cmdstanr_benchmark.R

library(rethinking)
library(cmdstanr)
library(microbenchmark)

X <- data.frame(x = rnorm(100, 9, 2))

fit_mean <- function(cmdstan) {
  ulam(
    alist(
      x ~ normal( mu , Sigma ),
      mu ~ normal(0, 5),
      Sigma ~ exponential( 1 )
    ), data=X , chains=4 , cores=4 , cmdstan=cmdstan )
}

microbenchmark("cmdstan" = fit_mean(cmdstan = TRUE),
               "rstan" = fit_mean(cmdstan = FALSE),
               times = 10)

# Unit: seconds
#     expr       min        lq      mean    median       uq      max neval
#  cmdstan 14.811175 14.981982 15.402317 15.177852 15.90549 16.54465    10
#    rstan  3.722226  3.805964  9.465478  3.900478 21.54825 22.83466    10