add TODO + 2 more examples

Author: kevinushey

TODO

@@ -0,0 +1,10 @@
+Boost.SIMD
+==========
+
+- Make it easy for client packages to use Boost.SIMD (plugin? Also need LinkingTo: BH?)
+- Enable 'parallelFor', 'parallelReduce'-style calls to SIMD
+- See 'Wizard' (http://www.evanmiller.org/) for inspiration
+- Need to think about NA (idiom that handled NA)
+- Setting of compiler flags for CRAN
+- Don't overlap with Armadillo, Eigen
+

examples/boost-simd-capabilities.cpp

@@ -0,0 +1,37 @@
+// [[Rcpp::depends(RcppParallel)]]
+#include <RcppParallel.h>
+#include <Rcpp.h>
+using namespace Rcpp;
+
+#include <boost/simd/sdk/simd/extensions/meta/tags.hpp>
+#include <boost/simd/sdk/config/is_supported.hpp>
+#include <boost/config.hpp>
+
+// Try playing around with the '-march' flag to see what support is detected.
+// Note that most modern processors will implement instructions up to SSE3,
+// but certain AVX and FMA instructions are only available on newer CPUs.
+
+// [[Rcpp::export]]
+void simd_capabilities() {
+   
+   using boost::simd::is_supported;
+   using namespace boost::simd::tag;
+   
+   std::cout << "SIMD Capabilities\n"
+             << "=================\n\n";
+   
+   std::cout << "AVX2:   " << is_supported<avx2_>()   << "\n"
+             << "AVX:    " << is_supported<avx_>()    << "\n"
+             << "FMA4:   " << is_supported<fma4_>()   << "\n"
+             << "FMA3:   " << is_supported<fma3_>()   << "\n"
+             << "SSE4a:  " << is_supported<sse4a_>()  << "\n"
+             << "SSE4_2: " << is_supported<sse4_2_>() << "\n"
+             << "SSE4_1: " << is_supported<sse4_1_>() << "\n"
+             << "SSE3:   " << is_supported<sse3_>()   << "\n"
+             << "SSE2:   " << is_supported<sse2_>()   << "\n"
+             << "SSE:    " << is_supported<sse_>()    << "\n";
+}
+
+/***R
+simd_capabilities()
+*/

examples/boost-simd-transform.cpp

@@ -0,0 +1,44 @@
+// [[Rcpp::depends(RcppParallel)]]
+#include <RcppParallel.h>
+#include <Rcpp.h>
+using namespace Rcpp;
+
+#include <boost/simd/sdk/simd/algorithm.hpp>
+
+struct plus
+{
+   template <class T>
+   T operator()(const T& lhs, const T& rhs) const
+   {
+      return lhs + rhs;
+   }
+};
+
+// [[Rcpp::export]]
+NumericVector simd_add(NumericVector lhs,
+                       NumericVector rhs)
+{
+   NumericVector result = no_init(lhs.size());
+   boost::simd::transform(lhs.begin(), lhs.end(), rhs.begin(), result.begin(), plus());
+   return result;
+}
+
+// [[Rcpp::export]]
+NumericVector cpp_add(NumericVector lhs, NumericVector rhs)
+{
+   return lhs + rhs;
+}
+
+/***R
+n <- 1024 * 1000
+lhs <- rnorm(n)
+rhs <- rnorm(n)
+simd <- simd_add(lhs, rhs)
+cpp <- cpp_add(lhs, rhs)
+all.equal(simd, cpp)
+microbenchmark::microbenchmark(
+   simd = simd_add(lhs, rhs),
+   cpp  = cpp_add(lhs, rhs),
+   R    = lhs + rhs
+)
+*/