Based on the definitions in Section 2. we devise aframework that breaks the verification process into threeinvariants that correspond to the three steps necessaryfor processing a memory operation (shown in Figure 1).First, memory operations are read from the instructionstream in program order (operations access the (highest level) cache in a possiblydifferent order, which we call cache order (enable hardware optimizations such as write buffers.Some time after accessing the cache, operations performand become visible in the globally shared memory. Thisoccurs when the affected data is written back to memoryor accessed by another processor. At the global memory,cache orders from all processors are combined into oneglobal memory order (
การแปล กรุณารอสักครู่..