浙江大学多核计算课件03.pptVIP

Fundamental Concepts of Parallel Programming Designing for Threads Task Decomposition Data Decomposition Data Flow Decomposition Challenges Synchronization The process by which two or more threads coordinate their activities Communication The bandwidth and latency issues associated with exchanging data between threads Load balancing The distribution of work across multiple threads so that they all perform roughly the same amount of work Scalability The challenge of making efficient use of a larger number of threads when software is run on more-capable systems Parallel Programming Patterns * ZJU-Intel Embedded Technology Center * Special care is needed to eliminate startup and shutdown latencies One thread’s output is the input to a second thread Data Flow Common in audio processing, imaging, and in scientific programming Multiple thread performing the same operation but on different blocks of data Data Common in GUI apps Different activities assigned to different threads Task Comments Design Decomposition Data Flow Wavefront Data Flow Pipeline Data Geometric Decomposition Task/Data Divide and Conquer Task Task-level parallelism Decomposition Pattern