Parallel MLEM on Multicore Architectures

Abstract. The efficient use of multicore architectures for sparse matrixvector multiplication (SpMV) is currently an open challenge. One algorithm which makes use of SpMV is the maximum likelihood expectation maximization (MLEM) algorithm. When using MLEM for positron emission tomography (PET) image reconstruction, one requires a particularly large matrix. We present a new storage scheme for this type of matrix which cuts the memory requirements by half, compared to the widelyused compressed sparse row format. For parallelization we combine the two partitioning techniques recursive bisection and striping. Our results show good load balancing and cache behavior. We also give speedup measurements on various modern multicore systems.

• Type: Journal
• Year: 2009
• Where: ICCS
• Authors: Tilman Küstner, Josef Weidendorfer, Jasmine Schirmer, Tobias Klug, Carsten Trinitis, Sybille Ziegler
• Keywords: Applied Computing, ICCS 2009, Likelihood Expectation Maximization, Positron Emission Tomography, Sparse Matrixvector Multiplication