Linear algebra subroutines for large SSD-resident dense and sparse matrices
Перейти к файлу
Harsha Vardhan Simhadri 9dac81f39c
Update README.md
2020-12-14 14:15:25 -08:00
drivers rename *.h -> bof_*.h 2019-03-20 19:41:59 +05:30
include rename *.h -> bof_*.h 2019-03-20 19:41:59 +05:30
misc renamed types.h to bof_types.h 2019-03-20 13:44:51 +00:00
src rename *.h -> bof_*.h 2019-03-20 19:41:59 +05:30
.gitignore first commit 2018-08-22 16:07:15 +05:30
CMakeLists.txt added mmap gemm benchmark 2018-10-10 22:00:15 +05:30
LICENSE first commit 2018-08-22 16:07:15 +05:30
README.md Update README.md 2020-12-14 14:15:25 -08:00
_clang-format first commit 2018-08-22 16:07:15 +05:30
nsdi2019_paper.pdf Added NSDI paper 2018-10-26 16:56:42 +05:30

README.md

This project is archived. Please reach to authors by email for questions.

BLAS-on-Flash Project Files

BLAS-on-Flash implements a small subset of BLAS routines (BLAS-2/3) that work with matrices/vectors stored on flash storage. In addition to BLAS routines, BlasonFlash also provides other routines like kmeans, sort, map, reduce for large-scale processing on disk-resident data. Currently, only the following routines are supported.

  • _gemm
  • _csrmm
  • _csrgemv
  • _csrcsc
  • sort
  • kmeans
  • map
  • reduce

Requirements

  • Ubuntu 16.04 or newer running Linux Kernel v4.13 or newer (Older kernels have issues setting nr_requests parameter)
  • cat /sys/block/<dev>/queue/nr_requests is at least 32768
  • Intel MKL 2017+ is installed (usually in /opt/intel) and added to LD_PRELOAD path; See CMakeLists.txt for finer control over MKL paths
  • libaio-dev package installed
  • make and cmake installed with cmake > 3.0.0

Build instructions

  • git clone ssh://git@github.com/Microsoft/BLAS-on-flash blas-on-flash
  • cd blas-on-flash
  • mkdir bin && cd bin
  • cmake -DCMAKE_BUILD_TYPE=X .. where X=Debug or X=Release
  • make -Bj

BLAS-on-Flash library design, architecture, and some performance numbers can be found in this paper.

Contributing

This project welcomes contributions and suggestions. Most contributions require you to agree to a Contributor License Agreement (CLA) declaring that you have the right to, and actually do, grant us the rights to use your contribution. For details, visit https://cla.microsoft.com.

When you submit a pull request, a CLA-bot will automatically determine whether you need to provide a CLA and decorate the PR appropriately (e.g., label, comment). Simply follow the instructions provided by the bot. You will only need to do this once across all repos using our CLA.

This project has adopted the Microsoft Open Source Code of Conduct. For more information see the Code of Conduct FAQ or contact opencode@microsoft.com with any additional questions or comments.