### Table 3. Parallel runtime (seconds). Old or New solver means using the serial or parallel symbolic factorization algorithm, respectively.

"... In PAGE 3: ... Now, the numerical phase consumes more memory. Table3 shows the runtime of the parallel symbolic algorithm, and that reasonable speedups are obtained. One remark worth noting is that the fill quality of ParMeTiS is worsening with increasing processor count.... In PAGE 5: ... The overall operation count is reduced from O(n3) operations to O(n2k) for 2D problems, where k is tolerance-dependent constant. Table3 compares the performance of traditional multifrontal (MF) solver to that of the superfast multifrontal (fast MF) solver. As can be seen, with increasing problem size, the superfast solver can be more than three times faster than the traditional one.... ..."

### Table 5: Maximum achievable e#0Eciency of the parallel multifrontal algorithm due to load imbalance.

1994

"... In PAGE 15: ... Because each nCUBE 2 processor has a total of 16MB of memory,we performed the simulation on a Sun Sparc-2 workstation. Table5 shows these e#0Eciencies for the test problems of Table 2. For 16 processors, for most of the problems, MD produced orderings whose e#0Eciency is comparable to the e#0Eciency achieved Name MD MND KLND SND woodw 3,461,499 4,301,889 3,657,900 5,170,396 cycle 5,713,494 9,015,082 4,398,097 6,468,351 ken-11 4,074,171 4,651,280 10,620,980 8,962,257 d2q06c 29,696,436 34,553,561 14,911,941 12,866,482 pilot 42,380,243 57,822,291 67,974,968 58,684,694 gosh 51,583,850 63,173,750 76,440,590 64,244,883 pilot87 195,776,414 244,039,517 214,215,718 263,492,967 pds-06 208,223,868 302,185,761 495,597,665 1,192,302,692 cre-d 297,383,390 288,993,512 312,874,063 498,409,786 maros-r7 557,860,867 1,068,417,963 691,770,450 675,704,811 Table 4: The number of operations performed during the Cholesky factorization for four ordering algorithms.... ..."

Cited by 10

### Table 15: Software environment characteristics of the massively parallel systems involved in the Perfect BenchmarksTM e ort.

1993

Cited by 1

### Table 1. Distribution of tag abundance from massively parallel sequencing signature (MPSS) 559

2004

"... In PAGE 3: ..., 2002) and 84 given by: 85 + = x x x f 1 2 exp ) ( 2 (1) 86 87 where x is the base-10 logarithm of tag abundance in transcripts per million (tpm). 88 89 Table1 illustrates this phenomenon with the distribution of tag abundance and 90 microarray noise. This coincidence provides further evidence of the existence of a universal 91 distribution associated with gene expression (Ueda et al.... In PAGE 4: ...replacements from the available transcripts, and the total number of genes that are being 121 captured is then recorded. 122 123 We urge the reader to recall Table1 where the similarity between the distribution of 124 tag abundance from MPSS and that of microarray noise is shown. The former originates from 125 Jongeneel et al.... ..."

### Table 3. Speedup Parallel Solver compared with EISPACK

1995

"... In PAGE 11: ...Table 3. Speedup Parallel Solver compared with EISPACK Table3 contains the speedup of the presented algorithm as compared to the corre- sponding sequence of EISPACK routines where the rst ten eigenpairs are computed. n is the dimension of the problem and p is the number of processors.... ..."

Cited by 10

### Table 7: Available values of solver for the hypre solvers.

2006

"... In PAGE 31: ....3.4.3.2 hypre hypre [10][11] is a package of parallel iterative solvers and preconditioners from Lawrence Livermore National Laboratories. The hypre solvers are listed in Table7 and preconditioners are listed in Table 8. Note that the BoomerAMG solver does not use a preconditioner, and the ParaSails preconditioner cannot be used with the PCG solver.... ..."

### Table 4: CPU Times (seconds) for Uzawa-like method with BBT preconditioner solved by: multifrontal (MF) or IC conjugate gradient (PCG).

1996

"... In PAGE 6: ... Incomplete cholesky preconditioned conjugate gradient. In Table4 we analyze the solution of problem (7) for the constant permeability case ( quot; = 0) with the two proposed solvers for BBT . We give the number of outer iterations and the total solution times.... ..."

Cited by 2

### Table 1: Error rates of the GaBP solver versus those of the parallel sequential solver and SVMlight

in Jerusalem

"... In PAGE 2: ...Table 1: Error rates of the GaBP solver versus those of the parallel sequential solver and SVMlight Table1 describes the seven datasets we used to compare the al- gorithms and the classification accuracy obtained. These computa- tions were done using five processing nodes (3.... ..."