README.md

cuBLAS Level-1 APIs - cublas<t>rotmg

Description

This code demonstrates a usage of cuBLAS rotmg function to apply the Givens rotation matrix to vector x and y

A = 1.0
B = 5.0

See documentation for further details.

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

• cublas<t>rotmg 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 cublas_examples.sln project in Visual Studio and build

Usage

$  ./cublas_rotmg_example

Sample example output:

A
1.00
=====
B
5.00
=====
X
2.10
=====
Y
1.20
=====
param
1.00 5.00 6.00 7.00 8.00 
=====
A
3.10
=====
B
0.62
=====
X
1.94
=====