Am 29.08.2019 um 10:38 schrieb Marcin Błażejowski:
Hi,
suppose we have two kx1 matrixes (bhat and bvar, in fact). I'd like to
ask what is more efficient (in terms of memory consumption): having in
bundle two separate matrixes or just one kx2?
I think the overhead of having one additional matrix object is
negligible, especially for large k.
(I assumed you are talking about the xml representation of the bundle,
to pass it around in MPI. You can look at the xml yourself if you store
the bundle to a file, possibly after unzipping. There is one more xml
node for the extra matrix, but this is just a fixed amount of text,
shouldn't really matter.)
hth,
sven