Tools
Sorted by:
Try your query at:
 |
 |
 |
 |
 |
 |
Results 1 - 10
of
1,126
Table 5: Main PHY specifications of very high speed radio access at 17 GHz
"... In PAGE 30: ... The MVC address is assigned as in Table 5. Table5 : MVC address range use. Address Use 0 MASCARA MPDUs 1-2 Assigned to (default) ATM signalling VCs 3-254 Assigned to user plane SVCs (or PVCs) 255 Reserved Table 6 presents a short summary on different addresses used in the WAND system.... In PAGE 50: ... Both scenarios will require the application of directive antennas because of the higher bandwidth, lower allowed power and higher path loss. Table5 summarizes the main specifications and the differences of the 17 GHz very high speed radio interface compared to the WAND demonstrator. Table 5: Main PHY specifications of very high speed radio access at 17 GHz... ..."
Table 17: Performance of the very high speed switch with input and output bu#0Bers. Cell
Table 18: Performance of the very high speed switch with input and crosspoint bu#0Bers.
Table 7: Comparison of interconnection network graphs One of the trade-o s of xed-degree graphs is an increased diameter. However, the n-SCC can be built with very high speed buses in the local links. Therefore, the n-SCC can present communication delays comparable to the n-star, if we consider that the lateral links often use serial links for making their lay-out simpler. In addition, many practical algorithms present locality of operation and require just a limited region of the interconnection network to run. This reduces even more the requirements for long communication paths in the graph and contributes to high performance in parallel computers. Another disadvantage of xed-degree graphs is a reduced fault tolerance in comparison to variable- degree graphs. The fault tolerance of the n-SCC graph is 2, while the fault tolerance of the n-star and the n-cube is respectively equal to (n ? 2) and (n ? 1). However, since the underlying topology 33
"... In PAGE 35: ... Such algorithm is actually an extension of the routing algorithm presented in this report and can be based on dynamic fault-tolerant routing and broadcasting algorithms already developed for the n-star [3], [16], [17] and the cube-connected cycles [18]. Table7 also shows another type of I/O-bounded interconnection network, namely the cube- connected cycles or CCC. An n-CCC graph can be built by replacing each node of an n-cube with a ring of n or more nodes.... In PAGE 35: ... An n-CCC graph can be built by replacing each node of an n-cube with a ring of n or more nodes. Table7 shows typical values for n-CCC graphs containing n nodes in each ring. The number of nodes and diameter of an n-CCC graph formed under such structure are given respectively by N = n2n and dCCC = 2n + bn=2c ? 2 [19].... In PAGE 35: ...-SCC. However, it has been shown that the n-star outperforms the n-cube on these aspects [1], [3]. If we recall that the n-SCC not only uses the n-star as its quotient Cayley graph but also has fewer nodes in each ring, then it seems that we should expect favorable results when compared with the CCC. Table7 also compares the n-SCC with other graphs from the viewpoint of one-port broadcasting algorithms. Theorem 11 shows that the selection of a proper value for the ratio of the transmission rates in the local and lateral links of an n-SCC graph can reduce the running time of a broadcasting algorithm to about twice the time spent in the lateral link steps.... In PAGE 35: ... This results in an O(n) running time, while a one-port broadcasting in an n-star has a O(n log n) running time. The values shown for one-port broadcasting in the n-star in Table7 are optimal and have been extracted from [15]. By inspecting Table 7, we notice that the one-port broadcasting in an n-SCC graph can be accomplished with running time better than or equal to that of an n-star containing (n ? 1) times fewer nodes.... In PAGE 35: ... The values shown for one-port broadcasting in the n-star in Table 7 are optimal and have been extracted from [15]. By inspecting Table7 , we notice that the one-port broadcasting in an n-SCC graph can be accomplished with running time better than or equal to that of an n-star containing (n ? 1) times fewer nodes. For n = 4 and n = 6, the total number of steps required by the one-port broadcasting algorithm is about 50% greater than the diameter of the corresponding SCC graph.... In PAGE 36: ... Since the diameter of an n-star is less than that of a hypercube of similar size, we simply extend this result to conclude that broadcasting in the n-SCC graph can be achieved in shorter running time than in a CCC of similar size. As a matter of fact, an inspection of Table7 allows us to con rm that.... ..."
