摘要
Parallel loops account for the greatest amount of parallelism in numerical programs.Executing nested loops in parallel with low run-time overhead is thus very important for achieving high perform- ance in parallel processing systems.However,in parallel processing systems with caches or local memo- ries in memory hierarchies,“thrashing problem”may arise whenever data move back and forth between the caches or local memories in different processors. Previous techniques can only deal with the rather simple cases with one linear function in the perfect- ly nested loop.In this paper,we present a parallel program optimizing technique called hybrid loop inter- change(HLI)for the cases with multiple linear functions and loop carried data dependences in the nested loop.With HLI we can easily eliminate or reduce the thrashing phenomena without reducing the program parallelism.
Parallel loops account for the greatest amount of parallelism in numerical programs.Executing nested loops in parallel with low run-time overhead is thus very important for achieving high perform- ance in parallel processing systems.However,in parallel processing systems with caches or local memo- ries in memory hierarchies,“thrashing problem”may arise whenever data move back and forth between the caches or local memories in different processors. Previous techniques can only deal with the rather simple cases with one linear function in the perfect- ly nested loop.In this paper,we present a parallel program optimizing technique called hybrid loop inter- change(HLI)for the cases with multiple linear functions and loop carried data dependences in the nested loop.With HLI we can easily eliminate or reduce the thrashing phenomena without reducing the program parallelism.
