Two optimization techniques are explored for a parallelizing compiler that automatically generates parallel code for distributed-memory systems. Two problems arise from message-passing code generated by the compiler: 1) messages contain redundant data, and 2) the same data is sometimes transmitted to different processors, yet the messages are repacked for each processor. In this paper, the technique that eliminates the first problem, redundant data, is found to yield benefits. The steps necessary to implement the technique into the parallelizing compiler are provided in detail. The second optimization technique yields negligible benefits and is not implemented. Testing of the implementation is described and the results of the testing are given to show the benefits achieved by the first optimization technique.

Recommended Citation: Martin J., Ferner C., Guinn C., Janicki T., (2010). Suppressing Independent Loops in Packing/Unpacking Loop Nests to Reduce Message Size for Message-Passing Code. UNCW MS CSIS Proceedings. V. 4 , N. 9 .