See how parallel computing is working in practice vs optimal speedup

parallel_analysis.R

library(furrr)

x <- numeric()

for (i in 1:20) {
  future::plan(multisession, workers = i)
  result <- system.time({
    r <- furrr::future_map(1:100,
                           ~ sum(rnorm(1e7)),
                           .options = furrr_options(seed=TRUE))
  })
  print(result[[3]])
  x <- c(x, result[[3]])
}

# PLOT
plot(1:20, x, type = "b", col = "blue", pch = 16, lty = 1, ylim = range(c(x, x / (1:20))),
     xlab = "Index", ylab = "Values", main = "Actual vs Optimal")

lines(1:20, x / (1:20), type = "b", col = "red", pch = 17, lty = 2)

legend("topright", legend = c("x", "x / (1:20)"), col = c("blue", "red"), 
       pch = c(16, 17), lty = c(1, 2))