A framework for resource-constrained rate-optimal software pipelining

被引:20
作者
Govindarajan, R
Altman, ER
Gao, GR
机构
[1] INDIAN INST SCI, DEPT COMP SCI & AUTOMAT, BANGALORE 560012, KARNATAKA, INDIA
[2] IBM CORP, THOMAS J WATSON RES CTR, YORKTOWN HTS, NY 10598 USA
[3] MCGILL UNIV, SCH COMP SCI, MONTREAL, PQ H3A 2A7, CANADA
关键词
instruction-level parallelism; instruction scheduling; integer linear programming; software pipelining; superscalar and VLIW architectures;
D O I
10.1109/71.544355
中图分类号
TP301 [理论、方法];
学科分类号
081202 ;
摘要
The rapid advances in high-performance computer architecture and compilation techniques provide both challenges and opportunities to exploit the rich solution space of software pipelined loop schedules. In this paper, we develop a framework to construct a software pipelined loop schedule which runs on the given architecture (with a fixed number of processor resources) at the maximum possible iteration rate (a la rate-optimal) while minimizing the number of buffers-a close approximation to minimizing the number of registers. The main contributions of this paper are: First, we demonstrate that such problem can be described by a simple mathematical formulation with precise optimization objectives under a periodic linear scheduling framework. The mathematical formulation provides a clear picture which permits one to visualize the overall solution space (for rate-optimal schedules) under different sets of constraints. Secondly, we show that a precise mathematical formulation and its solution does make a significant performance difference. We evaluated the performance of our method against three leading contemporary heuristic methods. Experimental results show that the method described in this paper performed significantly better than these methods. The techniques proposed in this paper are useful in two different ways: 1) As a compiler option which can be used in generating faster schedules for performance-critical loops (if the interested users are willing to trade the cost of longer compile time with faster runtime). 2) As a framework for compiler writers to evaluate and improve other heuristics-based approaches by providing quantitative information as to where and how much their heuristic methods could be further improved.
引用
收藏
页码:1133 / 1149
页数:17
相关论文
共 43 条
[1]  
AIKEN A, 1991, RES MG PAR, P274
[2]  
AIKEN A, 1988, 88922 CORN U DEP COM
[3]  
AIKEN A, 1988, P SIGPLAN 88 C PROGR, V23
[4]  
Allen J. R., 1983, P 10 ACM SIGACT SIGP, P177
[5]  
ALTMAN ER, 1995, P ACM SIGPLAN 95 C P, V30, P139
[6]  
ALTMAN ER, 1995, THESIS MCGILL U MONT
[7]   COMPILING FOR THE CYDRA-5 [J].
DEHNERT, JC ;
TOWLE, RA .
JOURNAL OF SUPERCOMPUTING, 1993, 7 (1-2) :181-227
[8]  
DEHNERT JC, 1989, OPERATING SYSTEMS RE, V23
[9]  
DEHNERT JC, 1989, SIGPLAN NOTICES, V24
[10]  
DEHNERT JC, 1989, P 3 INT C ARCH SUPP, V17, P26