Image Rotation using NVIDIA NPP with CUDA

Overview

Took an existing, non-working example that tried to do image rotations and fixed it to rotate an image. Above and beyond this, figured out how the translation of rotated images works to create a pretty pinwheel effect.

More complicated part (and you can see this from the github history) was the big puzzle about what getRotationBoundingBox() returned and the offsets used in nppiRotate_8u_C1R() (also didn't help that nothing said nppi::loadImage() only understood 8-bit grey images. Learned how to use the offsets render the results correctly in the final image. It turns out that the rotation point is always the top left corner of the image, which threw me off for a long time.

Code Organization

bin/ This folder should hold all binary/executable code that is built automatically or manually. The executable is pinwheelNPP.

data/ This folder should hold all example data in any format. If the original data is rather large or can be brought in via scripts, this can be left blank in the respository, so that it doesn't require major downloads when all that is desired is the code/structure.

lib/ Any libraries that are not installed via the Operating System-specific package manager should be placed here, so that it is easier for inclusion/linking.

src/ The source code should be placed here in a hierarchical fashion, as appropriate.

README.md This file should hold the description of the project so that anyone cloning or deciding if they want to clone this repository can understand its purpose to help with their decision.

INSTALL This file should hold the human-readable set of instructions for installing the code so that it can be executed. If possible it should be organized around different operating systems, so that it can be done by as many people as possible with different constraints.

Makefile or CMAkeLists.txt or build.sh There should be some rudimentary scripts for building your project's code in an automatic fashion.

run.sh An optional script used to run your executable code, either with or without command-line arguments.

Key Concepts

Performance Strategies, Image Processing, NPP Library

Supported SM Architectures

SM 3.5 SM 3.7 SM 5.0 SM 5.2 SM 6.0 SM 6.1 SM 7.0 SM 7.2 SM 7.5 SM 8.0 SM 8.6

Supported OSes

Linux, Windows

Supported CPU Architecture

x86_64, ppc64le, armv7l

CUDA APIs involved

Dependencies needed to build/run

FreeImage, NPP

Prerequisites

Download and install the CUDA Toolkit 11.4 for your corresponding platform. Make sure the dependencies mentioned in Dependencies section above are installed.

Build and Run

Windows

The Windows samples are built using the Visual Studio IDE. Solution files (.sln) are provided for each supported version of Visual Studio, using the format:

*_vs<version>.sln - for Visual Studio <version>

Each individual sample has its own set of solution files in its directory:

To build/examine all the samples at once, the complete solution files should be used. To build/examine a single sample, the individual sample solution files should be used.

Note: Some samples require that the Microsoft DirectX SDK (June 2010 or newer) be installed and that the VC++ directory paths are properly set up (Tools > Options...). Check DirectX Dependencies section for details."

Linux

The Linux samples are built using makefiles. To use the makefiles, change the current directory to the sample directory you wish to build, and run make:

$ cd cuda-class
$ make

The samples makefiles can take advantage of certain options:

• TARGET_ARCH= - cross-compile targeting a specific architecture. Allowed architectures are x86_64, ppc64le, armv7l. By default, TARGET_ARCH is set to HOST_ARCH. On a x86_64 machine, not setting TARGET_ARCH is the equivalent of setting TARGET_ARCH=x86_64.
  $ make TARGET_ARCH=x86_64
$ make TARGET_ARCH=ppc64le
$ make TARGET_ARCH=armv7l
  See here for more details.

• dbg=1 - build with debug symbols

  $ make dbg=1
  

• SMS="A B ..." - override the SM architectures for which the sample will be built, where "A B ..." is a space-delimited list of SM architectures. For example, to generate SASS for SM 50 and SM 60, use SMS="50 60".

  $ make SMS="50 60"
  

• HOST_COMPILER=<host_compiler> - override the default g++ host compiler. See the Linux Installation Guide for a list of supported host compilers.

    $ make HOST_COMPILER=g++

Running the Program

After building the project, you can run the program using the following command:

./bin/pinwheelNPP

or

- Copy code
./run.sh 

The ./run.sh example includes other command-line args for you to play with.

This command will execute the compiled binary, generating a pinwheel of the (data/Lena.p) starting at a 5 degree rotation from a position on lower right of the image and maing 12 degree additional rotation until completing a full 360 (maybe a little less) circuit. The result is saved in Lena_rotated.png in the data/ directory.

If you wish to run the binary directly with custom input/output files, you can use:

- Copy code
./bin/pinwheelNPP -input=data/Lena.pgm -output=data/Lena_rotated.pgm

To play around with different starting angles and rotation increments use:

- Copy code
./bin/pinwheelNPP -input=data/Lena.pgm -angle=15 -angleStep=15 -output=data/Lena_rotated.pgm

• Cleaning Up To clean up the compiled binaries and other generated files, run:

- Copy code
make clean

This will remove all files in the bin/ directory.

You'll need to manually remove any files you create either in the data/ directory or with the -output= argument.