The goal of quickr is to make your R code run quicker.
R is an extremely flexible and dynamic language, but that flexibility and dynamicism can come at the expense of speed. This package lets you trade back some of that flexibility for some speed, for the context of a single function.
The main exported function is quick(), here is how you use it.
library(quickr)
convolve <- quick(function(a, b) {
declare(type(a = double(NA)),
type(b = double(NA)))
ab <- double(length(a) + length(b) - 1)
for (i in seq_along(a)) {
for (j in seq_along(b)) {
ab[i+j-1] <- ab[i+j-1] + a[i] * b[j]
}
}
ab
})quick() returns a quicker R function. How much quicker? Let’s
benchmark it! For reference, we’ll also compare it to a
pure-C
implementation.
slow_convolve <- function(a, b) {
declare(type(a = double(NA)),
type(b = double(NA)))
ab <- double(length(a) + length(b) - 1)
for (i in seq_along(a)) {
for (j in seq_along(b)) {
ab[i+j-1] <- ab[i+j-1] + a[i] * b[j]
}
}
ab
}
library(quickr)
quick_convolve <- quick(slow_convolve)
convolve_c <- inline::cfunction(
sig = c(a = "SEXP", b = "SEXP"), body = r"({
int na, nb, nab;
double *xa, *xb, *xab;
SEXP ab;
a = PROTECT(Rf_coerceVector(a, REALSXP));
b = PROTECT(Rf_coerceVector(b, REALSXP));
na = Rf_length(a); nb = Rf_length(b); nab = na + nb - 1;
ab = PROTECT(Rf_allocVector(REALSXP, nab));
xa = REAL(a); xb = REAL(b); xab = REAL(ab);
for(int i = 0; i < nab; i++) xab[i] = 0.0;
for(int i = 0; i < na; i++)
for(int j = 0; j < nb; j++)
xab[i + j] += xa[i] * xb[j];
UNPROTECT(3);
return ab;
})")
a <- runif (100000); b <- runif (100)
timings <- bench::mark(
r = slow_convolve(a, b),
quickr = quick_convolve(a, b),
c = convolve_c(a, b),
min_time = 2
)
timings
#> # A tibble: 3 × 6
#> expression min median `itr/sec` mem_alloc `gc/sec`
#> <bch:expr> <bch:tm> <bch:tm> <dbl> <bch:byt> <dbl>
#> 1 r 1.22s 1.22s 0.819 782KB 0.819
#> 2 quickr 1.89ms 2.22ms 467. 782KB 10.2
#> 3 c 4.21ms 4.49ms 227. 782KB 5.22
plot(timings) + bench::scale_x_bench_time(base = NULL)
In the case of convolve(), quick() returns a function approximately
200 times quicker, giving similar performance to the C function.
quick() can accelerate any R function, with some restrictions:
- Function arguments must have their types and shapes declared using
declare(). - Only atomic vectors, matrices, and array are currently supported:
integer,double,logical, andcomplex. - The return value must be an atomic array (e.g., not a list)
- Named variables must have consistent shapes throughout their lifetimes.
NAvalues are not supported.- Only a subset of R’s vocabulary is currently supported.
#> [1] - : ! != (
#> [6] [ [<- [<<- { *
#> [11] / & && %*% %/%
#> [16] %% %o% ^ + <
#> [21] <- <<- <= = ==
#> [26] > >= | || Arg
#> [31] Conj Fortran Im Mod Re
#> [36] abs acos array as.double asin
#> [41] atan backsolve break c cat
#> [46] cbind ceiling character chol chol2inv
#> [51] cos crossprod declare diag dim
#> [56] double exp floor for forwardsolve
#> [61] if ifelse integer length log
#> [66] log10 logical matrix max min
#> [71] ncol next nrow numeric outer
#> [76] print prod raw rbind repeat
#> [81] runif seq seq_along seq_len sin
#> [86] solve sqrt sum t tan
#> [91] tcrossprod which.max which.min while
Many of these restrictions are expected to be relaxed as the project matures. However, quickr is intended primarily for high-performance numerical computing, so features like polymorphic dispatch or support for complex or dynamic types are out of scope.
