For instance: Humanities and Science University Journal
About consortium subscription Contacts
(812) 4095364 Non-commercial partnership
St. Petersburg


"Humanities and Science University Journal" №13 (Physical and mathematical, biological and technical science), 2015

GPU Memory Transfer Optimization for Computed Tomography Image Processing

V. S. Chukanov, I. V. Shturts
Price: 50 руб.
 Modern graphics processor units (GPUs) provide massive parallelism capabilities with hundreds of parallel threads with SIMD-oriented architecture which is well-suited for big data processing. In this paper we consider huge and complex data processing tasks where original dataset cannot be easily separated into independent parts and updated in parallel. We propose a pipelined method of data processing for such application
as computed tomography 3D image reconstruction. The method runs memory transfers concurrently with GPU processing and allows breaking the sequence “copy data chunk to the GPU — do update — copy updated data to the CPU”. It is a common case for computed tomography (CT) image reconstruction algorithms which require multiple iterations over the projection data in order to update a subset of voxels. Due to CT system geometry specifications data dependencies are complex and classic double
buffering technique cannot be applied.
Keywords: GPU, double buffering, image processing, computed tomography.
1. Nickolls, J. (2007, August). GPU Parallel Computing Architecture and CUDA
Programming Model. IEEE Hot Chips 19, Stanford, CA.
2. Lindholm, E., Nickolls, J., Oberman, S., & Montrym, J. NVIDIA Tesla: A unified
graphics and computing architecture. IEEE Micro, 2008, 28(2), 39–55.
3. Jeon, H., Xia, Y., & Prasanna, V.K. Parallel Exact Inference on a CPU-GPGPU
Heterogenous System. Proceeding of the 39th International Conference on Parallel
Processing, 2010, pp. 61–70.
4. Bauer, M., Cook, H., & Khailany, B. CudaDMA: optimizing GPU memory
bandwidth via warp specialization. Proceedings of 2011 International Conference for High Performance Computing, Networking, Storage and Analysis, 2011.
5. Mokhtari, R., & Stumm, M. BigKernel -- High Performance CPU-GPU Communication Pipelining for Big Data-Style Applications. Proceedings of the 2014
IEEE 28th International Parallel and Distributed Processing Symposium, 2014,
pp. 819–828.
6. Scherl, H., Keck, B., Kowarschik, M., & Hornegger, J. Fast GPU-based CT
reconstruction using the common unified device architecture (CUDA). Nuclear Science
Symposium Conference Record, 2007, Vol. 6, pp. 4464–4466.
Price: 50 рублей
To order