## ----packages, results='hide', message=FALSE----------------------------------
library("clrng")

## ----streams, eval=TRUE-------------------------------------------------------
get_system_info()
if (detectGPUs()) {
  setContext(grep("gpu", listContexts()$device_type)[1])
    ## check gpu information
    currentDevice()
    
    ## create 4 streams on CPU as R matrix (with package's default initial seed)
    streamsonCpu <- createStreamsCpu(n=4)
    
    ## Important: streams are of type integer in R
    typeof(streamsonCpu) 
    
    ## Attention: when converting streams from CPU to GPU, 
    ## should set type = "integer", or leave it as default `NULL' as below
    t(as.matrix(vclMatrix(streamsonCpu)))
    t(as.matrix(2*vclMatrix(streamsonCpu)))
    t(as.matrix(2*vclMatrix(streamsonCpu, type="integer"))) 
    
    type = c('float','double')[1+gpuR::deviceHasDouble()]
    ## setting streams as "double" or other type can cause problems, see the following 
    t(as.matrix(2*vclMatrix(streamsonCpu, type=type))) 
    t(as.matrix(vclMatrix(2*streamsonCpu)))
    
    ## continue to create 6 streams on GPU
    streamsonGpu <- createStreamsGpu(n=6)
    t(as.matrix(streamsonGpu))
    
    ## save the created streams as .rds object on CPU 
    saveRDS(as.matrix(createStreamsCpu(n = 4)), "myStreams.rds")
    
    ## load saved streams
    streams_saved <- vclMatrix(readRDS("myStreams.rds"))

} else {
   message("No GPU detected. Skipping GPU-dependent code.")
}