Local mesh refinement multilevel techniques
SIAM Journal on Scientific and Statistical Computing
ACM Transactions on Mathematical Software (TOMS)
Thread scheduling for cache locality
Proceedings of the seventh international conference on Architectural support for programming languages and operating systems
Downwind numbering: robust multigrid for convection—diffusion problems
Applied Numerical Mathematics - Special issue on multilevel methods
A Supernodal Approach to Sparse Partial Pivoting
SIAM Journal on Matrix Analysis and Applications
Performance of Various Computers Using Standard Linear Equations Software
Performance of Various Computers Using Standard Linear Equations Software
| Hi-index | 0.00 |
We present a new method for storing sparse matrices as a block-matrix graph where the block are stored as local valuue arrays which are indexed via a set of a compact row-ordered schemes. This combines the flexibility of the graph structure with higher efficiency due to higher data locality. The inner compact pattern also allows identification of entries, which can lead to further advantages with respect to memory and computing time. To examine the efficiency of this new method, we study of a model case the performance of a matrix-vector multiplication which is the basic building block for most iterative methods. It turns out that our technique is competitive except for very small inner blocks. We also present two more realistic reactive flow applications to which this technique was applied.