Multipass communication systems for tiled processor architectures
Name
72692209-MIT.pdf
Description
Full printable version
Size
30.54 MB
Format
Adobe PDF
Checksum (MD5)
fe2e030a41631e5465f9b7c9e6b8b960
Author(s)
Shnidman, Nathan R. (Nathan Robert)
Advisor(s)
Anant Agarwal.
Date Issued
2006
Publisher
Massachusetts Institute of Technology
Abstract
Multipass communication systems utilize multiple sets of parallel baseband receiver functions to balance communication data rates and available computation capabilities. This is achieved by spatially pipelining baseband functions across parallel resources to perform multiple processing passes on the same set of received values, thus allowing the system to simultaneously convey multiple sequences of data using a single wireless link. The use of multiple passes mitigates the effects of data rate on receiver processing bottlenecks, making the use of general-purpose processing elements for high data rate communication functions viable. The flexibility of general-purpose processing, in turn, allows the receiver composition to trade-off resource usage and required processing rate. For instance, a communication system could be distributed across 2 passes using 2x the overall area, but reducing the data rate for each pass and the resultant overall required processing rate, and hence clock speed, by 1/2. Lowering the clock speed can also be leveraged to reduce power through voltage scaling and/or the use of higher Vt devices. The characteristics of general-purpose parallel processors for communications processing are explored, as well as the applicability of specific parallel designs to communications processing.
(Cont.) In particular, an in depth look is taken of the Raw processor's tiled architecture as a general-purpose parallel processor particularly well suited to portable communications processing. An example of a multipass system, based on the 802.11a baseband, implemented on the Raw processor along with the accompanying hardware implementation is presented as both a proof-of-concept, as well as a means to explore some of the advantages and trade-offs of such a system. A bit-error rate study is presented which shows this multipass system to be within a small fraction of dB of the performance of an equivalent data rate single pass system, thus demonstrating the viability of the multipass algorithm. In addition, the capability of tiled processors to maximize processing capabilities at the system block level, as well as the system architectural level, is shown. Parallel implementations of two processing intensive functions: the FFT and the Viterbi decoder are shown. A parallelized assembly language FFT utilizing 16 tiles is shown to have a 1,000x improvement , and a parallelized 48-tile assembly language Viterbi decoder is shown to have a 10, 000x improvement over corresponding serial C implementations.
Description
Thesis (Ph. D.)--Massachusetts Institute of Technology, Dept. of Electrical Engineering and Computer Science, 2006.
Includes bibliographical references (p. 191-202).
Subjects
Electrical Engineering and Computer Science.
MIT Department
Massachusetts Institute of Technology. Department of Electrical Engineering and Computer Science
Terms of Use
M.I.T. theses are protected by copyright. They may be viewed from this source for any purpose, but reproduction or distribution in any format is prohibited without written permission. See provided URL for inquiries about permission.
Persistent DSpace Link
