Multivariate distributions and arma::rowvec
Closed this issue · 1 comments
For the multivariate normal and t, I need to benchmark transposing mu in the loop vs assigning the transpose to a rowvec outside the loop. For both, I should go ahead and overload the functions to take mu as a rowvec or as a vec, as it may be more convenient for users to pass one versus the other in various circumstances. Of course, all of this is on hold as I can't move back into dev until I hear back from CRAN on v 0.1 (submitted October 2).
It turns out (unsurprisingly) that transposing mu in the loop is slower than assigning the transpose to a rowvec outside the loop; see attached files (renamed so the suffix is .txt rather than .cpp so I could attach -- see the rules). A commit will be along shortly changing how this is done, and I'll likely add overloads that take mu as a rowvec later today or tomorrow.