TY - GEN

T1 - Macro tree transformations of linear size increase achieve cost-optimal parallelism

AU - Morihata, Akimasa

PY - 2011/12/26

Y1 - 2011/12/26

N2 - This paper studies parallel evaluation of tree transformations, in particular accumulative ones. Accumulation is a ubiquitous programming pattern. However, since accumulation usually imposes restrictions on evaluation orders, accumulative tree transformations appear to be unsuitable for parallel evaluation. We propose a parallel evaluation method for a large class of tree-to-tree recursive functions, which may contain accumulations, higher-order terms, and function compositions. Our parallel evaluation method achieves optimal parallel speedup if the transformation is of linear size increase, namely, the size of each output is linearly bounded by the size of the corresponding input. Our result is based on the theory of macro tree transducers and that of parallel tree contractions. The main contribution is to reveal a good collaboration between them.

AB - This paper studies parallel evaluation of tree transformations, in particular accumulative ones. Accumulation is a ubiquitous programming pattern. However, since accumulation usually imposes restrictions on evaluation orders, accumulative tree transformations appear to be unsuitable for parallel evaluation. We propose a parallel evaluation method for a large class of tree-to-tree recursive functions, which may contain accumulations, higher-order terms, and function compositions. Our parallel evaluation method achieves optimal parallel speedup if the transformation is of linear size increase, namely, the size of each output is linearly bounded by the size of the corresponding input. Our result is based on the theory of macro tree transducers and that of parallel tree contractions. The main contribution is to reveal a good collaboration between them.

UR - http://www.scopus.com/inward/record.url?scp=84055193137&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84055193137&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-25318-8_17

DO - 10.1007/978-3-642-25318-8_17

M3 - Conference contribution

AN - SCOPUS:84055193137

SN - 9783642253171

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 204

EP - 219

BT - Programming Languages and Systems - 9th Asian Symposium, APLAS 2011, Proceedings

T2 - 9th Asian Symposium on Programming Languages and Systems, APLAS 2011

Y2 - 5 December 2011 through 7 December 2011

ER -