From d7e561c6c8bac8edb0bfe55ad7214db811c5edc3 Mon Sep 17 00:00:00 2001 From: Edgar Solomonik Date: Thu, 13 Oct 2016 16:39:09 -0500 Subject: [PATCH] cleaned up README.md a bit and added build link --- README.md | 26 +++++++------------------- 1 file changed, 7 insertions(+), 19 deletions(-) diff --git a/README.md b/README.md index e8444894..899f64fa 100644 --- a/README.md +++ b/README.md @@ -1,27 +1,15 @@ -A distributed memory library for efficient decomposition of tensors of arbitrary type and parallel MPI+OpenMP execution of tensor contractions/functions. - -Basic build process: ./configure; make; - -Basic test process: sequential: 'make test', parallel: 'make test2' - -See examples/ directory to learn usage, or src/ directory to examine library source code. - - -Links in documentation below redirect to Doxygen documentation hosted at http://solomon2.web.engr.illinois.edu/ctf/index.html - -
-
-
Cyclops Tensor Framework Documentation
-

-Introduction

-

Cyclops Tensor Framework (CTF) is a distributed-memory library that provides support for tensor algebra. CTF tensors are multidimensional distributed arrays of arbitrary type, which may be used to represent graphs or hypergraphs. The primary means of specifying CTF tensor operations is assigning `iterator' indices for each way of each tensor and specifying an elementwise function to apply. Partitioning and efficient redistribution of tensor data is also possible via CTF.

-

The software is available on GitHub (github.com/solomonik/ctf) and maybe obtained via the command

-

git clone https://github.com/solomonik/ctf

+Cyclops Tensor Framework +

+

Cyclops Tensor Framework (CTF) is a distributed-memory library that provides support for tensor algebra. CTF tensors are multidimensional distributed arrays of arbitrary type, which may be used to represent graphs or hypergraphs.

+

To get an idea of how CTF can be used, browse through some of the short sample codes in the examples folder (or Examples doxygen module)!

+

Basic build process: ./configure; make;

+

Basic test process: sequential: 'make test', parallel: 'make test2'

+

The primary means of specifying CTF tensor operations is assigning `iterator' indices for each way of each tensor and specifying an elementwise function to apply. Partitioning and efficient redistribution of tensor data is also possible via CTF.

CTF requires the BLAS and MPI to be built, with MPI being the main parallel execution and communication mechanism. OpenMP and CUDA may additionally be provided for threading and accelerator support, respectively, but CTF will also build without them.

As a guide to modern usage of CTF for sparse matrix computations, graph computations, and tensor computations, we recommend the following paper

Edgar Solomonik and Torsten Hoefler; Sparse Tensor Algebra as a Parallel Programming Model; arXiv, Nov 2015. (link)