| | |
Summary: Minimum Register Instruction Sequence Problem:
Revisiting Optimal Code Generation for DAGs
R. Govindarajan y, H.Yang z, J. N. Amaral ], C. Zhang z, G. R. Gao z
y Supercomputer Education & Research Centre z Electrical and Computer Engineering ] Dept. of Computing Science
Dept. of Computer Science & Automation University of Delaware University of Alberta
Indian Institute of Science Newark, DE 19716 Edmonton T6G 2E8
Bangalore, 560 012, INDIA Delaware, U.S.A. Alberta, CANADA
govind@csa.iisc.ernet.in fhyang,czhang,ggaog@capsl.udel.edu amaral@cs.ualberta.ca
Abstract
We revisit the optimal code generation or evaluation or
der determination problem --- the problem of generating an
instruction sequence from a data dependence graph (DDG).
In particular, we are interested in generating an instruction
sequence S that is optimal in terms of the number of regis
ters used by the sequence S. We call this MRIS (Minimum
Register Instruction Sequence) problem.
We developed an efficient heuristic solution for the MRIS
problem based on the notion of instruction lineage. This so
lution facilitates the sharing of registers among instructions
within a lineage and across lineages by exploiting the struc
|
