OpenMP parallel model

OpenMP uses the fork-join parallel execution mode. It is executed by a main thread at the beginning. ProgramThis thread is executed in a serial manner until the first parallel guidance statement is encountered. The process is as follows: ① fork: the main thread creates a thread to execute Code; ② Join: after each thread is executed, it is synchronized or interrupted, and only the main thread is executed.
OpenMP programming is relatively simple. It makes full use of the features of the shared storage Architecture and avoids the overhead of message passing. Although it also supports coarse-grained parallelism, it mainly targets fine-grained loop-level parallelism. Another feature of OpenMP is that you do not need to make major changes to the Code when converting a serial program into a parallel program. Its shortcomings are that it can only run on machines with shared storage structures. Improper data placement may lead to other problems. Low loop granularity of parallelism may increase system overhead.