NEWS | R Documentation |
Use the SHR3
generator for the default implementation
just like Leong et al do, making our default implementation
identical to theirs (but 32- and 64-bit compatible)
Switched generators from float
to double
ensuring that results are identical on 32- and 64-bit platforms
Simplified builds with respect to GSL use via the RcppGSL package; added a seed setter for the GSL variant
Corrected use of fabs()
to abs()
on integer
variables, with a grateful nod to Brian Ripley for the hint (based
on CRAN checks on the beloved Slowlaris machines)
Accelerated Travis CI tests by relying exclusively on
r-cran-*
packages from the PPAs by Michael Rutter and myself
Updated DESCRIPTION
and NAMESPACE
according to
current best practices, and R-devel CMD check --as-cran
checks
Add a new generator deploying R's unif_rand which faciliates use of Ziggurat as the user-supplied generator for N(0,1)
Update a 'local' demo comparing normal distribution RNGs from Boost, C++11 and Armadillo none of which are particularly speedy
Add declaration to import a symbol from Rcpp to
NAMESPACE
to ensure proper instantiation with the upcoming
Rcpp version
Set dependency on R (>= 3.0.0) needed by vignette builder
Remove pdf files in vignette on cleanup to ensure rebuilds
Testing framework added with a standard RNG test, a new and similar test particular for Normal distributions, and the Chi^2 test from the Leong et al paper (JSS, 2005)
New demo()
functionality showing the three added tests.
Code has been reorganized into separate headers implementing
the generators, using a common (virtual) base class; and accessors
and tests in the src/
directory.
Two new ziggurat generators implementations have been adapted from, respectively, GNU Gretl and QuantLib
Added a vignette
Initial basic package version, wrapping five different implementations