MAGMA Downloads

MAGMA provides implementations for CUDA, HIP, Intel Xeon Phi, and OpenCL. The latest releases are MAGMA 2.7.0 for CUDA and HIP, MAGMA MIC 1.4.0 for Intel Xeon Phi, and clMAGMA 1.3 for OpenCL. The libraries available for download are listed below in the order of their release dates.

Please use any of the following publications to reference MAGMA.

MAGMA Bitbucket repository:


MAGMA 2.7.0

MAGMA 2.7.0 is now released. This release includes:

  • Add support for builds targeting NVIDIA's Hopper architecture;
  • New routine: magma_dshposv_gpu and magma_dshposv_native solve Ax = b,
    for a symmetric positive definite matrix 'A', using FP16 during the Cholesky
    factorization. GMRES-based iterative refinement is used to recover the solution
    up to double precision accuracy. The '_gpu' suffix denotes a hybrid CPU-GPU
    factorization, while '_native' denotes a GPU-only factorization;
  • Performance improvement for the batch QR factorization routine;
  • Performance improvement for the variable size batch LU factorization routine;
  • Bug fixes, performance optimizations, benchmark additions, and maintenance
    updates to support current and new MAGMA routines, latest NVIDIA and AMD
    math libraries and GPU hardware.
magma-2.7.0.tar.gz   Download View License

MagmaDNN 1.2

MagmaDNN 1.2 is now available. MagnaDNN provides HP data analytics and machine learning tools using MAGMA as its computational backend. Updates in this release include:

  • Bug fixes and performance improvements;
  • oneDNN (MKL DNN) support including fully connected, convolutional, and pooling layers;
  • CMake build system;
  • Added NN examples including CNN, ResNet, AlexNet, LeNet, MNIST and CIFAR interactive, and VGG16;
  • Added examples for Tensor Math;
  • C++ style formatter;
  • Modularized distributed optimizer;
  • CIFAR10, CIFAR100, and MNIST data loaders;
  • CUDA streams added;
  • Model summary prinout;
  • Spack package manager installation support.

More information on MagmaDNN 1.2 is given in this paper and tutorial.

Cite as:

Daniel Nichols, Rocco Febbo, Florent Lopez, Kwai Wong, Stanimire Tomov, and Jack Dongarra. (2020, July 30). MagmaDNN (Version 1.2). Zenodo.

MagmaDNN's repository is on Bitbucket:

release-magmadnn-v1.2.tar.gz   Download View License


MAGMA MIC 1.4.0 is now available. This release provides implementations for MAGMA's one-sided (LU, QR, and Cholesky) and two-sided (Hessenberg, bi- and tridiagonal reductions) dense matrix factorizations, as well as linear and eigenproblem solver for Intel Xeon Phi Coprocessors. More information on the approach is given in this presentation.

magmamic-1.4.0.tar.gz   Download View License

clMAGMA 1.3

clMAGMA is an OpenCL port of MAGMA. It supports AMD GPUs. The clMAGMA library dependancies, in particular optimized GPU OpenCL BLAS and CPU optimized BLAS and LAPACK for AMD hardware, can be found in the AMD clMath Libraries (formerly APPML).

Included in the clMAGMA 1.3 release are routines for the following algorithms:

  • LU, QR, and Cholesky factorizations in both real and complex  arithmetic (single and double);
  • Linear and least squares solvers based on correspondingly the LU/Cholesky and QR factorizations in both real and complex  arithmetic (single and double);
  • Reductions to Hessenberg, bidiagonal, and tridiagonal forms using orthgonal similarity transformationsin both real and complex arithmetic (single and double);
  • Eigen and singular value problem solvers in both real and complex arithmetic (single and double);
  • Orthogonal transformation routines.
clmagma-1.3.0.tar.gz   Download View License


Copyright © 2022 The University of Tennessee. All rights reserved.

Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met:
· Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer.
· Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer listed in this license in the documentation and/or other materials provided with the distribution.
· Neither the name of the copyright holders nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission.

This software is provided by the copyright holders and contributors "as is" and any express or implied warranties, including, but not limited to, the implied warranties of merchantability and fitness for a particular purpose are disclaimed. in no event shall the copyright owner or contributors be liable for any direct, indirect, incidental, special, exemplary, or consequential damages (including, but not limited to, procurement of substitute goods or services; loss of use, data, or profits; or business interruption) however caused and on any theory of liability, whether in contract, strict liability, or tort (including negligence or otherwise) arising in any way out of the use of this software, even if advised of the possibility of such damage.

Nov 26 2022 Admin Login