[PDF][PDF] Application specific mesh partition improvement
SIAM Journal on Scientific Computing, 2015•scorec.rpi.edu
Partitioning unstructured meshes for parallel adaptive workflows running on leadership
class systems requires the application of scalable partitioning tools that meet the needs of
each workflow step. The combination of the (hyper) graph methods with two criteria partition
improvement to hundreds of thousands of cores was previously presented [23, 51]. This
work generalizes the partition improvement to multiple criteria and presents results to more
than one million cores of Mira BlueGene/Q on meshes with over 12 billion elements.
class systems requires the application of scalable partitioning tools that meet the needs of
each workflow step. The combination of the (hyper) graph methods with two criteria partition
improvement to hundreds of thousands of cores was previously presented [23, 51]. This
work generalizes the partition improvement to multiple criteria and presents results to more
than one million cores of Mira BlueGene/Q on meshes with over 12 billion elements.
Abstract
Partitioning unstructured meshes for parallel adaptive workflows running on leadership class systems requires the application of scalable partitioning tools that meet the needs of each workflow step. The combination of the (hyper) graph methods with two criteria partition improvement to hundreds of thousands of cores was previously presented [23, 51]. This work generalizes the partition improvement to multiple criteria and presents results to more than one million cores of Mira BlueGene/Q on meshes with over 12 billion elements.
scorec.rpi.edu
以上显示的是最相近的搜索结果。 查看全部搜索结果