How to dynamically optimize the memory access in multiprocessor computing systems, programs, and computing system

摘要

An apparatus, method, and program product for optimizing a multiprocessor computing system by sampling memory reference latency and adjusting system components accordingly. While a process is executing in a computing system, memory reference sampling of memory locations is performed from the shared memory of the computing system referenced in the running process. Each sampled memory reference collected from sampling is associated with a latency and a physical memory location in shared memory. Each sampled memory reference is analyzed to identify a segment of the memory location in shared memory corresponding to the sub-optimal latency, and one or more identified based on the sampled memory reference analyzed The physical location of the segment, the processor on which one or more processes referencing the identified segment are executed, the status associated with the identified segment or segments, or any combination thereof are dynamically adjusted , Thereby optimizing the memory access of the multiprocessor computing system. [Selection] Figure 3