Skip to content
forked from HipGraph/FusedMM

Implementation of FusedMM method for IPDPS 2021 paper titled "FusedMM: A Unified SDDMM-SpMM Kernel for Graph Embedding and Graph Neural Networks"

Notifications You must be signed in to change notification settings

SNU-ARC/FusedMM

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

88 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

FusedMM

This is the official implementation of FusedMM method accepted for publication in IEEE IPDPS 2021 titled "FusedMM: A Unified SDDMM-SpMM Kernel for Graph Embedding and Graph Neural Networks".

PDF is available in arXiv

System Requirements

Users need to have the following software/tools installed in their PC/server. The source code was compiled and run successfully in Linux (Ubuntu and Debian distributions).

GCC version >= 4.9
OpenMP version >= 4.5

Some helpful links can be found at GCC, OpenMP and Environment Setup.

Compile and Run FusedMM

$ ./configure

It will probe the system and generate the Makefile and necessary configuration inside kernels for generated optimized kernels. This step will generate Makefile even if it fails to detect the hardware and users will be able to update the Makefile and kernels/Make.inc manually with appropriate hardware flags.

To build the generated library for single precision float, use

$ make

To test FusedMM for single precision float, use

$ make test 

Compiling step will generate all executable files inside the bin folder. To run the tester and timer with a specific kernel using FusedMM, please use the following format:

./bin/xsOptFusedMMtime_fr_pt -input dataset/harvard.mtx 

The optimized kernels have the prefix xsOptFusedMM* and the generalized kernels have the prefix xsFusedMM*. There are several parameters which can be provided as follows:

-input <string>, full path of input file (required).
-K <int>, dimension of the embedding.
-C <int> Cachesize in KB to use cache flushing in timer
-nrep <int> Number of repetition in timer  
-T <1,0> want to run the tester along with timer  

Download All Datasets of FusedMM

To conduct experiments using all the datasets of FusedMM paper, please download it from the following link: Datasets

Note for double precision floating point:

Configure step detects the SIMD width for single precision. For double precision, it is normally half the width. Update "pre" (d for double) and "vlen" (SIMD width) in Makefile accordingly and use make command.

Citation

If you find this repository helpful, please cite the following paper:

@inproceedings{rahman2020fusedmm,
  title={FusedMM: A Unified SDDMM-SpMM Kernel for Graph Embedding and Graph Neural Networks},
  author={Rahman, Md and Sujon, Majedul Haque and Azad, Ariful and others},
  booktitle={35th Proceedings of IEEE IPDPS},
  year={2021}
}

Contact

Please contact the following person if you have any questions: Ariful Azad (azad@iu.edu), Majedul Haque Sujon (msujon@iu.edu) or, Md. Khaledur Rahman (morahma@iu.edu).

About

Implementation of FusedMM method for IPDPS 2021 paper titled "FusedMM: A Unified SDDMM-SpMM Kernel for Graph Embedding and Graph Neural Networks"

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages

  • C 55.0%
  • C++ 42.8%
  • Shell 1.9%
  • CMake 0.3%