Xsimd C++ Header-Only Library Files

This header-only library provides modern, portable C++ wrappers for SIMD intrinsics and parallelized, optimized math implementations (SSE, AVX, NEON, AVX512). By placing this library in this package, we offer an efficient distribution system for Xsimd < https://github.com/xtensor-stack/xsimd> for R packages using CRAN.

RcppXsimd

Overview

This library provides an Rcpp wrapper around Xsimd that is a modern, portable C++14 header-only library for SIMD intrinsics and parallelized, optimized math implementations.

Example

library(RcppXsimd)
?supportsAVX

if (supportsAVX()) {
  Sys.setenv(PKG_CPPFLAGS = "-mavx")
  Rcpp::sourceCpp(verbose = TRUE, code='
    // [[Rcpp::plugins(cpp14)]]
    // [[Rcpp::depends(RcppXsimd)]]

    #include <Rcpp.h>
    #include "xsimd/xsimd.hpp"

    // [[Rcpp::export]]
    void demoAVX() {
      xsimd::batch<double, 4> a(1.0);
      xsimd::batch<double, 4> b(1.0);
      Rcpp::Rcout << a << " + " << b << " = " << (a + b) << std::endl;
    }')
  demoAVX()
} else {
  message("AVX is not supported")
}

Development status

RcppXsimd is being developed in R Studio.

Acknowledgements

• This project is supported in part through the National Science Foundation grant DMS 1264153.