2026-09-19T05:03:23Z https://www.ideals.illinois.edu/oai-pmh

oai:www.ideals.illinois.edu:2142/23253 2023-07-10 col_2142_5131 col_2142_8888 com_2142_5130 com_2142_8887 com_2142_234

Hornick, Scot W. 2011-05-07T14:07:35Z 2011-05-07T14:07:35Z 10000-01-01 1989 This thesis considers the mesh of trees architecture as both a special-purpose and a general-purpose parallel computer. A family of special-purpose VLSI architectures for computing an ($n\sb1 \times n\sb2 \times \cdots \times n\sb{d}$)-point multidimensional DFT over $\doubz\sb{M}$, the ring of integers modulo $M$, is proposed. Using the two-dimensional mesh of trees as a component, these architectures achieve optimal VLSI area $A$ = $\Theta((N\sp2\log\sp{2}M)/T\sp2)$ for any given computation time $T\ \epsilon$ ($\Omega(\log N),O(\sqrt{N\log M})\rbrack.$ The convergence properties of Newton's method are studied. By introducing and formalizing the notion of attunement of a linear system of equations, it is shown that Newton's method provides polylog-time solutions for a broader class of linear systems than was previously supposed. In particular, the system matrix need not be well-conditioned; all that is required is that the known vector be well-attuned to the system matrix. It is then shown that Newton's method can be implemented on a special-purpose architecture based on the three-dimensional mesh of trees. This same architecture can be used to construct the stiffness equations arising from a finite element approximation. Furthermore, it can be hybridized with a systolic array to achieve a processor-time or area-time tradeoff. Then, in a different vein, the two-dimensional mesh of trees is studied as a general-purpose parallel computer. It is shown that this architecture can afford finer memory granularity and, thereby, reduce the memory redundancy required for deterministic P-RAM simulation. A distributed-memory, bounded-degree network model of parallel computation is proposed that allows one to take greater advantage of the potential for fine-grain memories without sacrificing other aspects of realism. The simulation scheme presented is admitted by this new model and achieves constant memory redundancy. Made available in DSpace on 2011-05-07T14:07:35Z (GMT). No. of bitstreams: 2 license.txt: 4922 bytes, checksum: 910b249b4beec47e7ab768910c8f966f (MD5) 8924842.pdf: 4069878 bytes, checksum: a429fd5e69e455130261aa6590dc275a (MD5) Previous issue date: 1989 Item marked as restricted to the 'UIUC Users [automated]' Group (id=2) by Howard Ding (hding2@illinois.edu) on 2011-05-07T15:03:13Z Item is restricted indefinitely. Restriction data tranferred 2014-07-01T11:30:07-05:00 Original Data Group with Access UIUC Users [automated] Release Date: none Reason: ETDs are only available to UIUC Users without author permission ETDs are only available to UIUC Users without author permission U of I Only AAI8924842 (UMI)AAI8924842 http://hdl.handle.net/2142/23253 eng Copyright 1989 Hornick, Scot Wayne Engineering, Electronics and Electrical Computer Science The mesh of trees architecture for parallel computation text Electrical and Computer Engineering Electrical and Computer Engineering University of Illinois at Urbana-Champaign Dissertation Ph.D.