syevjBatched

cuSOLVER Batched Standard Symmetric Dense Eigenvalue solver (via Jacobi method) example

Description

This code demonstrates a usage of cuSOLVER syevjBatched function for using syevjBatched to compute spectrum of a pair of dense symmetric matrices by

A_jx = λx

where A0 and A1 is a 3x3 dense symmetric matrices

A0 = |  1.0 | -1.0 | 0.0 |
     | -1.0 |  2.0 | 0.0 |
     |  0.0 |  0.0 | 0.0 |

A1 = | 3.0 | 4.0 | 0.0 |
     | 4.0 | 7.0 | 0.0 |
     | 0.0 | 0.0 | 0.0 |

The following code uses syevjBatched to compute eigenvalues and eigenvectors

A_jx = V_j * W_j * V^T_j

The user can disable/enable sorting by the function cusolverDnXsyevjSetSortEig.

Supported SM Architectures

All GPUs supported by CUDA Toolkit (https://developer.nvidia.com/cuda-gpus)

Supported OSes

Linux
Windows

Supported CPU Architecture

x86_64
ppc64le
arm64-sbsa

CUDA APIs involved

• cusolverDnSsyevjBatched_bufferSize API
• cusolverDnSsyevjBatched API

Building (make)

Prerequisites

• A Linux/Windows system with recent NVIDIA drivers.
• CMake version 3.18 minimum
• Minimum CUDA 9.0 toolkit is required.

Build command on Linux

$ mkdir build
$ cd build
$ cmake ..
$ make

Make sure that CMake finds expected CUDA Toolkit. If that is not the case you can add argument -DCMAKE_CUDA_COMPILER=/path/to/cuda/bin/nvcc to cmake command.

Build command on Windows

$ mkdir build
$ cd build
$ cmake -DCMAKE_GENERATOR_PLATFORM=x64 ..
$ Open cusolver_examples.sln project in Visual Studio and build

Usage

$  ./cusolver_syevjBatched_example

Sample example output:

A0 = (matlab base-1)
1.00 -1.00 0.00
-1.00 2.00 0.00
0.00 0.00 0.00
=====
A1 = (matlab base-1)
3.00 4.00 0.00
4.00 7.00 0.00
0.00 0.00 0.00
=====
matrix 0: syevj converges
matrix 1: syevj converges
====
W0[0] = 0.381966
W0[1] = 2.618034
W0[2] = 0.000000
====
W1[0] = 0.527864
W1[1] = 9.472136
W1[2] = 0.000000
====
V0 = (matlab base-1)
0.85 -0.53 0.00
0.53 0.85 0.00
0.00 0.00 1.00
V1 = (matlab base-1)
0.85 0.53 0.00
-0.53 0.85 0.00
0.00 0.00 1.00