R-Friendly Threading in C++

Provides a C++11-style thread class and thread pool that can safely be interrupted from R. See Nagler (2021) .

RcppThread

Provides R-friendly threading functionality:

• thread safe versions of Rcpp's checkUserInterrupt(), Rcout, and Rcerr,
• an interruptible thread class that otherwise behaves like std::thread,
• classes for the thread pool pattern and parallel for loops for easy and flexible parallelism,
• thread safe progress tracking,
• state-of-the art speed, see benchmarks.

The library is header-only, platform-independent, and only requires a C++11-compatible compiler.

Functionality

For a detailed description of its functionality and examples, see the associated JSS paper or the API documentation.

Since then, the following new features have been added:

• Printing to the error stream with Rcerr.

• Free-standing functions like parallelFor() now dispatch to a global thread pool that persists for the entire session. This significantly speeds up programs that repeatedly call these functions.

• Faster runtimes due to lock-free work stealing queue and loops (from quickpool).

• Option to resize a thread pool.

• An R function RcppThread::detectCores() to determine the number of (logical) cores on your machine.

• C++ classes ProgressCounter and ProgressBar for tracking progress in long-running loops.
  Example usage:

    // 20 iterations in loop, update progress every 1 sec
    RcppThread::ProgressBar bar(20, 1);
    RcppThread::parallelFor(0, 20, [&] (int i) {
        std::this_thread::sleep_for(std::chrono::milliseconds(200));
        bar++;
    });
  

  Output: (just one line that is continuously updated)

  ...
  Computing: [==========================              ] 65% (~1s remaining)       
  ...
  Computing: [========================================] 100% (done) 
  

Installation

Release version from CRAN:

install.packages("RcppThread")

Latest development version from github:

# install.packages("devtools")
devtools::install_github("tnagler/RcppThread")

How to use it

with cppFunction

Pass "RcppThread" to the depends argument and "cpp11" to the plugins argument. For example:

Rcpp::cppFunction('void func() { /* actual code here */ }', 
                       depends = "RcppThread", plugins = "cpp11")

with sourceCpp

Add

// [[Rcpp::plugins(cpp11)]]
// [[Rcpp::depends(RcppThread)]]

before including any headers in your source code.

in another R package

1. Add the line CXX_STD = CXX11 to the src/Makevars(.win) files of your package.
2. Add RcppThread to the LinkingTo field of your DESCRIPTION file.

For optimal portability, you might also want to add

PKG_LIBS = `"$(R_HOME)/bin/Rscript" -e "RcppThread::LdFlags()"`

to your src/Makevars (not .win). This adds -latomic/-lpthread flags as necessary and available.

Automatic override of std::cout, std::cerr, and std::thread

There are preprocessor options to replace all occurrences of std::cout, std::cerr, and std::thread with calls to RcppThread::Rcout, RcppThread::Rcerr, and RcppThread::Thread (provided that the RcppThread headers are included first). To enable this, use

#define RCPPTHREAD_OVERRIDE_COUT 1    // std::cout override
#define RCPPTHREAD_OVERRIDE_CERR 1    // std::cerr override
#define RCPPTHREAD_OVERRIDE_THREAD 1  // std::thread override

before including the RcppThread headers.

References

Nagler, T. (2021). "R-Friendly Multi-Threading in C++." Journal of Statistical Software, Code Snippets, 97(1), 1-18. doi: 10.18637/jss.v097.c01