Thanks, Dirk, for your reply.
> 'data + data' may work for Armadillo because Armadillo generally has
> much better support for math operations. We know of some shortcomings
> in our code there (see TODO and/or Issue list on Github, help is welcome).
OK, thanks for the hint, that's certainly good to know.
> NumericMatrix tmp = results.slice(i); // even wrap as neeed
"auto-wrapping" doesn't seem to work (compile time error: conversion
from ‘arma::Mat<double>’ to non-scalar type ‘Rcpp::NumericMatrix {aka
Rcpp::Matrix<14>}’ requested). However,
NumericMatrix tmp = wrap(results.slice(i)); // even wrap as neeed
compiles complaintlessly. But then, memory is not reused:
tmp(0,0) = 666.;
Rprintf("%f / %f\n", tmp(0,0), results(0,0,i));
gives:
666.000000 / 0.000000
Bummer.
I guess I will end up rewriting myfun and all code depending on in to use old
school arrays (but only after three cups of coffee...).
Best,
Gregor
_______________________________________________
Rcpp-devel mailing list
[email protected]
https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel
