CPU-parallelization

DESCRIPTION

@@ -28,7 +28,7 @@ Imports:
   R6 (>= 2.4.1)
 License: GPL-3
 Encoding: UTF-8
-RoxygenNote: 7.3.2
+RoxygenNote: 7.3.3
 URL: https://github.com/MarselScheer/bootGOF
 BugReports: https://github.com/MarselScheer/bootGOF/issues
 Suggests: 

R/GOF_model.R

@@ -1,3 +1,5 @@
+library(parallel)
+
 ##' @title Convenience function for creating a GOF-test for statistical models
 ##'
 ##' @description Simplifies the creation of an instance of
@@ -18,6 +20,13 @@
 ##'   class used for performing the GOF test (\link{GOF_model_test})
 ##'   is injected. This parameter simply makes it easier to test the
 ##'   convenience function properly.
+##' @param n_cores positive integer specifying the number of CPU cores to use
+##'   for parallel resampling. If bigger than 1, the L'Ecuyer-CMRG is used;
+##'   if 'NULL' or 1, one core is used with the current RNG.
+##'   Default is "NULL".
+##' @param seed integer intended to seed the internally setup
+##'   L'Ecuyer-CMRG, but will also be applied when RNG not replaced.
+##'   Default is "NULL", which will not alter the seed.
 ##' @export
 ##' @return instance of \link{GOF_model_test}
 ##' @examples
@@ -54,7 +63,9 @@ GOF_model <- function(model, # nolint
                       y_name,
                       Rn1_statistic, # nolint
                       gof_model_resample_class = GOF_model_resample,
-                      gof_model_test_class = GOF_model_test
+                      gof_model_test_class = GOF_model_test,
+                      n_cores = NULL,
+                      seed = NULL
                       ) {
   checkmate::assert_subset(
     x = simulator_type,
@@ -67,8 +78,6 @@ GOF_model <- function(model, # nolint
     ))
   }
 
-
-
   simulators <- list(
     lm = list(
       parametric = GOF_lm_sim_param,
@@ -113,6 +122,9 @@ GOF_model <- function(model, # nolint
     y_name = y_name,
     Rn1_statistic = Rn1_statistic,
     gof_model_info_extractor = mie,
-    gof_model_resample = model_resample)
+    gof_model_resample = model_resample,
+    n_cores = n_cores,
+    seed = seed)
+
   return(ret)
 }

