high performance computing on graphics processing units: hgpu.org

Computed tomography (CT) is a widespread method used to study the internal structure of objects. The method has applications in medicine, industry and other fields of human activity. In particular, Electronic Imaging, as a species CT, can be used to restore the structure of nanosized objects. Accurate and rapid results are in high demand in modern science. However, there are computational limitations that bound the possible usefulness of CT. On the other hand, the introduction of high-performance calculations using Graphics Processing Units (GPUs) provides improving quality and performance of computed tomography investigations. Moreover, parallel computing with GPUs gives significantly higher computation speeds when compared with (Central Processing Units) CPUs, because of architectural advantages of the former. In this paper a computed tomography method of recovering the image using parallel computations powered by NVIDIA CUDA technology is considered. The implementation of this approach significantly reduces the required time for solving the CT problem.