Tsinghua University

Beijing Polystar Hitech Co., Ltd.

Yuanwei Wang

Huanqi Cao

Zhenyu Wang

Xiaogang Hu

This disclosure provides a graph computation method based on distributed parallel computing, a distributed parallel computing system, and a computer-readable medium. The graph computation method includes the following steps: obtaining the data of the graph to be computed, where the graph consists of multiple vertices and edges. For a subgraph where the degrees of both the source vertex and the target vertex are greater than a predetermined threshold, and for the reduce-scatter type communication for the target vertex, the following operations are performed: for reductions along rows and columns, a ring algorithm is used; for global reductions, the message is first transposed locally to change the data order from row-major to column-major. Then, the row-wise reduce-scatter is performed first, followed by the column-wise reduce-scatter, to reduce communication across super node along columns. The graph computation method of this disclosure optimizes collective communication to reduce communication across super node along columns. The hierarchical reduce-scatter eliminates redundant data within the super node in the first step of row-wise reduction, so that the communication across the super node in the second step is minimal and without redundancy.

Collective communication optimization method for global high-degree vertices, and application

Provided are a distributed parallel computing-based graph computing method, a distributed parallel computing system, and a computer readable medium. The graph computing method comprises: obtaining data of a graph to be computed, dividing vertexes into a first extremely high degree class E, a second high degree class H, and a third regular class R, forming a super node by a predetermined number of compute nodes, the communication between nodes in a super node being faster than the communication between nodes across the super node, and dividing the nodes into grids, one node in each grid, the internal nodes in one super node being logically arranged in a row, vertices being evenly divided into the nodes according to numbers, Ri and Ro being maintained by the nodes to which Ri and Ro belong, the state of an Ho vertex being synchronously maintained on a column, the state of an Hi vertex being synchronously maintained on the column and the row, and the states of Eo and Ei vertexes being globally and synchronously maintained. According to the three-level vertex division method, sub-graphs are well balanced between the nodes, thereby facilitating elimination of repeated sending of a message to a same super node, and avoiding a high-cost global delegate.

Three-class vertex degree aware-based 1.5-dimensional graph division method and application

The present invention relates to method of preparing a monocrystalline diamond abrasive grain comprising the steps of: sufficiently dispersing a micron-scale monocrystalline diamond and at least a micron-scale metal powder in a sol containing at least a nano-scale metal powder, and controlling the suitable ratio of the two kinds of metal powder with different order of magnitude of particle size, accordingly the micron-scale metal powder can be stuffed into voids among the monocrystalline diamond grains, and only a suitable amount of nano-scale metal powder is required to let the micron-scale metal powder and nano-scale metal powder coat together on the surface of the micron-scale diamond to form an integral and uniform coating layer including at least two kinds of metal grains, thus the resulting monocrystalline diamond abrasive grains have a unique rough-surface morphology, a plurality of contact points and contact surfaces resenting in the grinding process and good self-sharpening.

Growing community of inventors

Beijing, China

Yuanwei Wang