Unlocked memory application releasing method

Abstract

The invention provides an unlocked memory application releasing method. According to the unlocked memory application releasing method, a block of memory pool is occupied by each thread independently to serve as a cache, and all memory blocks applied from the caches are finally released back into the caches; when a memory application is carried out, an application thread obtains a block of memory from a corresponding cache, the thread number of the application thread is marked on the distributed memory block, and then the application is accomplished; after the memory block is used, the thread is released and the thread number, marked on the memory block, of the application thread is checked, the memory block is returned to the corresponding cache if the application belongs to the corresponding cache, and then releasing is accomplished; if the application does not belongs to the cache, comparison and replacement are carried out in sequence, the memory block is mounted on a one-way chain table of the application thread, and then releasing is accomplished. In addition, each time when a message is sent or received by a task, periodical detection is carried out on the application thread, and all the memory blocks on the one-way chain table are returned to the caches.

Description

technical field [0001] The invention relates to the technical field of embedded systems, in particular to a memory application release method for a multi-thread system. Background technique [0002] Embedded systems often use multithreading to complete different tasks, and the threads are connected to each other and communicate with each other using messages. By using shared memory, inter-thread communication, the destination thread can directly read the message of the source thread without copying. However, this leads to the need to use a mutex when applying for and releasing the memory used by the message, which affects efficiency. At present, there are generally two methods for applying and releasing heap memory. One is to directly call the memory application interface (ptmalloc) and release interface (free) of the general function library (glibc), and the other is to establish its own memory pool for management. [0003] The logic of the general function library (glibc...

AI Technical Summary

Problems solved by technology

Not to mention that multiple threads compete to acquire the lock of the global shared memory pool, even if an independent cache pool is defined for each thread, the overhead of the lock is unavoidable

Images