GPGPU – Visual Computing Lab

Papers published in the 2014 IEEE International Ultrasonics Symposium proceedings

Thomas Kim Kjeldsen — Thu, 27 Nov 2014 08:08:10 +0000

In our Advanced Technology Foundation project “FutureSonic”, we recently presented two papers at the 2014 IEEE International Ultrasonics Symposium together with our research partners at the Technical University of Denmark.

The first paper [1] presents how ultrasound images can be computed efficiently on GPUs and on multicore CPUs that support Single Instruction Multiple Data (SIMD) extensions. We were able to accelerate a reference implementation in C from around 700 ms/frame to 5.4 ms/frame using the same multicore CPU. The speedup was achieved primarily by optimizing the memory access patterns and by utilizing AVX instructions. On a high-end GPU the fastest computation time was less than 0.5 ms/frame.

The results obtained above were utilized in the second paper [2] where the GPU implementation was ported to mobile devices. We showed that modern mobile GPUs provide enough computing power to produce ultrasound images in real-time. Furthermore, we showed that the WiFi throughput is sufficient for real-time reception of raw data from a wireless ultrasound trandsucer.

References
[1] Synthetic Aperture Sequential Beamforming implemented on multi-core platforms
IEEE International Ultrasonics Symposium (IUS), p2181 – 2184 (2014)

[2] Implementation of synthetic aperture imaging on a hand-held device
IEEE International Ultrasonics Symposium (IUS), p2177 – 2180 (2014)

Real-time finite element modelling using CUDA

Peter Trier Mikkelsen — Thu, 13 Aug 2009 10:59:15 +0000

Here is a little movie showing real time simulation of non-linear elastic material properties using the Total Lagrangian Explicit Dynamic FEM. Three different sets of material parameters were used. Our implementation is done in CUDA. Thanks to Brian Bunch Christensen and Jens Rimestad for cooperation on the implementation.

The source code is under the LGPL licence and can be found here. Please acknowledge if you use it for your research. Thanks to Movania Muhammad Mobeen for tidying up the project so it should compile out of the box.

Dirk Fortmeier made this project work in linux using CMAKE. You can find the code for this at https://github.com/fortmeier/CUDA-FE-With-Haptics which also includes haptics interaction.

How we used the model in previous research can be read in the following paper:

Solid mesh registration for radiotherapy treatment planning.
K. Ø. Noe, T.S. Sørensen.
2010 International Symposium on Biomedical Simulation (ISBMS).
Lecture Notes in Computer Science, 2010; 5958: 59-70.
[abstract]

Some videos showing our use can be found here.

4. DIVISION

Jesper Mosegaard — Tue, 16 Jun 2009 13:42:06 +0000

Interactive 4. division.

Cardiac Surgery Simulator

Jesper Mosegaard — Wed, 06 May 2009 14:01:01 +0000

Ray marching smoke

Peter Trier Mikkelsen — Wed, 06 May 2009 13:36:20 +0000

Nice real-time smoke visualization, rendered by our CUDA tracer including a clear sky model. The smoke dataset was delivered by Michael Bang and Brian Bunch Christensen from the Department of Computer Science, University of Aarhus.

Raytracing on the GPU

Peter Trier Mikkelsen — Fri, 01 May 2009 16:32:14 +0000

We have recently begun initial experiments and implementation of a GPU assisted raytracer (through CUDA). Our vision is a progressive global illumination renderer with real time updates.

CUDA raytracer – refraction and reflection test

New employee: Karsten Noe

Jesper Mosegaard — Fri, 01 May 2009 09:56:41 +0000

We are happy to announce that Karsten Noe has signed the letter of appointement, and will thus join the computer graphics group at the Alexandra Institute from the 1st of august 2009. Karsten was previously a PhD student (expecting a PhD degree soon) from the Department of Computer Science, University of Aarhus, working on GPU accelerated registration of deformable tissue in radiation therapy.

Simulation of Congenital Heart Surgeryproject

Jesper Mosegaard — Thu, 30 Apr 2009 20:49:12 +0000

Cardiac Surgery Simulation - Illustrative Heart

The real-time simulation of congenital heart surgery has been a research project since 2003. We have successfully designed and implemented a very detailed cardiac surgical simulation through the utilization of the GPU for general purpose calculations, e.g. soft tissue deformation, displacement mapping and haptic feedback. We have written a number of papers on these subject (see http://www.jespermosegaard.dk/Publications).

Cardiac Surgical Simulation - Open Thorax

The research project was a collaboration between computer scientists and surgeons from Aarhus, London and Tübingen.

Easy GPGPU program – OpenGL and CG

Jesper Mosegaard — Wed, 29 Apr 2009 18:43:56 +0000

As part of the GPGPU course at the University of Aarhus in 2005 we developed a very simple set of base-classes for General Purpose Computation using the Graphics Processing Unit (GPGPU) through OpenGL, Nvidia CG, and either framebuffer objects or PBuffers for render-to-texture functionality. Today you should ideally use Nvidia CUDA or OpenCL for GPGPU – but the code might still be of interest for older hardware or a pure OpenGL/CG based approach to GPGPU:

SimpleReactionDiffusion (framebuffer_object).zip

The archive file includes the EasyGPUProgram class that has methods to initiate data in a 2d grid layout, do computation (as a cg fragment shader), and retrieve the data. We have included a reaction-diffusion example based on GPU Gems 2 chapter 31 using the EasyGPUProgram class.