alcf-mpi-benchmarks's Introduction
ALCF MPI benchmarks suite consists of five independent programs: mmps, pingpong, aggregate, bisection, and collectives, which measure and report, correspondingly, a zero length messaging rate, point-to-point communication latency with ping-pong benchmark, single node aggregate communication bandwidth, bisection bandwidth, and the latency of certain collective operations. Messaging rate “mmps”: To maximize the rate, the offeror may choose appropriate number of communicating neighbors XNBORS, the number of MPI tasks placed on the reference node, and the window size - the parameter, which determines the number of messaging sent simultaneously to each neighbor. The placement of the neighbors is defined by proper modification of the “getranks.c” file according to specifics of the interconnect subsystem. Point-to-point communication latency is measured by a “pingpong” benchmark for three cases: intranode, nearest neighbor, and farthest path. The offeror can modify the “getranks.c” file to specify the communicating tasks according to specifics of the interconnect. For intranode benchmark, the communicating tasks must reside on the same node. For nearest neighbor benchmark, the communicating tasks must be placed on two distinct nodes, connected by the nearest path, provided by the interconnect subsystem. For the farthest path benchmark, the two communicating tasks must be placed on two distinct nodes with the longest interconnect path between them. Single node aggregate bandwidth benchmark measures and reports the total aggregate interconnect bandwidth, available to a task by aggregating the bandwidth of all available links on the node. The benchmark utilizes point-to-point communications. To maximize the bandwidth, the offeror may specify “N” - the number of communicating tasks. The offeror may also modify the “getranks.c” file to place the communicating tasks according to specifics of their proposed interconnect subsystem. Bi-section bandwidth benchmark is measures and reports the aggregate bandwidth between the worse-case bi-section of the partition. The offeror should generate the bi-section by modifying the “getranks.c” file according to specifics of their interconnect subsystem. The latency of collective operations, specifically Barrier, Broadcast, and Allreduce, is measured and reported by a “collective” benchmark. The measurements are performed across entire partition via World communicator, as well as the two non-overlapping sub partitions, dividing the World in two equal parts. The offeror may specify the division of the World into parts by modifying the “split.c” file, according to their preference and using specifics of the interconnect subsystems.
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.