Mapping multirate multidimensional Digital Signal Processing systems to efficient computer architectural models by multidimensional unfolding
被引:0
作者:
Peng, DM
论文数: 0引用数: 0
h-index: 0
机构:
Univ Nebraska, Dept Elect & Comp Engn, Lincoln, NE USAUniv Nebraska, Dept Elect & Comp Engn, Lincoln, NE USA
Peng, DM
[1
]
Sharif, H
论文数: 0引用数: 0
h-index: 0
机构:
Univ Nebraska, Dept Elect & Comp Engn, Lincoln, NE USAUniv Nebraska, Dept Elect & Comp Engn, Lincoln, NE USA
Sharif, H
[1
]
机构:
[1] Univ Nebraska, Dept Elect & Comp Engn, Lincoln, NE USA
来源:
PARALLEL AND DISTRIBUTED COMPUTING SYSTEMS, PROCEEDINGS
|
2003年
关键词:
D O I:
暂无
中图分类号:
TP3 [计算技术、计算机技术];
学科分类号:
0812 ;
摘要:
A direct mapping of multirate Digital Signal Processing (DSP) algorithms to hardware would require data to move at different rates in circuits on the chip, which involves low hardware utilization, complicate routing and synchronization of multiple clock signals. This paper contributes to mapping multidimensional multirate DSP algorithms to uniform-rate special-purpose efficient computer architectural models. Based on the analysis of the reduction of the translated MultiRate Multidimensional Data Flow Graphs, we map multirate multidimensional DSP algorithms to practical architectural models with an extended Push-Up scheduling method.