I do believe that arma uses the correct back solve when everything is called
correctly (I haven’t profiled the code recently to check what BLAS functions
are being called though). I get better performance from Rcpp than R when I
change the back solve line to be
RI = solve(trimatu(R_upptri),I);
Very nice discussion, thanks to all involved.
We generally have Conrad's ear for simple-enough changes so if someone wants
to help getting the triangular solver access to BLAS into Armadillo we can
probably make that happen. Otherwise we can either hit BLAS directly, or use
Eigen as Gabor sugges
On Fri, Jun 27, 2014 at 7:14 PM, Peter Rossi wrote:
> Folks-
>
> In my package, bayesm, I use backsolve() to invert upper-triangular
> arrays. I am in the process of converting my package to rccp-arma.
>
> I thought I would test the analogous operation in arma by declaring
> the matrix as upper-t
Hi Hao,
from what I understand Armadillo doesn't have a forward or back-solver.
In any case it's pretty easy to create one, for example you can have a look
here:
https://github.com/mfasiolo/mvnfast/blob/master/src/mahaCpp.cpp
Have a nice weekend,
Matteo
On Sat, Jun 28, 2014 at 12:25 AM, Hao
Hi Peter,
I think Soren had a similar issue in February (about Armadillo not having an
optimized solver for triangular matrices):
http://lists.r-forge.r-project.org/pipermail/rcpp-devel/2014-February/007147.html
I don't think there was ever a reply though...
In contrast, the notes for backsolve
Folks-
In my package, bayesm, I use backsolve() to invert upper-triangular
arrays. I am in the process of converting my package to rccp-arma.
I thought I would test the analogous operation in arma by declaring
the matrix as upper-triangular and inverting using solve().
To my surprise, pure R co
This may be too-simpleminded for subscribers to this list, but I've
posted my two examples to:
https://github.com/DavisDaddy/RcppLinAlg
-- Mike
On Fri, Jun 27, 2014 at 11:38 AM, Kevin Ushey wrote:
> Hi Michael,
>
> Thanks for reporting -- things like this signal that we could do better
> er
Hi Michael,
Thanks for reporting -- things like this signal that we could do better
error reporting, e.g. it might be more appropriate to see
Error in eval(expr, envir, enclos) (from srcConn#6) :
Wrong R type for mapped vector (expected 'double', got 'integer')
or something to that effec
Thanks, Dirk. Last night I woke up with a cramp in my leg and the
inspiration that the problem was in the R test code, not in the call
to Eigen. I was absent-mindedly creating integer matrices in my test
code. After I changed to:
amat <- matrix(as.numeric(1:9), nrow=3)
etc., the Eigen stuf
Hi,
sorting an Rcpp Vector leaves it's name attribute unsorted.
See this code:
require(Rcpp)
require(inline)
f <- cxxfunction(sig=c(xx="integer"), plugin="Rcpp", body='
IntegerVector x(xx);
x.sort();
return x;')
x <- c(a=1, b=2, c=0)
f(x)
sort(x)
I would love to have this available, but I am no
On 26 June 2014 at 16:01, Michael Hannon wrote:
| Greetings. I've been trying to get familiar with RcppArmadillo and RcppEigen.
| As my "Hello, world!" example I'm using matrix multiplication. I got the
| Armadillo version to work, but I'm having trouble with the Eigen version.
|
| I'm trying t
11 matches
Mail list logo
