JeffFessler
Some tiny performance improvements
(M for modified)
M mri/mri_sensemap_denoise.m
Forcing parfor to avoid queueing sMaps estimation, which could be slow in, e.g., 32-channel case.
M penalty/Cdiff.m
Matlab ndims()'s return always >= 2.
M systems/@fatrix2/cat.m
Avoid nesting @fatrix2 when numel(varargin)==1
M systems/@fatrix2/mtimes.m
Enabling A*c, where A is a @fatrix2, c is a scalar.
M systems/@fatrix2/private/fatrix2_subsref_colon.m
Use parfor for fatrix2-matrix product, when parpool is already init'ed, (initialize a parpool can be expensive by itself, hence only use parfor when parpool is already available)
M systems/Gmri.m
Enabling default nufft_args to support 3D;
Substituting repmat w/ bsxfun, RAM friendly.
Seems like this adds the dependency for parallel toolbox which MIRT doesn't assume?
Seems like this adds the dependency for parallel toolbox which MIRT doesn't assume?
Right, but I think it is safe to assume user who needs sMap estimation has access to the parallel toolbox.