Mentions légales du service

Skip to content
Snippets Groups Projects
hhakim's avatar
hhakim authored
Remove the light GPU implementation (aka FaustGPU) as it was an experimental implementation based on gpu_mod (while GPU2 is the full implementation).

Before removing I verified there was no loss in performance (with pyfaust-3.0.17).

(latest_pyfaust) hhadjdji@v100alpha:~$ python ./test_gpu_mod.py | grep -v changed
F.shape (1024, 1024) F.numfactors: 5 fac_type= sparse
***toarray*** cpu time: 1.6364107518456876 FaustGPU time: 1.6207437822595239 GPU2 time: 0.609403109177947
***F@M*** cpu time: 2.066476149018854 FaustGPU time: 1.0155699900351465 GPU2 time: 0.8688869220204651
***F.norm(2)*** cpu time: 0.03744047088548541 FaustGPU time: 0.20482274424284697 GPU2 time: 0.19360520970076323
F.shape (1024, 1024) F.numfactors: 5 fac_type= dense
***toarray*** cpu time: 6.241591406054795 FaustGPU time: 6.68755641579628 GPU2 time: 0.23564746510237455
***F@M*** cpu time: 7.5440095462836325 FaustGPU time: 1.0803766320459545 GPU2 time: 0.9672970180399716
***F.norm(2)*** cpu time: 4.2132647559046745 FaustGPU time: 0.24246237566694617 GPU2 time: 0.24203181779012084

The script is hosted on gitlab : https://gitlab.inria.fr/faustgrp/faust/-/snippets/708
610de11b
History

pipeline status FAµST Logo

FAuST Toolbox -- Flexible Approximate Multi-Layer Sparse Transform

General purpose

The FAuST toolbox contains a C++ code implementing a general framework designed to factorize matrices of interest into multiple sparse factors. It contains a template CPU/GPU C++ code and a Matlab wrapper. A Python wrapper is also available. The algorithms implemented here are described in details in [1]- Le Magoarou

For more information on the FAuST Project, please visit the website of the project: FAµST website


Installation

Please refer to the document "./gettingStartedFAuST-version2_0.pdf" to install the FAUST toolbox. The FAUST toolbox has been tested on the following environments:

  • LINUX (fedora 24 - 33 / centos 7 / Ubuntu)
  • MACOS X
  • WINDOWS (windows 10)

Quick install on UNIX

Unpack the directory.
mkdir ./build
cd ./build
cmake .. OR ccmake .. (with Graphical User Interface)
make
make install

Warning: The Matlab interface of FAuST requires compiling mex files. The mex compiler compatible with specific versions of gcc depending on the platform used. For more information, please refer to the Mathworks website.


Quickest Install on Linux, Windows and macOS

Pre-compiled packages from Gitlab Continuous Integration are also available. Except of course PIP packages, all packages include matlab and python wrappers, below are the latest release links.


License

Cf. license.txt


Contacts

Rémi Gribonval: remi.gribonval@inria.fr
Hakim: hakim.hadj-djilani@inria.fr

Credits

Luc Le Magoarou
Remi Gribonval
Nicolas Bellot
Adrien Leman
Thomas Gautrais
Hakim Hadj-Djilani

References

[1] Le Magoarou L. and Gribonval R., "Flexible multi-layer sparse approximations of matrices and applications", Journal of Selected Topics in Signal Processing, 2016.