rcpp - solving compiled ODE in desolve package in R -

i solving ode's in r using desolve package. in order speed calculations want use compiled code, using instructions here

i showing below example of ode system - coding using rcpp below. details of ode system taken example matlab code (can found here). wanted simulate non-trivial set of ode in order see difference in r , compile code. following drive file, calculated mass-balances in 2 different ways (only r , compiled code)

library(desolve) library(ggplot2) library(microbenchmark)  source('parameters_gprotein.r') p   <- parameters()  source('ic_gprotein.r') ic  <- initial_conditions()  time = seq(from = 0, = 600)  source('odes_gprotein.r') sim.data.df  <- as.data.frame(vode(ic,time,ode_gprotein,p,                                     mf = 22, rtol=1e-3,atol=1e-6,maxord = 5,                                     verbose = f))  rcpp::sourcecpp("odes_gprotein.cpp") sim.data.df <- as.data.frame(vode(ic,time,odes_gprotein,p,                                       mf = 22, rtol = 1e-3, atol = 1e-6, maxord = 5,                                       verbose = f)) 

my question since vode call made in r. mean equations solved in compiled code if mass balances formed in cpp , speed gains realized, or have make vode call in cpp file.

certainly, microbenchmarking results show there speed gain when using odes_gprotein.cpp

unit: milliseconds                                                                                                                                expr   sim.data.df1 <- as.data.frame(vode(ic, time, ode_gprotein, p,      mf = 22, rtol = 0.001, atol = 1e-06, maxord = 5, verbose = f))  sim.data.df2 <- as.data.frame(vode(ic, time, odes_gprotein, p,      mf = 22, rtol = 0.001, atol = 1e-06, maxord = 5, verbose = f))          min        lq      mean    median        uq      max neval  27.801954 29.543624 31.213758 30.565434 31.399140 86.28537   100   8.188846  8.577824  9.177491  8.817025  9.437214 18.94304   100 

thanks

reproducible example

in future, please link / provide written code question based on.

comparisons

when writing comparison of r , rcpp function code should append end of c++ function name _cpp call different implementations clear. e.g. ode_gprotein() or odes_gprotein() c++ implementation?

in case of benchmark, appears if odes_gprotein() cpp() call since microbenchmark lower. means vode() function calling rcpp wrapper introduced namespace via rcpp::sourcecpp.

gaining speed

if possible, code should embedded within cpp greatest gain. means may have implementation scratch if speed huge factor (e.g. working large data or performing computationally intensive simulations). recommendation comes result of decreasing relay between r code front , c++ code. in essence, when c++ code called r objects must copied , reformatted different objects behind scenes before can worked upon. particularly case if using objects outside of default rcpp types (e.g. rcpparmadillo, rcppgsl, et cetera) copy must happen r's object structure importing structure.

what's happening in rcpp::sourcecpp (behind scenes)

let's dive little bit deeper happening behind rcpp::sourcecpp(). specifically, when method called different attributes extracted (e.g. // [[rcpp::]]tags) , options enabled.

the common tag // [[rcpp::export] creates wrapper or abstraction layer around c++ code can work r objects. behavior important 3 reasons: 1. not need worry sexp types , memory protection, 2. automatic creation of r level function default .call("fname"), , 3. code caches enable compiled code recompiled on change.

thus, when call odes_gprotein() calling c++ function.