Table 1: Prediction Modes for Lossless JPEG (a is left-neighboring pixel, b is upper- neighboring pixel, c is upper-left-neighboring pixel) [11] Rice, R.F., Yeh, P.S., and Miller, W.H. \Algorithms for a Very High Speed Univer- sal Noiseless Coding Module. quot; JPL Publication 91-1. Pasadena, CA: JPL Publication O ce. February 15, 1991. [12] Venbrux, Jack, Yeh, Pen-Shu, and Liu, Muye N. \A VLSI Chip Set for High-Speed Lossless Data Compression. quot; IEEE Transactions on Circuits and Systems for Video Technology. Vol. 2. No. 4. December, 1992. pp. 381-391. [13] Arps, R.B. and Truong, T.K. \Comparison of International Standards for Lossless Still Image Compression. quot; Proceedings of the IEEE. Vol. 82. No. 6. June, 1994. pp. 889-899. [14] Memon, N.D. and Sayood, K. \Lossless Image Compression: A Comparative Study. quot; IS amp;T/SPIE Electronic Imaging Conference. San Jose, CA. February, 1995.
"... In PAGE 6: ...recision of these coe cients could be reduced by a factor of 1000 (i.e. three decimal digits) without a ecting the rst-order entropy of the error residuals; the advantage of this is a reduction in rst-order entropy of the DCT coe cients. The proposed method was compared to seven xed lters for the present lossless JPEG standard (given in Table1 ) [10], with the rst-order entropy of the error residuals for both methods given Table 2, where the proposed... ..."
Table 3. Employed Machine Models. In Table 1 and Table 2 we presented those two measurements of our experiments with the maximum number of machines. Since a network of several hundred machines permanently changes its con guration, it is impossible to get a xed number of machines with reproducable load situation for reference measurements. We had hands on about 1,000 machines but typically about 20% of them were shut down. Most of the LANs we used had processor idle percentages of at least 90% of the time when our parallel application did not run. Such values are very common as we also con rmed by logging CPU idle percentage data in our LAN for a period of several months. This average utilization remains almost constant from our experience, although the utilization of individual workstations varies permanently. However, some LANs with high speed workstations were heavily loaded with parallel cluster applications.
1993
"... In PAGE 8: ... Table3 shows the di erent machine models which have been incorporated in our con guration. There have been various di erent types of most machine models and even more di erent UNIX operating system versions.... ..."
Cited by 11
Table 18: Impact of Lost Cap-Ex (in 000s) from Section 5 simulation, Entire Sample
2007
"... In PAGE 37: ... For example, cable operators could be well-positioned to offer triple-play packages that support HDTV and very high-speed broadband if they upgrade their networks to the latest DOCSIS standard. This in turn would spur the incumbent (or indeed entrants who believe 37 JP Morgan, The Fibre Battle , December 4, 2006, Table18... ..."
Table 4. Encryption speed for 128-bit keys divided by the memory requirements for
"... In PAGE 14: ... The larger the value, the greater the indication that the algorithm may be more suitable for DSPs. Once again #28see Table4 #29 RC6 compares very well to the other #0Cnalists. Note further the exceptional advantages that RC6 o#0Bers in multi-block applications and therefore in very high-speed, streaming applications.... ..."
Table 5: MVC address range use.
"... In PAGE 50: ... Both scenarios will require the application of directive antennas because of the higher bandwidth, lower allowed power and higher path loss. Table5 summarizes the main specifications and the differences of the 17 GHz very high speed radio interface compared to the WAND demonstrator. Table 5: Main PHY specifications of very high speed radio access at 17 GHz 5 GHz (WAND demonstrator) 17 GHz High speed access Antennas omni Directional Coverage range 50 m 100 m Bit rate 20 Mbit/s 155 Mbit/s Modulation OFDM with 16 carriers OFDM with 64 carriers 8-PSK on each carrier 4-PSK * on each carrier Frequency range 5.... In PAGE 50: ... Table 5 summarizes the main specifications and the differences of the 17 GHz very high speed radio interface compared to the WAND demonstrator. Table5 : Main PHY specifications of very high speed radio access at 17 GHz 5 GHz (WAND demonstrator) 17 GHz High speed access Antennas omni Directional Coverage range 50 m 100 m Bit rate 20 Mbit/s 155 Mbit/s Modulation OFDM with 16 carriers OFDM with 64 carriers 8-PSK on each carrier 4-PSK * on each carrier Frequency range 5.15 - 5.... ..."
Table 2: calculation time for preprocessing and rendering
"... In PAGE 12: ...As we can see from Table2 preprocessing allows rendering individual copies at very high speed. The preprocessing stage is carried out about real time while the individually watermarked copies are rendered at the speed many times above real time: Rendering performs with the speed comparable to current CD-R burning facilities at 56x.... ..."
Results 1 - 10
of
1,126
