A Nested Loop Fusion Algorithm Based on Cost Analysis

When applying Loop Fusion technology in the existing parallelizing compiler systems, the combined loop may not gain better parallel efficiency due to the lack of the consideration to the cost analysis of parallel loops. By improving the execution process of parallelizing compiler systems, we proposed a nested loop fusion algorithm based on cost analysis. It can not only optimize the execution process of parallelizing compiler system but also ensure that the combined loop having been applied Loop Fusion technology can get positive speedup. We implemented this algorithm on a specific parallelizing compiler and compiled some benchmark programs by the compiler. The experimental results show that the efficiency of the parallel programs increases by 5% to 20%.