R/GOF_model_test.R

@@ -22,22 +22,33 @@ GOF_model_test <- R6::R6Class( # nolint
     ##' @param gof_model_resample an instance that implements
     ##'   \link{GOF_model_resample} in order to apply it to
     ##'   \code{model}
+    ##' @param n_cores positive integer specifying the number of CPU cores to
+    ##'   use for parallel resampling. If bigger than 1, the L'Ecuyer-CMRG is
+    ##'   used; if 'NULL' or 1, one core is used with the current RNG.
+    ##' @param seed integer intended to seed the internally setup
+    ##'   L'Ecuyer-CMRG, but will also be applied when RNG not replaced,
+    ##'   as long as it is not "NULL".
     ##' @return An instance of the Class
     initialize = function(model,
                           data,
                           nmb_boot_samples,
                           y_name,
                           Rn1_statistic, # nolint
                           gof_model_info_extractor,
-                          gof_model_resample) {
+                          gof_model_resample,
+                          n_cores,
+                          seed) {
       checkmate::assert_count(x = nmb_boot_samples, positive = TRUE)
+      checkmate::assert_count(x = n_cores, positive = TRUE, null.ok = TRUE)
       private$model_org <- model
       private$data_org <- data
       private$y_name <- y_name
       private$Rn1_statistic <- Rn1_statistic # nolint
       private$nmb_boot_samples <- nmb_boot_samples
       private$model_info_extractor <- gof_model_info_extractor
       private$model_resample <- gof_model_resample
+      private$n_cores <- n_cores
+      private$seed <- seed
       private$order_beta_dot_X_org <- order( # nolint
         private$model_info_extractor$beta_x_covariates(
           model = private$model_org
@@ -84,6 +95,8 @@ GOF_model_test <- R6::R6Class( # nolint
     nmb_boot_samples = NULL,
     model_info_extractor = NULL,
     model_resample = NULL,
+    n_cores = NULL,
+    seed = NULL,
     Rn1_statistic = NULL,
     Rn1_boot = NULL,
     Rn1_org = NULL,
@@ -108,5 +121,32 @@ GOF_model_test <- R6::R6Class( # nolint
           order_beta_x_covariates = private$order_beta_dot_X_org)
         return(Rn1_boot)
       }
-      private$Rn1_boot <- lapply(X = 1:private$nmb_boot_samples, FUN = f) # nolint
+
+      # Replace RNG with "L'Ecuyer-CMRG" if going parallel
+      replaced_rng <- FALSE
+      if (is.null(private$n_cores)) {
+        private$n_cores <- 1
+      } else if (private$n_cores > 1) {
+        # save and replace current RNG state
+        original_state <- if (exists(".Random.seed", .GlobalEnv))
+            .GlobalEnv$.Random.seed else NULL # nolint
+        RNGkind("L'Ecuyer-CMRG")
+        set.seed(NULL)
+        replaced_rng <- TRUE
+      }
+
+      if (!is.null(private$seed)) {
+        set.seed(private$seed)
+      }
+
+      private$Rn1_boot <- parallel::mclapply(X = 1:private$nmb_boot_samples, FUN = f, mc.cores = private$n_cores) # nolint
+
+      # Reset initial RNG if it has been replaced
+      if (replaced_rng) {
+        if (!is.null(original_state)) {
+          .GlobalEnv$.Random.seed <- original_state
+        } else {
+          RNGkind("default")
+        }
+      }
     }))

README.Rmd

@@ -81,6 +81,44 @@ mt$get_pvalue()
 
 ...
 
+## Parallelization
+
+The bootstrapping process can be accelerated using the `n_cores` parameter of
+the `GOF_model` function, that specifies the number of CPU cores to use.
+
+If this parameter is set to at least two cores, internally the currently used
+RNG is replaced by the L'Ecuyer-CMRG generator, which is safe to use
+in a parallel context.
+
+This internal generator can be seeded using the `seed` parameter of the
+`GOF_model` function, which however will also apply if the `n_cores` parameter
+is not used and thus the generator is not replaced.
+
+For example:
+```{r}
+set.seed(1)
+N <- 100
+X1 <- rnorm(N)
+X2 <- rnorm(N)
+d <- data.frame(
+  y = rpois(n = N, lambda = exp(4 + X1 * 2 + X2 * 6)),
+  x1 = X1,
+  x2 = X2)
+
+fit <- glm(y ~ x1 + x2, data = d, family = poisson())
+
+mt <- GOF_model(
+  model = fit,
+  data = d,
+  nmb_boot_samples = 100,
+  simulator_type = "parametric",
+  y_name = "y",
+  Rn1_statistic = Rn1_KS$new(),
+  n_cores = 2,
+  seed = 1)
+mt$get_pvalue()
+```
+
 ## Installation
 
 You can install it from CRAN

README.md

@@ -68,6 +68,45 @@ not be rejected by the GOF-test:
 
 …
 
+## Parallelization
+
+The bootstrapping process can be accelerated using the `n_cores`
+parameter of the `GOF_model` function, that specifies the number of CPU
+cores to use.
+
+If this parameter is set to at least two cores, internally the currently
+used RNG is replaced by the L’Ecuyer-CMRG generator, which is safe to
+use in a parallel context.
+
+This internal generator can be seeded using the `seed` parameter of the
+`GOF_model` function, which however will also apply if the `n_cores`
+parameter is not used and thus the generator is not replaced.
+
+For example:
+
+    set.seed(1)
+    N <- 100
+    X1 <- rnorm(N)
+    X2 <- rnorm(N)
+    d <- data.frame(
+      y = rpois(n = N, lambda = exp(4 + X1 * 2 + X2 * 6)),
+      x1 = X1,
+      x2 = X2)
+
+    fit <- glm(y ~ x1 + x2, data = d, family = poisson())
+
+    mt <- GOF_model(
+      model = fit,
+      data = d,
+      nmb_boot_samples = 100,
+      simulator_type = "parametric",
+      y_name = "y",
+      Rn1_statistic = Rn1_KS$new(),
+      n_cores = 2,
+      seed = 1)
+    mt$get_pvalue()
+    #> [1] 0.62
+
 ## Installation
 
