https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/mkl-ddiasv-complexity/m-p/815269#M4249 Dear Petros,<BR /><BR />The mkl_?diasv routine solves one of the systems of equations: A*x or A^T*x =f where A is a triangular upper or lower sparse matrix stored in the diagonal format. The Thomas algorithm is used to solve tridiagonal systems of equations. So we can see that the functionality is different. <BR /><BR />As concerns the complexity of mkl_ddiasv, it is equal to 2*bandwidth *(matrix size).<BR />Sergey<DIV style="background-color: #ffffff; font-style: normal; font-family: " courier="" new=""></DIV> Fri, 10 Feb 2012 09:09:14 GMT Sergey_K_Intel1 2012-02-10T09:09:14Z https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/mkl-ddiasv-complexity/m-p/815268#M4248 Hi,<BR />I would like to ask on the complexity of the mkl_ddiasv sarse diagonal solver.<BR />In particular, if my diagonal matrix is a tridiagonal does it have a comparable complexity with the Thomas algorithm?<BR />(I realize that it cannot be the same, but if mt'ed serious benefits could be gained in using this insead of a hand-written Thomas).<BR />Also for 5-diagonal systems, is the complexity something like bandwidth*matrixSize?<BR />Thank you in advance for your help,<BR />Petros<BR /><BR />PS: btw,both solvers (Thomas-tridiagonal and 5-diagonal ) could be very useful additions to mkl. My understanding is that, currently, the only other avail are the BLAS/LAPACK ones that only economize in storage.<BR /> Tue, 07 Feb 2012 04:06:31 GMT https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/mkl-ddiasv-complexity/m-p/815268#M4248 Petros_Mamales 2012-02-07T04:06:31Z https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/mkl-ddiasv-complexity/m-p/815269#M4249 Dear Petros,<BR /><BR />The mkl_?diasv routine solves one of the systems of equations: A*x or A^T*x =f where A is a triangular upper or lower sparse matrix stored in the diagonal format. The Thomas algorithm is used to solve tridiagonal systems of equations. So we can see that the functionality is different. <BR /><BR />As concerns the complexity of mkl_ddiasv, it is equal to 2*bandwidth *(matrix size).<BR />Sergey<DIV style="background-color: #ffffff; font-style: normal; font-family: " courier="" new=""></DIV> Fri, 10 Feb 2012 09:09:14 GMT https://community.intel.com/t5/Intel-oneAPI-Math-Kernel-Library/mkl-ddiasv-complexity/m-p/815269#M4249 Sergey_K_Intel1 2012-02-10T09:09:14Z