Method and device for memory recycle and allocation

Abstract

The invention discloses a method and device for memory recycle and allocation. The method comprises the following steps: a demand of memory allocation is received; after the demand of the memory allocation is received, if the memory of the current first memory management area dissatisfies the preset value, whether the recycle mode of the current first memory management area indicates recycle of the current first memory management area or not is judged, wherein the recycle mode of the current first memory management area is determined according to the topology condition of the node to which the current first memory management area belongs; if the recycle mode of the current first memory management area indicates recycle of the current first memory management area, the current first memory management area is recycled; the demand of the memory allocation is responded, and allocation is performed to the memory of the recycled current first memory management area. Through the mode, the method can ensure that the local memory can be in local allocation to the greatest extent, and avoids the unnecessary memory remote access to the greatest extent, so that the influence on the system performance is reduced.

Description

technical field [0001] The invention relates to the field of computer application technology, in particular to a method and device for reclaiming and allocating memory. Background technique [0002] With the rapid development of computer technology, people's demand for computers is increasing day by day. The computing power of a single processor (CPU) is limited, and people urgently need to combine multiple processors into a system. Non Uniform Memory Access Architecture (NUMA, Non Uniform Memory Access Architecture), because of its high scalability, is being It is increasingly used in the server field. [0003] NUMA is a distributed structure, which is characterized by: any processor can access any memory, peripherals, etc.; when each processor accesses different memories, there are different delays. Therefore, the main purpose of memory management is to allocate memory efficiently and quickly, so that each CPU can access local memory as much as possible, and release and ...

AI Technical Summary

Problems solved by technology

[0004] The R&D personnel of this application have discovered in the long-term research and development that in the existing technical solutions, the "recycling mode" is a global attribute, and the system determines the value of the "recycling mode" based on whether there are any two nodes in the current system The distance between them is greater than the preset "recycling distance". This scheme of determining the value of "recycling mode" makes the control granularity of the system too large, which cannot guarantee the local allocation of local memory well, resulting in unnecessary remote memory access , affecting system performance

Images