(PEN)-ULTIMATE TILING

被引:56
作者
BOULET, P
DARTE, A
RISSET, T
ROBERT, Y
机构
[1] Laboratoire LIP-IMAG, CNRS URA 1398, Ecole Normale Supérieure de Lyon
关键词
D O I
10.1016/0167-9260(94)90019-1
中图分类号
TP3 [计算技术、计算机技术];
学科分类号
0812 ;
摘要
In the framework of perfect loop nests with uniform dependences, tiling is a technique used to group elemental computation points so as to increase computation granularity and to reduce the overhead due to communication time. We review existing approaches from the literature, together with the optimization criteria that are used for determining a ''good'' or ''optimal'' tiling. Then we explain the need to introduce yet another criterion for defining ''optimal tiling'' in a scalable environment. Although our criterion is more complex than previously used ones, we are able to prove a theorem on optimality, and to provide a constructive method for defining the ''optimal tiling''.
引用
收藏
页码:33 / 51
页数:19
相关论文
共 14 条
[1]  
Banerjee U., 1988, Journal of Supercomputing, V2, P133, DOI 10.1007/BF00128174
[2]  
IRIGOIN F, 1987, THESIS ERCOLE NATION
[3]  
LAM MS, 1992, 5TH P SIAM C PAR PRO, P537
[4]   PARALLEL EXECUTION OF DO LOOPS [J].
LAMPORT, L .
COMMUNICATIONS OF THE ACM, 1974, 17 (02) :83-93
[5]  
PEIR JK, 1986, UIUCDCSR861259 REP
[6]  
Ramanujam J., 1992, Proceedings. Supercomputing '92. (Cat. No.92CH3216-9), P214, DOI 10.1109/SUPERC.1992.236692
[7]  
RAMANUJAM J, 1990, AUG P INT C PAR PROC, P179
[8]  
RAMANUJAM J, 1992, 5TH P SIAM C PAR PRO, P543
[9]  
SCHREIBER R, 1990, 9038 TECHN REP
[10]  
Schrijver A., 1986, THEORY LINEAR INTEGE