Results 11 - 20 of 29,433

Table 1. NCSA O2K Space Shared Hosts

in Abstract Characteristics of a Large Shared Memory Production Workload
by Su-hui Chiang, Mary K. Vernon
"... In PAGE 8: ... Based on the curves in Figure 6(b), Figure 7 provides the measured conditional distributions of requested memory for four different ranges of the requested number of pro- cessors, which can be used to create memory requests that have the observed correlation with job parallelism. Recall from Table1 that the average memory available per processor on the O2K is either 256 MB or 512 MB, depending on the host. As shown in Figures 5(a) and 6(a), 35-40% of all jobs have normalized requested memory greater than 256 MB; furthermore 15-20% of all jobs and 50% of the sequential jobs have normalized requested memory greater than 0.... ..."

Table 1. NCSA O2K Space Shared Hosts

in Abstract Characteristics of a Large Shared Memory Production Workload
by Su-hui Chiang, Mary K. Vernon
"... In PAGE 8: ... Based on the curves in Figure 6(b), Figure 7 provides the measured conditional distributions of requested memory for four different ranges of the requested number of pro- cessors, which can be used to create memory requests that have the observed correlation with job parallelism. Recall from Table1 that the average memory available per processor on the O2K is either 256 MB or 512 MB, depending on the host. As shown in Figures 5(a) and 6(a), 35-40% of all jobs have normalized requested memory greater than 256 MB; furthermore 15-20% of all jobs and 50% of the sequential jobs have normalized requested memory greater than 0.... ..."

Table 4.1: RMS for different weights and k-space trajectories after 1, 2, 5, and 10 iterations.

in A note on the iterative MRI reconstructionfrom nonuniform k-space data
by Tobias Knopp, Stefan Kunis, Daniel Potts

Table 4.2: CPU-time and memory usage for different iteration numbers # and sizes N of the k-space data and the reconstructed phantom.

in A note on the iterative MRI reconstructionfrom nonuniform k-space data
by Tobias Knopp, Stefan Kunis, Daniel Potts

Table 1. Algorithm Complexity for N Contacts, L Wrench Extremes in the Contact Model, J Wrench Samples in the Contact Model (Possibly Including Interior Samples),H Half-Spaces inCHorig,K Wrench Extremes in theTask Model, S Samples on the Object Surface, and F Planar Faces in the Object Surface Current paper Compute epsilon1 O((NL)4 + HK)(including computing CHorig) Compute W(G,epsilon1)

in Closure and Quality Equivalence for Efficient Synthesis of Grasps from Examples
by Nancy S. Pollard
"... In PAGE 13: ...Referring to Figure 4, the grasp synthesis algorithm has three parts: (1) compute parameter epsilon1; (2) compute grasp family W(G,epsilon1); (3) map W onto the surface of a new object to ob- tain contact regions. Here we show that all parts are polyno- mial in the number of contacts N ( Table1 ) and we compare the complexity of this algorithm with the elegant competing technique of Zhu, Ding, and Li (2001). Computing epsilon1 is trivial unless we are preserving a force- based quality value as in Section 7.... ..."

Table 1. Algorithm Complexity for N Contacts, L Wrench Extremes in the Contact Model, J Wrench Samples in the Contact Model (Possibly Including Interior Samples),H Half-Spaces inCHorig,K Wrench Extremes in theTask Model, S Samples on the Object Surface, and F Planar Faces in the Object Surface Current paper Compute epsilon1 O((NL)4 + HK)(including computing CHorig) Compute W(G,epsilon1)

in Closure and Quality Equivalence for Efficient Synthesis of Grasps from Examples
by Nancy S. Pollard
"... In PAGE 14: ...Referring to Figure 4, the grasp synthesis algorithm has three parts: (1) compute parameter epsilon1; (2) compute grasp family W(G,epsilon1); (3) map W onto the surface of a new object to ob- tain contact regions. Here we show that all parts are polyno- mial in the number of contacts N ( Table1 ) and we compare the complexity of this algorithm with the elegant competing technique of Zhu, Ding, and Li (2001). Computing epsilon1 is trivial unless we are preserving a force- based quality value as in Section 7.... ..."

