DBM
Section can be repeated.
Benchmark and test the dbm routines [Edit on GitHub]
Keywords
Keyword descriptions
- ALPHA
Type: real
Default: 1.00000000E+000
Usage: ALPHA 2.0Description: Multiplication factor
- ALWAYS_CHECKSUM
Type: logical
Default: F
Lone keyword: T
Usage: ALWAYS_CHECKSUMDescription: perform a checksum after each multiplication
- ASPARSITY
Type: real
Default: 0.00000000E+000
Usage: ASPARSITY 70Description: Sparsity of A matrix
- BETA
Type: real
Default: 0.00000000E+000
Usage: BETA 1.0Description: Product premultiplication factor
- BSPARSITY
Type: real
Default: 0.00000000E+000
Usage: BSPARSITY 80Description: Sparsity of B matrix
- BS_K
Type: integer[ ]
Default: 1 13 2 15
Usage: BS_K 1 13 2 5Description: Block sizes of inner dimension
- BS_M
Type: integer[ ]
Default: 1 13 2 15
Usage: BS_M 1 13 2 5Description: Row block sizes of C
- BS_N
Type: integer[ ]
Default: 1 13 2 15
Usage: BS_N 1 13 2 5Description: Column block sizes of C
- CSPARSITY
Type: real
Default: 0.00000000E+000
Usage: CSPARSITY 90Description: Sparsity of C matrix
- FILTER_EPS
Type: real
Default: -1.00000000E+000
Usage: FILTER_EPS 1.0Description: Threshold for on-the-fly and final filtering.
- K
Type: integer
Default: 256
Usage: K 1024Description: Inner dimension M
- KEEPSPARSE
Type: logical
Default: F
Lone keyword: T
Usage: KEEPSPARSEDescription: Keep product sparse
- M
Type: integer
Default: 256
Usage: M 1024Description: Dimension 1 of C
- N
Type: integer
Default: 256
Usage: N 1024Description: Dimension 2 of C
- N_LOOP
Type: integer
Default: 10
Usage: N_LOOP 10Description: Number of operations being timed (useful for small matrices).
- TRANSA
Type: logical
Default: F
Lone keyword: T
Usage: TRANSADescription: Transpose matrix A
- TRANSB
Type: logical
Default: F
Lone keyword: T
Usage: TRANSBDescription: Transpose matrix B