README.md

cuSOLVER orthgonalization by QR factorization example

Description

This code demonstrates a usage of cuSOLVER orgqr function for doing orthgonalization by QR factorization

A = Q * R

A is a 3x2 dense matrix

A = | 1.0 | 2.0 |
    | 4.0 | 5.0 |
    | 2.0 | 1.0 |

Examples perform following steps for both APIs:

• A = Q*R by GEQRF
• Form Q by ORGQR
• Check if Q is unitary or not

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

• cusolverDnDorgqr_bufferSize API
• cusolverDnDorgqr API
• cusolverDnDgeqrf API

Building (make)

Prerequisites

• A Linux/Windows system with recent NVIDIA drivers.
• CMake version 3.18 minimum

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_ormqr_example

Sample example output:

A = (matlab base-1)
1.00 2.00
4.00 5.00
2.00 1.00
=====
after geqrf: info = 0
after ormqr: info = 0
Q = (matlab base-1)
-0.22 0.53
-0.87 0.27
-0.44 -0.80
|I - Q**T*Q| = 1.414214E+00