Skip to main content
. Author manuscript; available in PMC: 2012 Feb 1.
Published in final edited form as: Magn Reson Med. 2010 Nov 16;65(2):363–369. doi: 10.1002/mrm.22690

Fig. 3.

Fig. 3

(a) Computation time for populating encoding matrices (eight transmitters, 64×64 excitation resolution, reduction factor of two) using one CPU thread (stars), four CPU threads (triangles), one GPU (squares) and four GPUs (circles) as a function of the coefficient matrix size. (b) CGLS computation time as a function of the coefficient matrix size. Coefficient matrices of different sizes were generated using different dwell time for RF pulses.