intermediate-vectors.r

# What is the cost of intermediate vector allocation?
library(Rcpp)
library(microbenchmark)

cppFunction('
  NumericVector add2a(NumericVector a, NumericVector b) {
    int n = a.size();
    NumericVector out(n);
    for (int i = 0; i < n; ++i) {
      out[i] = a[i] + b[i];
    }
    return(out);
  }
')

cppFunction('
  NumericVector add3a(NumericVector a, NumericVector b, NumericVector c) {
    int n = a.size();
    NumericVector out(n);
    for (int i = 0; i < n; ++i) {
      out[i] = a[i] + b[i] + c[i];
    }
    return(out);
  }
')

cppFunction('
  NumericVector add4a(NumericVector a, NumericVector b, NumericVector c, NumericVector d) {
    int n = a.size();
    NumericVector out(n);
    for (int i = 0; i < n; ++i) {
      out[i] = a[i] + b[i] + c[i] + d[i];
    }
    return(out);
  }
')

add2b <- function(a, b) a + b
add3b <- function(a, b, c) a + b + c
add4b <- function(a, b, c, d) a + b + c + d

x <- runif(1e5)

stopifnot(
  all.equal(add2a(x, x), add2b(x, x)),
  all.equal(add3a(x, x, x), add3b(x, x, x)),
  all.equal(add4a(x, x, x, x), add4b(x, x, x, x))
)

print(microbenchmark(
  add2a(x, x),
  add3a(x, x, x),
  add4a(x, x, x, x),
  add2b(x, x),
  add3b(x, x, x),
  add4b(x, x, x, x),
  times = 1000
))
# Unit: microseconds
#               expr   min  lq median   uq   max neval
#        add2a(x, x) 140.4 153    161  173  1307  1000
#     add3a(x, x, x) 161.8 175    185  198 20921  1000
#  add4a(x, x, x, x) 191.4 207    217  232 20783  1000
#        add2b(x, x)  90.8 100    106  115 20949  1000
#     add3b(x, x, x) 192.9 215    229  909 21513  1000
#  add4b(x, x, x, x) 297.9 337    375 1148 21226  1000