R/plot_past.R

In csgillespie/benchmarkme-data: Data Set for the 'benchmarkme' Package

globalVariables(c("cpu", "n_time", "past_results_v2", "time", "sysname", "ram"))
#' Scatter plot of past benchmarks
#'
#' Plot the previous benchmarks. This function creates two figures.
#' \itemize{
#' \item Figure 1: Total benchmark time over all benchmarks (in seconds) on the y-axis.
#' \item Figure 2: Relative time (compared to the smallest benchmark).
#' }
#' The data set used is \code{data(past_results_v2)}.
#' @param test_group One of "prog", "matrix_fun", "matrix_cal", "read5", "read50", "read200",
#' "write5", "write50" or "write200". Default value \code{prog}.
#' @param blas_optimize Default \code{NULL}. The default behaviour is to plot all results.
#' To plot only the BLAS optimized results, set to \code{TRUE}, otherwise \code{FALSE}.
#' @param cores Default \code{0}, i.e. no parallel.
#' @param log By default the y axis is plotted on the log scale. To change, set the
#' the argument equal to the empty parameter string, \code{""}.
#' @importFrom graphics abline grid par plot points legend
#' @importFrom grDevices palette rgb
#' @importFrom utils data
#' @importFrom stats aggregate
#' @import dplyr
#' @export
#' @examples
#' ## Plot all past results for the `prog` benchmark
#' plot_past("prog", blas_optimize = NULL)
plot_past = function(test_group,
                     blas_optimize = NULL,
                     cores = 0,
                     log = "y") {

  if (missing(test_group) || !(test_group %in% get_benchmarks())) {
    stop("test_group should be one of\n\t",
         get_benchmarks(collapse = TRUE),
         call. = FALSE)
  }

  results = select_results(test_group, blas_optimize = blas_optimize,
                           cores = cores)

  ## Arrange plot colours and layout
  op = par(mar = c(3, 3, 2, 1),
           mgp = c(2, 0.4, 0), tck = -.01,
           cex.axis = 0.8, las = 1, mfrow = c(1, 2))
  old_pal = palette()
  on.exit({
    palette(old_pal)
    par(op)
  })
  nice_palette()

  ymin = min(results$time)
  ymax = max(results$time)
  plot(results$time, xlab = "Rank", ylab = "Total timing (secs)",
       ylim = c(ymin, ymax), xlim = c(1, nrow(results) + 1), cex = 0.9,
       panel.first = grid(), log = log, pch = 21, bg = as.numeric(results$test_group))

  ## Relative timings
  fastest = min(results$time)
  ymax = ymax / fastest
  plot(results$time / fastest, xlab = "Rank", ylab = "Relative timing",
       ylim = c(1, ymax), xlim = c(1, nrow(results) + 1), cex = 0.9,
       panel.first = grid(), log = log, pch = 21, bg = as.numeric(results$test_group))
  abline(h = 1, lty = 3)

  invisible(results)
}