度来看,CCM不同于顺序一致(SC)的内存,只在由用户添加的幽灵代码是不允许引入新的写-写比赛。虽然CCM提供了SC存储器的大部分形式推理杠杆,但它的实现效率要高得多。作为一个例子,我们描述了一个简单的编程原则,提供了x86-TSO之上的CCM。该规程比确保SC所需的规程要宽松得多;例如,它不会为至多一个线程写入任何变量的程序带来任何负担。
英文摘要:
Coherent causal memory (CCM) is causal memory in which prefixes of an execution can be mapped to global memory states in a consistent way. While CCM requires conflicting pairs of writes to be globally ordered, it allows writes to remain unordered with respect to both reads and nonconflicting writes. Nevertheless, it supports assertional, state-based program reasoning using generalized Owicki-Gries proof outlines (where assertions can be attached to any causal program edge). Indeed, we show that from a reasoning standpoint, CCM differs from sequentially consistent (SC) memory only in that ghost code added by the user is not allowed to introduce new write-write races. While CCM provides most of the formal reasoning leverage of SC memory, it is much more efficiently implemented. As an illustration, we describe a simple programming discipline that provides CCM on top of x86-TSO. The discipline is considerably more relaxed than the one needed to ensure SC; for example, it introduces no burden whatsoever for programs in which at most one thread writes to any variable.
机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。