Table 1: All single features sfi in the feature space FS (C = 28 for audio files with 44.1 kHz sampling frequency)

in Digital Audio Forensics: A First Practical Evaluation on Microphone and Environment Classification
by Christian Kraetzer, Andrea Oermann, Jana Dittmann, Andreas Lang

Table 2: Tabulated performance data for the various molecular systems described in this paper using Blue Matter on BG/L. The total time per time-step and selected components are provided for both MPI and BG/L ADE SPI implementations. All of these data were taken in a dual core mode in which k-space and real-space operations are carried out on separate cores and used a complex to complex single precision 3D-FFT. The SOPE and Rhodopsin benchmarks used the velocity Verlet integrator and carried out the FFT-based P3ME operations on every time-step while the ApoA1 benchmark used RESPA[19] and only carried out P3ME on every fourth time- step. Performance is reported for two different processor mesh geometries at 4096 nodes because communication-intensive operations can be sensitive to the aspect ratio of the processor mesh.

in Blue Matter: Strong scaling of molecular dynamics on Blue Gene/L
by Blake G. Fitch, R Rayshubskiy Maria Eleftheriou, T. J. Christopher Ward, Mark Giampapa, Yuri Zhestkov, Michael C. Pitman, Frank Suits, Alan Grossfield, Jed Pitera, William Swope, Ruhong Zhou, Scott Feller, Robert S. Germain 2005
"... In PAGE 4: ... the context of Blue Matter of the time taken by the FFT as well by the neighborhood broadcast and reduce are provided in Table 2. 3 Performance Results We present detailed performance scaling results for three molecular systems whose sizes span the range from 10,000 to 100,000 atoms in Table2 for both MPI and BG/L ADE SPI communications protocols. Two of these systems, a small solvated lipid bilayer system, SOPE, and a solvated membrane protein system, Rhodopsin, have been used in published scientific work by the Blue Matter science team[5,23,24].... ..."
Cited by 4

Table 6 reports the percentage of cpu per platform re- quired to partition N GSM frequency channels (N 2 [1; 8]) at a sampling rate equals to 2.5 times the maximal fre- quency and with a 200kHz spacing scheme. The lter used in the implementation has 255 coe cients. The cor- responding SPEC performances are shown13 in Table 12, Section 5.1. The high CPU requirement14 is due to the high input data rate and the need to compute imaginary and real components of output samples for each frequency channel (requiring multiple oating point operations).

in Towards the Software Realization of a GSM Base Station
by Thierry Turletti Hans, Hans Bentzen, David Tennenhouse 1999
"... In PAGE 5: ... Table6 : Polyphase Transform performance (200 kHz spac- ing between channels) 4.2 Demodulation and Equalization The GSM speci cations do not impose a particular demod- ulation algorithm.... ..."
Cited by 4

Table 6 reports the percentage of cpu per platform re- quired to partition N GSM frequency channels (N 2 [1; 8]) at a sampling rate equals to 2.5 times the maximal fre- quency and with a 200kHz spacing scheme. The lter used in the implementation has 255 coe cients. The cor- responding SPEC performances are shown13 in Table 12, Section 5.1. The high CPU requirement14 is due to the high input data rate and the need to compute imaginary and real components of output samples for each frequency channel (requiring multiple oating point operations). Polyphase Transform Perf. (%cpu)

in Towards the Software Realization of a GSM Base Station
by Thierry Turletti, Hans Bentzen, David Tennenhouse 1999
"... In PAGE 5: ... Table6 : Polyphase Transform performance (200 kHz spac- ing between channels) 4.2 Demodulation and Equalization The GSM speci cations do not impose a particular demod- ulation algorithm.... ..."
Cited by 4
Results 11 - 20 of 29,433