[Rcpp-devel] Rcpp ISNAN slower than C ISNAN?

[Romain Francois]

I’d be interesting to see what this more C++idomatic version would perform 

nanCount = std::count_if( x.begin(), x.end(), ISNAN ) ;

Because despite all the inlining efforts that has been put in the implementation of NumericVector, its operator[] might not perform as well as using [] on a double*. 

Romain


> Le 13 déc. 2016 à 12:38, Johannes Kruisselbrink <johannes at kruisselbrink.eu> a écrit :
> 
> Here is a further reduced example (see below). Now it is a function to count NaNs in a vector, and it shows the same behaviour. Code is also available at https://github.com/jwkruisselbrink/rcpp-timings/tree/master/minimal.
> 
> Regarding your question:
> | Why not drop data and codes and use  sData1(i,k) - sData2(j,k)
> 
> Well, I wanted to rule out that Rcpp sugar was causing the slowdown. If it isn't we'll put it back in after having a fix for this issue.
> 
> /=============== call.c ===============/
> 
> #include <R.h>
> #include <Rinternals.h>
> 
> SEXP CountNans(SEXP sX, SEXP sLength) {
>  int i, n = *INTEGER(sLength);
>  int *nanCount;
>  double *x = REAL(sX);
> 
>  SEXP output = PROTECT(allocVector(INTSXP, 1));
>  nanCount = INTEGER(output);
>  *nanCount = 0;
>  for (i = 0; i < n; i++) {
>    if (ISNAN(x[i])) {
>      *nanCount += 1;
>    }
>  }
>  UNPROTECT(1);
>  return output;
> }
> 
> /=============== rcpp.cpp ===============/
> 
> #include <R.h>
> #include <Rcpp.h>
> using namespace Rcpp;
> 
> // [[Rcpp::export]]
> int CountNans(NumericVector x) {
>  int n = x.length();
>  int nanCount = 0;
>  for (int i = 0; i < n; i++) {
>    if (ISNAN(x[i])) {
>      nanCount++;
>    }
>  }
>  return nanCount;
> }
> 
> /=============== R code ===============/
> 
> library(Rcpp)
> library(microbenchmark)
> library(ggplot2)
> 
> sourceCpp('rcpp.cpp')
> 
> if (is.loaded(paste("call", .Platform$dynlib.ext, sep=""))) {
>  dyn.unload(paste("call", .Platform$dynlib.ext, sep=""))
> }
> system(paste("R CMD SHLIB call.c", sep=""))
> dyn.load(paste("call", .Platform$dynlib.ext, sep=""))
> 
> n <- as.integer(100000)
> x <- rnorm(n)
> mb <- microbenchmark(
>  rcpp <- CountNans(x),
>  call <- .Call("CountNans", x, n),
>  times = 10000
> )
> 
> autoplot(mb)
> 
> 
> _______________________________________________
> Rcpp-devel mailing list
> Rcpp-devel at lists.r-forge.r-project.org
> https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel