Merge pull request #321 from mrc-ide/i298-gpu-everywhere

.gitignore

@@ -16,7 +16,7 @@ inst/doc
 pkgdown
 inst/include/cub
 *.gcov
-vignettes_src/cuda.md
+vignettes_src/gpu.md
 
 .vscode/
 

DESCRIPTION

@@ -1,6 +1,6 @@
 Package: dust
 Title: Iterate Multiple Realisations of Stochastic Models
-Version: 0.11.4
+Version: 0.11.5
 Authors@R: c(person("Rich", "FitzJohn", role = c("aut", "cre"),
                     email = "[email protected]"),
              person("John", "Lees", role = "aut"),

Makefile

@@ -48,8 +48,8 @@ clean:
 		tests/testthat/example/*.o tests/testthat/example/*.so \
 		src/*.gcov src/*.gcda src/*.gcno
 
-vignettes/cuda.Rmd: vignettes_src/cuda.Rmd
-	./scripts/build_cuda_vignette
+vignettes/gpu.Rmd: vignettes_src/gpu.Rmd
+	./scripts/build_gpu_vignette
 
 vignettes: vignettes/dust.Rmd vignettes/rng.Rmd
 	${RSCRIPT} -e 'tools::buildVignettes(dir = ".")'

NEWS.md

@@ -1,3 +1,8 @@
+# dust 0.11.5
+
+* Major header reorganisation, the `dust::cuda` namespace is now `dust::gpu`, most user-facing uses of `cuda` and `device` now replaced with `gpu` (#298, #317)
+* New (static) method `real_size` on every dust class for accessing the size of `real_type`, in bits (#301)
+
 # dust 0.11.1
 
 * Support for the multinomial distribution. This differs from most other random number support because it returns a *vector* of values not a single number (#307)

R/compile.R

@@ -63,7 +63,7 @@ dust_code <- function(data, config) {
   if (config$has_gpu_support) {
     data_gpu <- data
     data_gpu$target <- "gpu"
-    data_gpu$container <- "DustDevice"
+    data_gpu$container <- "dust_gpu"
     dust_cpp <- c(dust_cpp,
                   substitute_dust_template(data_gpu, "dust_methods.cpp", NULL))
     dust_hpp <- c(dust_hpp,
@@ -155,7 +155,7 @@ dust_template_data <- function(model, config, cuda) {
        param = deparse_param(config$param),
        cuda = cuda$flags,
        target = "cpu",
-       container = "Dust",
+       container = "dust_cpu",
        has_gpu_support = as.character(config$has_gpu_support),
        methods_cpu = methods_cpu,
        methods_gpu = methods_gpu)

R/cpp11.R

@@ -76,12 +76,12 @@ dust_sir_capabilities <- function() {
   .Call(`_dust_dust_sir_capabilities`)
 }
 
-dust_sir_device_info <- function() {
-  .Call(`_dust_dust_sir_device_info`)
+dust_sir_gpu_info <- function() {
+  .Call(`_dust_dust_sir_gpu_info`)
 }
 
-dust_cpu_sir_alloc <- function(r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, device_config) {
-  .Call(`_dust_dust_cpu_sir_alloc`, r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, device_config)
+dust_cpu_sir_alloc <- function(r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, gpu_config) {
+  .Call(`_dust_dust_cpu_sir_alloc`, r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, gpu_config)
 }
 
 dust_cpu_sir_run <- function(ptr, step_end) {
@@ -148,12 +148,12 @@ dust_sirs_capabilities <- function() {
   .Call(`_dust_dust_sirs_capabilities`)
 }
 
-dust_sirs_device_info <- function() {
-  .Call(`_dust_dust_sirs_device_info`)
+dust_sirs_gpu_info <- function() {
+  .Call(`_dust_dust_sirs_gpu_info`)
 }
 
-dust_cpu_sirs_alloc <- function(r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, device_config) {
-  .Call(`_dust_dust_cpu_sirs_alloc`, r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, device_config)
+dust_cpu_sirs_alloc <- function(r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, gpu_config) {
+  .Call(`_dust_dust_cpu_sirs_alloc`, r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, gpu_config)
 }
 
 dust_cpu_sirs_run <- function(ptr, step_end) {
@@ -216,8 +216,8 @@ dust_cpu_sirs_n_state <- function(ptr) {
   .Call(`_dust_dust_cpu_sirs_n_state`, ptr)
 }
 
-dust_gpu_sirs_alloc <- function(r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, device_config) {
-  .Call(`_dust_dust_gpu_sirs_alloc`, r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, device_config)
+dust_gpu_sirs_alloc <- function(r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, gpu_config) {
+  .Call(`_dust_dust_gpu_sirs_alloc`, r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, gpu_config)
 }
 
 dust_gpu_sirs_run <- function(ptr, step_end) {
@@ -280,8 +280,8 @@ dust_gpu_sirs_n_state <- function(ptr) {
   .Call(`_dust_dust_gpu_sirs_n_state`, ptr)
 }
 
