The C Meataxe64 includes a number of parallel implementations of challenging computations. These take their inputs and return their results in disk files, which can be read and written using MTX64_WriteMatrix (1.5-2) and MTX64_ReadMatrix (1.5-2). They use a number of threads specified in the file src/mtx64/tuning.h. Each requires a tmpdir parameter which should be the pathname of a directory suitable for temporary files.
‣ MTX64_fMultiply( tmpdir, fn1, fn2, fn3 ) | ( function ) |
This function multiplies the matrices in files fn1 and fn2 (in that order) and writes the result into file fn3.
‣ MTX64_fMultiplyAdd( tmpdir, fn1, fn2, fn3, fn4 ) | ( function ) |
This function multiplies the matrices in files fn1 and fn2 (in that order) adds the result to teh matrix in fn3 and writes the result into file fn4.
‣ MTX64_fTranspose( tmpdir, fn1, fn2 ) | ( function ) |
This function computes the transpose of the matrix in filex fn1 and writes the result into file fn2.
‣ MTX64_fProduceNREF( tmpdir, fn1, fn2, fn3 ) | ( function ) |
This function computes the negative reduced echelon form of the matrix in file fn1, which is returned in two parts. A bitsstring in file fn2 which indicates the locations of pivot columns and a remnant in fn3 which contains the entries from the non-pivot columns of the pivot rows. The rank is returned
‣ MTX64_fEchelize( tmpdir, a, cs, rs, m, k, r ) | ( function ) |
This function computes the negative reduced echelon form of the matrix in file a, and writes both the result and the transforming matrix, in a total of five parts. The result is given as a bitsstring in file cs which indicates the locations of pivot columns and a remnant in r which contains the entries from the non-pivot columns of the pivot rows. In addition, the multiplier, cleaner and row select are returned in files m, k and rs, respectively. See 2 for more details. The rank is returned
generated by GAPDoc2HTML