We report on the development of a Synthetic Aperture Radar (SAR) backprojection algorithm implemented on multiple Graphics Processing Units (GPUs) in CUDA language and using a Non-Uniform FFT (NUFFT) routine to further numerically accelerating the calculations. The performance of the approach is analyzed in terms of computational speed and scalability on the GPU ”Jazz” cluster available at the Consorzio Interuniversitario per le Applicazioni del Supercalcolo per l’Universit`a e la Ricerca (CASPUR; Inter- University Consortium for the Application of Super-Computing for Universities and Research), Rome, Italy. The results, referring to the case of an individual node with 2-GPUs, show that the processing time scales approximately by 2 as compared to the case of a single GPU. Experimental results against the Air Force Research Laboratory (AFRL) airborne data delivered under the ”challenge problem for SAR-based Ground Moving Target Identification (GMTI) in urban environments” and collected under circular flight paths are also shown. The full processing of the data took approximately 19s on the mentioned 2-GPUs node.
NUFFT-based SAR backprojection on multiple GPUs / Capozzoli, Amedeo; Curcio, Claudio; Liseno, Angelo; P. V., Testa. - (2012), pp. 62-68. (Intervento presentato al convegno 2012 Tyrrhenian Works hop on Advances in Radar and Remote Sensing tenutosi a Napoli nel Sept. 12-14, 2012) [10.1109/TyWRRS.2012.6381104].
NUFFT-based SAR backprojection on multiple GPUs
CAPOZZOLI, AMEDEO;CURCIO, CLAUDIO;LISENO, ANGELO;
2012
Abstract
We report on the development of a Synthetic Aperture Radar (SAR) backprojection algorithm implemented on multiple Graphics Processing Units (GPUs) in CUDA language and using a Non-Uniform FFT (NUFFT) routine to further numerically accelerating the calculations. The performance of the approach is analyzed in terms of computational speed and scalability on the GPU ”Jazz” cluster available at the Consorzio Interuniversitario per le Applicazioni del Supercalcolo per l’Universit`a e la Ricerca (CASPUR; Inter- University Consortium for the Application of Super-Computing for Universities and Research), Rome, Italy. The results, referring to the case of an individual node with 2-GPUs, show that the processing time scales approximately by 2 as compared to the case of a single GPU. Experimental results against the Air Force Research Laboratory (AFRL) airborne data delivered under the ”challenge problem for SAR-based Ground Moving Target Identification (GMTI) in urban environments” and collected under circular flight paths are also shown. The full processing of the data took approximately 19s on the mentioned 2-GPUs node.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
