Tsinghua University

Intel Corporation

Wenguang Chen

Weimin Zheng

Jidong Zhai

Wei Xue

Hu Tiger Chen

Kang Chen

Eric Jian Huang

Rui Guo Zhang

Yeh-Ching Chung

Robert Kuhn

Bob Kuhn

Yuyang Jin

Bowen Yu

A data processing method, an execution workstation, a distributed computing system and a computer-readable storage medium. The method includes: splitting an aggregate into a plurality of aggregate fragments; sending a first portion of aggregate fragments to a first execution workstation to perform reduction on the first execution workstation; receiving, from a second execution workstation, aggregate fragments produced by the second execution workstation and corresponding to a second portion of aggregate fragments and performing reduction on the aggregate fragments received and the second portion of aggregate fragments to update the second portion of aggregate fragments; re-determining a new first portion of aggregate fragments and a new second portion of aggregate fragments from a remaining portion excluding the first portion of aggregate fragments that have been sent from the plurality of aggregate fragments, and performing the sending, receiving and reduction; sending the aggregate fragment completing all reduction to a management workstation.

Data processing method, execution workstation, distributed computing system, and storage medium

A computer executed parallel program scalability bottleneck detection method is provided, which includes: building a program structure graph for a program source code; collecting performance data based on a sampling technique during runtime; the performance data including: performance data of each vertex of the program structure graph and inter-process communication dependence of communication vertices; building a program performance graph by filling the program structure graph with the collected performance data, the program performance graph recording data and control dependence of each process as well as inter-process communication dependence; detecting problematic vertices from the program performance graph, and starting from some or all of the problematic vertices, backtracking through data/control dependence edges within a process and communication dependence edges between different processes, to detect scalability bottleneck vertices.

Parallel program scalability bottleneck detection method and computing device

A hybrid network system, a communication method and a network node are provided. The hybrid network system includes a combined network system architecture of a point-to-point network and a broadcast network, the point-to-point network is provided with a plurality of point-to-point channels, and each network node can send a message to a corresponding network node or receive the message from the corresponding network node through the point-to-point channel; the broadcast network is provided with a plurality of broadcast channels, and each broadcast channel can only have one sender at most at the same time, but can have a plurality of receivers; at least a part of network nodes have the capability of sending messages to be broadcast, at least a part of network nodes have the capability of receiving the broadcast messages, and at least some network nodes are configured to select the broadcast network or the point-to-point network at least according to communication tasks to be executed per se to complete the communication tasks to be executed. Preferably, each network node has the capability of sending and receiving the broadcast message. The hybrid network system of the embodiment of the present invention has high flexibility and high communication performance.

Hybrid network system, communication method and network node

Provided are a method for multiple accesses in a memory, an apparatus supporting multiple accesses in the memory and a storage system. The method comprises: receiving a number N of addresses in the memory, wherein N is an integer larger than 1 and the number N of addresses are discontinuous; performing a preset operation according to the number N of addresses; and outputting the result of the operation. As a consequence, the performances of a computer system can be improved, and a user can input and use the desired addresses just as required by the user.

Method and apparatus for multiple accesses in memory and storage system, wherein the memory return addresses of vertexes that have not been traversed

The present disclosure provides a method for computing a support of an itemset candidate based on graph structure data, a method for determining a frequent itemset using the same, and a method for determining a frequent itemset based on a prefix tree. The method for computing a support of an itemset candidate based on graph structure data comprises: converting data in a database to graph structure data (S); obtaining an itemset candidate from the database (S); obtaining a connected component corresponding to the itemset candidate in the graph structure data (S); determining the number of vertices included in the connected component (S); and determining the number of vertices as the support of the itemset candidate (S). The method for computing the support of the itemset candidate based on the graph structure data and the method for determining the frequent itemset using the same provided by the embodiments of the present disclosure may avoid spending too much time in obtaining the support of the itemset candidate by scanning the database, which improves efficiency of the algorithm while ensuring effectiveness of the algorithm.

Method for computing support of itemset candidate based on graph structure data and application thereof

Methods, apparatus, and articles of manufacture to perform runtime trace filtering associated with application performance analysis are disclosed. A disclosed example method involves generating a first performance value based on first performance data associated with a first function of a first application process. A difference value is generated based on the first performance value and a historical performance value associated with the first function. The difference value is compared to a threshold value, and first trace data associated with execution of the first application process is collected based on the comparison of the difference value to the threshold value.

Methods and apparatus to collect runtime trace data associated with application performance

Methods and apparatus to perform process placement for distributed applications are disclosed. An example method comprises determining a mapping between a communication graph representative of communications of a distributed application and a topology graph representative of communication costs associated with a computing network, and executing the distributed application with the processes of the distributed application assigned to the processing entities of the computing network based upon the mapping.

Growing community of inventors

Beijing, China

Wenguang Chen