The shape and mode of all function arguments must be declared. Local and return variables may optionally also be declared.
declare(type()) also has support for declaring size constraints, or
size relationships between variables. Here are some examples of declare
calls:
declare(type(x = double(NA))) # x is a 1-d double vector of any length
declare(type(x = double(10))) # x is a 1-d double vector of length 10
declare(type(x = double(1))) # x is a scalar double
declare(type(x = integer(2, 3))) # x is a 2-d integer matrix with dim (2, 3)
declare(type(x = integer(NA, 3))) # x is a 2-d integer matrix with dim (<any>, 3)
# x is a 4-d logical matrix with dim (<any>, 24, 24, 3)
declare(type(x = logical(NA, 24, 24, 3)))
# x and y are 1-d double vectors of any length
declare(type(x = double(NA)),
type(y = double(NA)))
# x and y are 1-d double vectors of the same length
declare(
type(x = double(n)),
type(y = double(n)),
)
# x and y are 1-d double vectors, where length(y) == length(x) + 2
declare(type(x = double(n)),
type(y = double(n+2)))The Viterbi algorithm is an example of a dynamic programming algorithm
within the family of Hidden Markov Models
(https://en.wikipedia.org/wiki/Viterbi_algorithm). Here, quick()
makes the viterbi() approximately 50 times faster.
slow_viterbi <- function(observations, states, initial_probs, transition_probs, emission_probs) {
declare(
type(observations = integer(num_steps)),
type(states = integer(num_states)),
type(initial_probs = double(num_states)),
type(transition_probs = double(num_states, num_states)),
type(emission_probs = double(num_states, num_obs)),
)
trellis <- matrix(0, nrow = length(states), ncol = length(observations))
backpointer <- matrix(0L, nrow = length(states), ncol = length(observations))
trellis[, 1] <- initial_probs * emission_probs[, observations[1]]
for (step in 2:length(observations)) {
for (current_state in 1:length(states)) {
probabilities <- trellis[, step - 1] * transition_probs[, current_state]
trellis[current_state, step] <- max(probabilities) * emission_probs[current_state, observations[step]]
backpointer[current_state, step] <- which.max(probabilities)
}
}
path <- integer(length(observations))
path[length(observations)] <- which.max(trellis[, length(observations)])
for (step in seq(length(observations) - 1, 1)) {
path[step] <- backpointer[path[step + 1], step + 1]
}
out <- states[path]
out
}
quick_viterbi <- quick(slow_viterbi)
set.seed(1234)
num_steps <- 16
num_states <- 8
num_obs <- 16
observations <- sample(1:num_obs, num_steps, replace = TRUE)
states <- 1:num_states
initial_probs <- runif (num_states)
initial_probs <- initial_probs / sum(initial_probs) # normalize to sum to 1
transition_probs <- matrix(runif (num_states * num_states), nrow = num_states)
transition_probs <- transition_probs / rowSums(transition_probs) # normalize rows
emission_probs <- matrix(runif (num_states * num_obs), nrow = num_states)
emission_probs <- emission_probs / rowSums(emission_probs) # normalize rows
timings <- bench::mark(
slow_viterbi = slow_viterbi(observations, states, initial_probs,
transition_probs, emission_probs),
quick_viterbi = quick_viterbi(observations, states, initial_probs,
transition_probs, emission_probs)
)
timings
#> # A tibble: 2 × 6
#> expression min median `itr/sec` mem_alloc `gc/sec`
#> <bch:expr> <bch:tm> <bch:tm> <dbl> <bch:byt> <dbl>
#> 1 slow_viterbi 202.52µs 213.73µs 4231. 1.59KB 12.6
#> 2 quick_viterbi 2.26µs 2.39µs 405918. 0B 0
plot(timings)
Simulate how heat spreads over time across a 2D grid, using the finite difference method applied to the Heat Equation.
Here, quick() returns a function over 100 times faster.
diffuse_heat <- function(nx, ny, dx, dy, dt, k, steps) {
declare(
type(nx = integer(1)),
type(ny = integer(1)),
type(dx = integer(1)),
type(dy = integer(1)),
type(dt = double(1)),
type(k = double(1)),
type(steps = integer(1))
)
# Initialize temperature grid
temp <- matrix(0, nx, ny)
temp[nx / 2, ny / 2] <- 100 # Initial heat source in the center
# Local helper that updates `temp` in-place.
apply_boundary_conditions <- function() {
temp[1, ] <<- 0
temp[nx, ] <<- 0
temp[, 1] <<- 0
temp[, ny] <<- 0
NULL
}
update_temperature <- function(temp, k, dx, dy, dt) {
temp_new <- temp
for (i in 2:(nx - 1)) {
for (j in 2:(ny - 1)) {
temp_new[i, j] <- temp[i, j] + k * dt *
((temp[i + 1, j] - 2 * temp[i, j] + temp[i - 1, j]) / dx ^ 2 +
(temp[i, j + 1] - 2 * temp[i, j] + temp[i, j - 1]) / dy ^ 2)
}
}
temp_new
}
# Time stepping
for (step in seq_len(steps)) {
apply_boundary_conditions()
temp <- update_temperature(temp, k, dx, dy, dt)
}
temp
}
quick_diffuse_heat <- quick(diffuse_heat)
# Parameters
nx <- 100L # Grid size in x
ny <- 100L # Grid size in y
dx <- 1L # Grid spacing
dy <- 1L # Grid spacing
dt <- 0.01 # Time step
k <- 0.1 # Thermal diffusivity
steps <- 500L # Number of time steps
timings <- bench::mark(
diffuse_heat = diffuse_heat(nx, ny, dx, dy, dt, k, steps),
quick_diffuse_heat = quick_diffuse_heat(nx, ny, dx, dy, dt, k, steps)
)
#> Warning: Some expressions had a GC in every iteration; so filtering is
#> disabled.
summary(timings, relative = TRUE)
#> Warning: Some expressions had a GC in every iteration; so filtering is
#> disabled.
