### Table 1: Architectural attributes of MIMD parallel computers

"... In PAGE 9: ...operational in November 1999. Table1 gives the definitions and values of the radar parameters and Table 2 shows the processor parameters. Table 1: Radar parameters Table 2: Processor parameters Figure 1 shows the block diagram of the GeoSAR range-Doppler signal processing and Figure 2 gives the numbers of floating-point operations per input sample at each processing stage ... In PAGE 9: ... Table 1 gives the definitions and values of the radar parameters and Table 2 shows the processor parameters. Table1 : Radar parameters Table 2: Processor parameters Figure 1 shows the block diagram of the GeoSAR range-Doppler signal processing and Figure 2 gives the numbers of floating-point operations per input sample at each processing stage ... In PAGE 21: ... Because of its centralized shared memory, the UMA model may limit scalability once built. Table1 compares the architectural attributes and performance of these five parallel architectures. ... ..."

### Table 1: Related Work. For the paradigm, S and M denote SPMD and MIMD, resp. For the architecture, S and D denote shared memory and distributed, respectively.

### Table 1 Characteristic parameters of MIMD computers

1998

"... In PAGE 16: ... These nodes run at 160 Mhz and are capable of a peak performance of 640 MFLOPS each. Our measurements of communication times (b0 and b1) for the CRAY T3E and the IBM SP computers are presented in Table1 . The maximum length of string is 1000 words.... ..."

Cited by 4

### Table 1 contains a summary of the performance of the partitioning strategies in the various cost models. The values are exact only if k divides n exactly and k has an integral square root which divides n exactly. It is clear that although the lines partition fares particularly badly in having a larger number of communications events and delays than either of the other partitions, if k2 gt; (k ? 1)n, it performs, in the time model, the best out of the three partitions. This means that the partition which has, out of the three, the most communication performs the best, and this result holds for any reasonable value of communications delay. The programmer who is faced with the problem of explicitly partitioning a longest common subsequence computation, should therefore make di erent qualitative decisions on the partition, and therefore write di erent programs, according to which cost model best approximates his architecture.

1992

"... In PAGE 9: ...t c d M Relevant Model event amp; delay event delay time Boxes n2=pk 2n(pk ? 1) 2(pk ? 1) n2=pk + 2(pk ? 1) Stripes n2=k n(k ? 1) k ? 1 n2=k + (k ? 1)n=k + (k ? 1) Lines n2=k n2 ? n n ? 1 n2=k + k + (k ? 1) Table1 : Summary of the Costs of the Partitions We have not discussed the relative appropriateness of event, delay and time costing, but in systems such as Distributed Memory MIMD architectures, as long as the locality of the computation can be matched to the locality of the architecture, the available communications bandwidth scales up linearly with the number of processors. In this case it need not cost to have a large number of communications events, nor need it cost if there exist long chains of communications events.... ..."

Cited by 1

### Table 2: As for Table 1, but for message passing implementations of parallel random number gen- erators on MIMD parallel computers.

1998

"... In PAGE 7: ... In some cases we have also tested the generators using alternate lattice sizes (64 64 and 256 256), which can probe for correlations at di erent scales. The results for the parallel tests are shown in Table 1 for the data parallel (SIMD) results and Table2 for the message passing (MIMD) results. One point to note from the results is that good initialization (or seeding) of parallel random number generators is crucial to their performance, particularly for parallel lagged Fibonacci genera- tors, where many seeds need to be assigned on each processor.... ..."

Cited by 2

### Table 15: Poverty and Inequality Group Weights Welfare Theil P0 P1 P2 P0*

"... In PAGE 47: ... The second (P0*) corresponds to the computation of the poverty rate under the standard assumption of a lognormal distribution of the within-group income, with endogenous mean and fixed variance. Table15 gives a static image of the differences between the two measures. At the aggregated level, P0* underestimates the poverty rate, but the results differ according to groups.... ..."

Cited by 1

### Table 2: Computational results for the warehouse problem (weak 0-1 model)

### TABLE 3. Performance of neural network implementations on workstations, parallel MIMD/ SIMD computers and dedicated neural network hardware (Adaptive Solution CNAPS).

1995

"... In PAGE 5: ... Cray T3E performance, communication overhead and scaleup measured with a TDNN network (4 layer, 4680 weights) used for promoter site detection, Backpropagation limit off-line learning algorithm and 3157 training patterns. System Software Performance [MCUPS] Comments TABLE3 . Performance of neural network implementations on workstations, parallel MIMD/... ..."

Cited by 1

### Table 2 Correlation coefficients (Pearson apos;s r) between MT+ contrast responsivity and behavioral measures

2007

"... In PAGE 3: ... 5). Given the considerable overlap between subtests included in these neuropsychological batteries, we relied primarily on 8 composite scores that span various aspects of reading and cognitive processing (see Table2 and Supplementary Table 1). We rely on the composite scores developed and validated by the authors of the test batteries, which have higher reliability and external validity than the component scores.... ..."

### Table 1 Summary of the local Grid deployment and distribution of the simulations in the testbed. Machine Computational Nodes Memory Avail. Nodes Simulations

"... In PAGE 6: ...the machines except Ramses, which runs the LCG-2 middleware and is part of the Biomed Virtual Organisation which will be explained later on. Table1 summarises the testbed (i.e.... ..."