TY - GEN
T1 - High-performance blob-based iterative reconstruction of electron tomography on multi-GPUs
AU - Wan, Xiaohua
AU - Zhang, Fa
AU - Chu, Qi
AU - Liu, Zhiyong
PY - 2011
Y1 - 2011
N2 - Three-dimensional (3D) reconstruction of electron tomography (ET) has emerged as a leading technique to elucidate the molecular structures of complex biological specimens. Blob-based iterative methods are advantageous reconstruction methods for 3D reconstruction of ET, but demand huge computational costs. Multiple Graphic processing units (multi-GPUs) offer an affordable platform to meet these demands, nevertheless, are not efficiently used owing to a synchronous communication scheme and the limited available memory of GPUs. We propose a multilevel parallel scheme combined with an asynchronous communication scheme and a blob-ELLR data structure. The asynchronous communication scheme is used to minimize the idle GPU time. The blob-ELLR data structure only needs nearly 1/16 of the storage space in comparison with ELLPACK-R (ELLR) data structure and yields significant acceleration. Experimental results indicate that the multilevel parallel scheme allows efficient implementations of 3D reconstruction of ET on multi-GPUs, without loss any resolution.
AB - Three-dimensional (3D) reconstruction of electron tomography (ET) has emerged as a leading technique to elucidate the molecular structures of complex biological specimens. Blob-based iterative methods are advantageous reconstruction methods for 3D reconstruction of ET, but demand huge computational costs. Multiple Graphic processing units (multi-GPUs) offer an affordable platform to meet these demands, nevertheless, are not efficiently used owing to a synchronous communication scheme and the limited available memory of GPUs. We propose a multilevel parallel scheme combined with an asynchronous communication scheme and a blob-ELLR data structure. The asynchronous communication scheme is used to minimize the idle GPU time. The blob-ELLR data structure only needs nearly 1/16 of the storage space in comparison with ELLPACK-R (ELLR) data structure and yields significant acceleration. Experimental results indicate that the multilevel parallel scheme allows efficient implementations of 3D reconstruction of ET on multi-GPUs, without loss any resolution.
KW - blob
KW - electron tomography (ET)
KW - iterative methods
KW - multi-GPUs
KW - three-dimensional (3D) reconstruction
UR - http://www.scopus.com/inward/record.url?scp=79955860783&partnerID=8YFLogxK
U2 - 10.1007/978-3-642-21260-4_10
DO - 10.1007/978-3-642-21260-4_10
M3 - Conference contribution
AN - SCOPUS:79955860783
SN - 9783642212598
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 61
EP - 72
BT - Bioinformatics Research and Applications - 7th International Symposium, ISBRA 2011, Proceedings
T2 - 7th International Symposium on Bioinformatics Research and Applications, ISBRA 2011
Y2 - 27 May 2011 through 29 May 2011
ER -