| | |
Summary: Compiler and Runtime Support for Static and Adaptive Irregular
Reductions on a Multithreaded Architecture
Gary Zoppetti Gagan Agrawal y
Department of Computer and Information Sciences
University of Delaware, Newark DE 19716
zoppetti@eecis.udel.edu
y Department of Computer and Information Sciences
Ohio State University, Columbus, OH 43210
agrawal@cis.ohio-state.edu
Abstract
Computations from many scienti c and engineering domains use irregular meshes and/or
sparse matrices. The codes expressing these computations involve irregular reductions. Irreg-
ular reductions can be static, in which the problem geometry does not change in the time-step
loop, or adaptive, in which the problem geometry does change. Both static and adaptive irreg-
ular reductions pose many challenges to parallel architectures and their compilers in terms of
parallelization, locality management, and communication optimization.
Multithreaded architectures o er rich support for local synchronization, overlapping of com-
munication and computation, and low overhead communication and thread switching. Therefore,
they appear to be promising for scalable parallelization of irregular reductions. We have devel-
oped an execution strategy for supporting irregular reductions on a ne-grained multithreaded
|