-test_cuda_pars <- function(r_device_config, n_particles, n_particles_each, n_state, n_state_full, n_shared_int, n_shared_real, data_size, shared_size) {
-  .Call(`_dust_test_cuda_pars`, r_device_config, n_particles, n_particles_each, n_state, n_state_full, n_shared_int, n_shared_real, data_size, shared_size)
+test_cuda_pars <- function(r_gpu_config, n_particles, n_particles_each, n_state, n_state_full, n_shared_int, n_shared_real, data_size, shared_size) {
+  .Call(`_dust_test_cuda_pars`, r_gpu_config, n_particles, n_particles_each, n_state, n_state_full, n_shared_int, n_shared_real, data_size, shared_size)
 }
 
 test_xoshiro_run <- function(name) {
@@ -296,12 +296,12 @@ dust_variable_capabilities <- function() {
   .Call(`_dust_dust_variable_capabilities`)
 }
 
-dust_variable_device_info <- function() {
-  .Call(`_dust_dust_variable_device_info`)
+dust_variable_gpu_info <- function() {
+  .Call(`_dust_dust_variable_gpu_info`)
 }
 
-dust_cpu_variable_alloc <- function(r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, device_config) {
-  .Call(`_dust_dust_cpu_variable_alloc`, r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, device_config)
+dust_cpu_variable_alloc <- function(r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, gpu_config) {
+  .Call(`_dust_dust_cpu_variable_alloc`, r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, gpu_config)
 }
 
 dust_cpu_variable_run <- function(ptr, step_end) {
@@ -364,8 +364,8 @@ dust_cpu_variable_n_state <- function(ptr) {
   .Call(`_dust_dust_cpu_variable_n_state`, ptr)
 }
 
-dust_gpu_variable_alloc <- function(r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, device_config) {
-  .Call(`_dust_dust_gpu_variable_alloc`, r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, device_config)
+dust_gpu_variable_alloc <- function(r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, gpu_config) {
+  .Call(`_dust_dust_gpu_variable_alloc`, r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, gpu_config)
 }
 
 dust_gpu_variable_run <- function(ptr, step_end) {
@@ -432,12 +432,12 @@ dust_volatility_capabilities <- function() {
   .Call(`_dust_dust_volatility_capabilities`)
 }
 
-dust_volatility_device_info <- function() {
-  .Call(`_dust_dust_volatility_device_info`)
+dust_volatility_gpu_info <- function() {
+  .Call(`_dust_dust_volatility_gpu_info`)
 }
 
-dust_cpu_volatility_alloc <- function(r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, device_config) {
-  .Call(`_dust_dust_cpu_volatility_alloc`, r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, device_config)
+dust_cpu_volatility_alloc <- function(r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, gpu_config) {
+  .Call(`_dust_dust_cpu_volatility_alloc`, r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, gpu_config)
 }
 
 dust_cpu_volatility_run <- function(ptr, step_end) {
@@ -504,12 +504,12 @@ dust_walk_capabilities <- function() {
   .Call(`_dust_dust_walk_capabilities`)
 }
 
-dust_walk_device_info <- function() {
-  .Call(`_dust_dust_walk_device_info`)
+dust_walk_gpu_info <- function() {
+  .Call(`_dust_dust_walk_gpu_info`)
 }
 
-dust_cpu_walk_alloc <- function(r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, device_config) {
-  .Call(`_dust_dust_cpu_walk_alloc`, r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, device_config)
+dust_cpu_walk_alloc <- function(r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, gpu_config) {
+  .Call(`_dust_dust_cpu_walk_alloc`, r_pars, pars_multi, step, r_n_particles, n_threads, r_seed, deterministic, gpu_config)
 }
 
 dust_cpu_walk_run <- function(ptr, step_end) {

R/cuda.R

@@ -1,12 +1,11 @@
-##' Detect CUDA configuration. This is experimental and subject to
-##' change. This function tries to compile a small program with `nvcc`
-##' and confirms that this can be loaded into R, then uses that
-##' program to query the presence and capabilities of your NVIDIA
-##' GPUs. If this works, then you can use the GPU-enabled dust
-##' features, and the information returned will help us.  It's quite
-##' slow to execute (several seconds) so we cache the value within a
-##' session.  Later versions of dust will cache this across sessions
-##' too.
+##' Detect CUDA configuration. This function tries to compile a small
+##' program with `nvcc` and confirms that this can be loaded into R,
+##' then uses that program to query the presence and capabilities of
+##' your NVIDIA GPUs. If this works, then you can use the GPU-enabled
+##' dust features, and the information returned will help us.  It's
+##' quite slow to execute (several seconds) so we cache the value
+##' within a session.  Later versions of dust will cache this across
+##' sessions too.
 ##'
 ##' Not all installations leave the CUDA libraries on the default
 ##' paths, and you may need to provide it. Specifically, when we link
@@ -165,7 +164,7 @@ cuda_configuration <- function(path_cuda_lib = NULL, path_cub_include = NULL,
     pkg <- pkgload::load_all(dat$path, export_all = FALSE, quiet = quiet,
                              helpers = FALSE, attach_testthat = FALSE)
     on.exit(pkgload::unload(dat$name))
-    info <- pkg$env$dust_device_info()
+    info <- pkg$env$dust_gpu_info()
     path_cub_include <-
       cuda_path_cub_include(info$cuda_version, path_cub_include)
     paths <- list(