 You can install it from CRAN
@@ -89,9 +128,9 @@ package in your environment by calling:
 # sessionInfo
 
     sessionInfo()
-    #> R Under development (unstable) (2025-08-19 r88650)
+    #> R Under development (unstable) (2025-10-19 r88945)
     #> Platform: x86_64-pc-linux-gnu
-    #> Running under: Ubuntu 24.04.2 LTS
+    #> Running under: Ubuntu 24.04.3 LTS
     #> 
     #> Matrix products: default
     #> BLAS:   /usr/lib/x86_64-linux-gnu/openblas-pthread/libblas.so.3 
@@ -109,15 +148,15 @@ package in your environment by calling:
     #> tzcode source: system (glibc)
     #> 
     #> attached base packages:
-    #> [1] stats     graphics  grDevices datasets  utils     methods   base     
+    #> [1] parallel  stats     graphics  grDevices utils     datasets  methods  
+    #> [8] base     
     #> 
     #> other attached packages:
-    #> [1] bootGOF_0.1.1
+    #> [1] bootGOF_0.1.1.9000
     #> 
     #> loaded via a namespace (and not attached):
     #>  [1] digest_0.6.37     desc_1.4.3        backports_1.5.0   R6_2.6.1         
     #>  [5] fastmap_1.2.0     xfun_0.53         knitr_1.50        htmltools_0.5.8.1
-    #>  [9] rmarkdown_2.29    cli_3.6.5         renv_1.1.5        withr_3.0.2      
-    #> [13] pkgload_1.4.0     compiler_4.6.0    rprojroot_2.1.0   tools_4.6.0      
-    #> [17] pkgbuild_1.4.8    checkmate_2.3.3   evaluate_1.0.4    yaml_2.3.10      
-    #> [21] rlang_1.1.6
+    #>  [9] rmarkdown_2.30    cli_3.6.5         pkgload_1.4.1     compiler_4.6.0   
+    #> [13] rprojroot_2.1.1   tools_4.6.0       pkgbuild_1.4.8    checkmate_2.3.3  
+    #> [17] evaluate_1.0.5    yaml_2.3.10       rlang_1.1.6

inst/tinytest/test_GOF_model.R

@@ -33,7 +33,9 @@ GOF_model_test_dummy <- R6::R6Class( # nolint
                           y_name,
                           Rn1_statistic, # nolint
                           gof_model_info_extractor,
-                          gof_model_resample) {
+                          gof_model_resample,
+                          n_cores,
+                          seed) {
     }))
 
 GOF_model_error_if_fit_class_is_not_lm_or_glm <- function() { # nolint
@@ -86,7 +88,9 @@ GOF_model_uses_lm_info_extractor <- function() { # nolint
                             y_name,
                             Rn1_statistic, # nolint
                             gof_model_info_extractor,
-                            gof_model_resample) {
+                            gof_model_resample,
+                            n_cores,
+                            seed) {
         inject_lm_info_extractor <<- inherits(
           x = gof_model_info_extractor,
           what = "GOF_lm_info_extractor")
@@ -209,7 +213,9 @@ GOF_model_uses_glm_info_extractor <- function() { # nolint
                             y_name,
                             Rn1_statistic, # nolint
                             gof_model_info_extractor,
-                            gof_model_resample) {
+                            gof_model_resample,
+                            n_cores,
+                            seed) {
         inject_glm_info_extractor <<- inherits(
           x = gof_model_info_extractor,
           what = "GOF_glm_info_extractor")