>> benchmark(fun1(), fun2(), fun3(), fun4(), order = "relative", > replications = 1e5L) > test replications elapsed relative user.self sys.self user.child > sys.child > 4 fun4() 100000 6.16 1.000 6.06 0 NA > NA > 3 fun3() 100000 6.19 1.005 5.91 0 NA > NA > 2 fun2() 100000 6.92 1.123 6.75 0 NA > NA > 1 fun1() 100000 13.01 2.112 12.73 0 NA > NA
An interesting thread, thanks for the benchmarks. I just wanted to point out that even with 100,000 replications, noise is altering the ordering: fun3() is quicker when considering user.self instead of elapsed. (Are there any ways, built-in to benchmark, to avoid this? E.g. do multiple runs, and average, or just use quickest run, etc.?) Darren _______________________________________________ Rcpp-devel mailing list [email protected] https://lists.r-forge.r-project.org/cgi-bin/mailman/listinfo/rcpp-devel
