anvl

Package website: release | dev

Composable code transformation framework for R, allowing you to run numerical programs at the speed of light. It currently implements JIT compilation for very fast execution and reverse-mode automatic differentiation. Programs can run on various hardware backends, including CPU and GPU.

Installation

{anvl}¹ can be installed from GitHub or r-universe. During runtime, we require libprotobuf. Source installation requires a C++20 compiler and protoc (protobuf compiler).

You can install the latest release from r-universe (prebuilt binary) or GitHub:

install.packages("anvl", repos = c("https://r-xla.r-universe.dev", getOption("repos")))
pak::pak("r-xla/anvl@*release")

To get the latest dev version, run:

pak::pak("r-xla/anvl")

In order to install CUDA support (linux amd64 only), requiring only a compatible driver, run:

install.packages("cuda12.8", repos = "https://mlverse.r-universe.dev")

Prebuilt Docker images are also available. See the Installation vignette on the package website for more details.

Quick Start

We define an R function operating on AnvlArrays – the primary data type of {anvl}. It can be executed in either eager mode (each operation is performed immediately) or jit mode (the whole function is compiled into a single executable via jit()).

library(anvl)
f <- function(a, b, x) {
  a * x + b
}

a <- nv_scalar(1.0, "f32")
b <- nv_scalar(-2.0, "f32")
x <- nv_scalar(3.0, "f32")

# Eager mode
f(a, b, x)
#> AnvlArray
#>  1
#> [ CPUf32{} ]

# JIT mode
f_jit <- jit(f)
f_jit(a, b, x)
#> AnvlArray
#>  1
#> [ CPUf32{} ]

Through automatic differentiation, we can also obtain the gradient of the above function.

g_jit <- jit(gradient(f, wrt = c("a", "b")))
g_jit(a, b, x)
#> $a
#> AnvlArray
#>  3
#> [ CPUf32{} ] 
#> 
#> $b
#> AnvlArray
#>  1
#> [ CPUf32{} ]

For more complex examples, such as implementing a Gaussian Process, see the package website.

Main Features

• Automatic Differentiation:
  • Gradients for functions with scalar outputs are supported.
• Eager and JIT mode:
  • Run operations eagerly for interactive use, or jit() entire functions into a single fused kernel for performance.
• Fast:
  • Code is JIT compiled into a single kernel.
  • Runs on different hardware backends, including CPU and GPU.
  • Asyncronous allocation and execution, allowing the accelerator to do it’s job while R interpretes.
• Extendable:
  • It is possible to add new primitives, transformations, and (with some effort) new backends.
  • The package is written almost entirely in R.
• Multi-backend:
  • The backend supports execution via XLA as well as an experimental {quickr}-based Fortran backend (CPU only).

Platform Support

• Linux (x86_64)
  • ✅ CPU backend is fully supported.
  • ✅ CUDA (NVIDIA GPU) backend is fully supported.
• Linux (ARM)
  • ✅ CPU backend is fully supported.
  • ❌ GPU is not supported.
• Windows
  • ✅ CPU backend is fully supported.
  • ⚠️ GPU is only supported via Windows Subsystem for Linux (WSL2).
• macOS (ARM)
  • ✅ CPU backend is supported.
  • ⚠️ Metal (Apple GPU) backend is available but not fully functional.
• macOS (x86_64)
  • ❌ Not supported.

Acknowledgments

• This work is supported by MaRDI.
• The design of this package was inspired by and borrows from:
  • JAX, especially the autodidax tutorial.
  • The microjax project.
• For JIT compilation, we leverage the OpenXLA project.

Footnotes

1. A real anvil is a tool blacksmiths use to reshape metal; this package reshapes R code in a similar spirit. We spell the package name anvl (without the i) because AnVIL is already taken by a Bioconductor package. ↩