11. |
Novel approach for implementing convolutions with small tables |
|
IEE Proceedings E (Computers and Digital Techniques),
Volume 138,
Issue 4,
1991,
Page 255-259
A.Skavantzos,
Preview
|
PDF (570KB)
|
|
摘要:
We present a new algorithm, ‘the one over eight squared algorithm’, which can be useful in computing tasks with convolutional complexity, such as convolutions, correlations and complex multiplications. The new algorithm reduces the address space requirement and hardware investment for table lookup implementations of such tasks from 22nto 2nwherenis the width of the operands. Owing to considerable reduction in the required table size, convolutions and correlations with table lookups can be cost-effectively extended to wider operands.
DOI:10.1049/ip-e.1991.0033
出版商:IEE
年代:1991
数据来源: IET
|
12. |
Families of sequences and arrays with good periodic correlation properties |
|
IEE Proceedings E (Computers and Digital Techniques),
Volume 138,
Issue 4,
1991,
Page 260-268
D.H.Green,
S.K.Amarasinghe,
Preview
|
PDF (1120KB)
|
|
摘要:
Families of binary sequences with good periodic correlation properties and large linear span are described which are derived from certain multiple-valued linear feedback shift registers with composite characteristic polynomials. Multiple-valued pseudorandom arrays constructed from parentm-sequences are used to describe and analyse the formation and properties of these families. A general construction is presented which includes a number of well known families as special cases. A simple method for synthesising sequence generation is also indicated.
DOI:10.1049/ip-e.1991.0034
出版商:IEE
年代:1991
数据来源: IET
|
13. |
New protocol for multistage interconnection networks |
|
IEE Proceedings E (Computers and Digital Techniques),
Volume 138,
Issue 4,
1991,
Page 269-275
N.M.Patel,
Preview
|
PDF (1007KB)
|
|
摘要:
Multistage interconnection networks (MINs) are a cost-effective way of providing interprocessor communication in multiprocessor systems. Many MIN protocols have been studied in the literature but all are known to suffer from the ‘hot-spot’ effect which causes network throughput to drop when a particular output pin becomes more heavily utilised (hotter) than the other pins. The paper outlines a protocol which is designed to be less susceptible to such effects. The new protocol is a variant of the circuit switching class and simulation results indicate that it gives better overall performance than other protocols in the same class.
DOI:10.1049/ip-e.1991.0035
出版商:IEE
年代:1991
数据来源: IET
|
14. |
Fault-tolerant serial-parallel multiplier |
|
IEE Proceedings E (Computers and Digital Techniques),
Volume 138,
Issue 4,
1991,
Page 276-280
L.G.Chen,
T.H.Chen,
Preview
|
PDF (653KB)
|
|
摘要:
The paper presents a novel fault-tolerant circuit design using a time-redundancy method for a serial-parallel multiplier, which is useful in DSP applications with serial data transmission. It utilises the RECO (REcomputing with Circularly shifted Operands) technique to detect errors concurrently. A simple OR-gate based circuit is used as the location table to identify faulty bit-slice pairs. The reconfiguration technique is then introduced to bypass the potential faulty bit-slices. This design can have the maximum detectable error region ( n/2 bits), without appending extra computing elements. The latency from error detection to location is only about two clock cycles, i.e. almost real-time detecting can be achieved. Pipe-lined timing for two computations is illustrated. The analyses of performance and complexity are described. The results show that this is an efficient design methodology for fault-tolerant multiplication with serial data.
DOI:10.1049/ip-e.1991.0036
出版商:IEE
年代:1991
数据来源: IET
|
15. |
Multiple bus network with overlapping connectivity |
|
IEE Proceedings E (Computers and Digital Techniques),
Volume 138,
Issue 4,
1991,
Page 281-284
B.Wilkinson,
Preview
|
PDF (506KB)
|
|
摘要:
A multiple bus multiprocessor network with overlapping connectivity is presented. An overlapping connectivity network allows each processor to directly reach a group of memory modules. The groups of memory modules of adjacent processors overlap. Using simulation, it is demonstrated that the bandwidth of the over lapping connectivity multiple bus network significantly exceeds that of a conventional multiple bus network when the memory modules requested are close.
DOI:10.1049/ip-e.1991.0037
出版商:IEE
年代:1991
数据来源: IET
|
16. |
Multipeak histogram analysis in region splitting: a regularisation problem |
|
IEE Proceedings E (Computers and Digital Techniques),
Volume 138,
Issue 4,
1991,
Page 285-288
SrimantaPal,
PritimoyBhattacharyya,
Preview
|
PDF (381KB)
|
|
摘要:
The paper describes an application of regularisation to the problem of estimating the cutoff threshold parameter for splitting the regions within an image by using its multimodal histogram. The analysis involves generating a multiple description of a histogram by smoothing it with cubic splines, locating peaks and valleys from it, and finally selecting the cutoff threshold by computing a score.
DOI:10.1049/ip-e.1991.0038
出版商:IEE
年代:1991
数据来源: IET
|