### Table 1 Alliant FX2816: LU, Cholesky, and QR factorizations of a 1024 1024 matrix.

"... In PAGE 5: ... All other level 3 BLAS used are from the GEMM-based library [19, 20]. In Table1 , performance results on one to eight processors for the three routines are shown. By restricting the size of the cluster to eight, only the two processors (of four) that have distinct cache controllers on each processor module are used.... ..."

### Table 3 IBM 3090 VF/600J: LU, Cholesky, and QR factorizations of a 1200 1200 matrix.

"... In PAGE 5: ... The level 3 BLAS used are from ESSL [16], except for a tuned FORTRAN implementation of DTRMM [22]. Results for IBM 3090 are presented in Table3 together with previously presented results, prev, for machine speci c... ..."

### Table 4. Complexity of Cholesky factorization with matrix dimension JL.

2005

"... In PAGE 58: ... If the algorithm contains a matrix inverse, it is always executed on a hermitian symmet- ric matrix, which in the current context is also always positive de nite and has a Cholesky factorization. The matrix inverse is assumed to be computed through the inverse of the Cholesky factorization of the argument matrix and to have the computational complexity listed in Table4 . In some cases computational complexity and the memory requirements of an algorithm can be exchanged, but the memory requirements of the algorithms are not... ..."

### Table 5.1: Number of oating point operations in the LU, Cholesky, and QR factorizations.

in Ring-oriented Block Matrix Factorization Algorithms for Shared and Distributed Memory Architectures

1992

Cited by 1

### Table 5.2: Results for the LU, Cholesky, and QR factorizations of a 1024 1024 matrix on Alliant FX2816.

in Ring-oriented Block Matrix Factorization Algorithms for Shared and Distributed Memory Architectures

1992

Cited by 1

### Table 3.1: Number of oating point operations in the LU, Cholesky, and QR factorizations.

### Table 4.5: Modeled times in secs for the LU, Cholesky, and QR factorizations of a 1000 1000 matrix.

### Table 2 Intel iPSC/2: LU, Cholesky, and QR factorizations of a 1000 1000 matrix.

"... In PAGE 5: ... The best performance obtained from the level 3 BLA routine DGEMM is around 0:5 M ops [5]. Results for iPSC/2 are presented in Table2 together with previously presented results, prev, for machine speci c implementations [5]. IBM 3090 VF/600J.... ..."

### Table 5.3: Results for the LU, Cholesky, and QR factorizations of a 1000 1000 matrix on Intel iPSC/2.

in Ring-oriented Block Matrix Factorization Algorithms for Shared and Distributed Memory Architectures

1992

Cited by 1

### Table 5.4: Results for the LU, Cholesky, and QR factorizations of a 1200 1200 matrix on IBM 3090 VF/600J.

1992

Cited by 1