A Case Study of Implementing Supernode Transformations

作者：Steinbrecher Johann^*; Philippidis Cesar J; Shang Weijia

来源：International Journal of Parallel Programming, 2014, 42(2): 320-342.

DOI：10.1007/s10766-013-0248-7

• Linear scheduling
• Wavefront scheduling
• Supernode transformation
• Loop tiling
• Loop blocking
• Multithreading
• Loop skewing

Supernode transformation is a technique to decrease the communication overhead by partitioning and scheduling a loop nest to a multi-processor system. This is achieved by grouping a number of iterations in a perfectly nested loop with regular dependences as a . Previous work has been focusing on finding the optimal supernode size and shape as well as an optimal execution schedule for multi-processor systems with unbounded resources. This paper emphasizes on the actual implementation strategies of supernode transformations on multi-core systems with limited resources. Using an example, the longest common subsequence (LCS) problem, we present and compare three different multithreading implementations. A formula for the total execution time of each method is presented. The techniques are benchmarked on a 12-core and a 4-core machine. On the 12-core machine our first technique, which yields increased data locality, speeds up the unaltered sequential loop nest 16.7 times. Combining this technique with skewing the loop by changing the linear schedule scores a 42.6 speedup. A more sophisticated method that executes entire rows of the loop nest in one thread scores a 59.5 speedup. Concepts presented and discussed in this paper on the LCS problem serve as basic foundation for implementations at regular dependence algorithms.

• 出版日期2014-4