#> # A tibble: 2 × 6
#> expression min median `itr/sec` mem_alloc `gc/sec`
#> <bch:expr> <dbl> <dbl> <dbl> <dbl> <dbl>
#> 1 diffuse_heat 224. 219. 1 514. Inf
#> 2 quick_diffuse_heat 1 1 220. 1 NaN
plot(timings)
Here is quickr used to calculate a rolling mean. Note that the CRAN package RcppRoll already provides a highly optimized rolling mean, which we include in the benchmarks for comparison.
slow_roll_mean <- function(x, weights, normalize = TRUE) {
declare(
type(x = double(NA)),
type(weights = double(NA)),
type(normalize = logical(1))
)
out <- double(length(x) - length(weights) + 1)
n <- length(weights)
if (normalize)
weights <- weights/sum(weights)*length(weights)
for(i in seq_along(out)) {
out[i] <- sum(x[i:(i+n-1)] * weights) / length(weights)
}
out
}
quick_roll_mean <- quick(slow_roll_mean)
x <- dnorm(seq(-3, 3, len = 100000))
weights <- dnorm(seq(-1, 1, len = 100))
timings <- bench::mark(
r = slow_roll_mean(x, weights),
rcpp = RcppRoll::roll_mean(x, weights = weights),
quickr = quick_roll_mean(x, weights = weights)
)
#> Warning: Some expressions had a GC in every iteration; so filtering is
#> disabled.
timings
#> # A tibble: 3 × 6
#> expression min median `itr/sec` mem_alloc `gc/sec`
#> <bch:expr> <bch:tm> <bch:tm> <dbl> <bch:byt> <dbl>
#> 1 r 160.42ms 173.12ms 5.54 124.24MB 14.8
#> 2 rcpp 25.56ms 25.89ms 38.8 4.46MB 0
#> 3 quickr 6.07ms 6.36ms 154. 781.35KB 3.99
timings$expression <- factor(names(timings$expression), rev(names(timings$expression)))
plot(timings) + bench::scale_x_bench_time(base = NULL)
Use declare(parallel()) to annotate the next for loop or sapply()
call for OpenMP parallelization. Parallel loops must be
order-independent: avoid shared-state updates or inter-iteration
dependencies. OpenMP adds overhead, so it can be slower for small
workloads, but substantially faster for larger ones.
Here is a concrete example using colSums(). At smaller sizes, the
quickr serial version is fastest (even faster than base colSums()). As
sizes grow, the two serial versions converge, and the parallel version
pulls ahead. However, the speedup is not linear with core count (e.g.,
with 12 cores, the speedup is closer to ~6x).
colSums_quick_parallel <- quick(function(x) {
declare(type(x = double(NA, NA)))
declare(parallel())
sapply(seq_len(nrow(x)), \(r) sum(x[, r]))
})
colSums_quick <- quick(function(x) {
declare(type(x = double(NA, NA)))
sapply(seq_len(nrow(x)), \(r) sum(x[, r]))
})
r <- bench::press(
n = 2^(4:14),
{
m <- array(runif(n * n), c(n, n))
bench::mark(
parallel = colSums_quick_parallel(m),
quick = colSums_quick(m),
base = colSums(m),
)
},
.quiet = TRUE
)
library(ggplot2)
library(dplyr, warn.conflicts = FALSE)
r |>
mutate(.before = 1,
desc = attr(expression, "description")) |>
select(desc, n, median) |>
ggplot(aes(x = n, y = median, color = desc)) +
geom_point() + geom_line() +
scale_x_log10() + bench::scale_y_bench_time()
quickr does not set OpenMP thread counts. To control threads, set
OMP_NUM_THREADS (and optionally OMP_THREAD_LIMIT or OMP_DYNAMIC)
before calling a compiled function,
e.g. Sys.setenv(OMP_NUM_THREADS = "4").
When called in a package, quick() will pre-compile the quick functions
and place them in the ./src directory. Run devtools::load_all() or
quickr::compile_package() to ensure that the generated files in
./src and ./R are in sync with each other.
In a package, you must provide a function name to quick(). For
example:
my_fun <- quick(name = "my_fun", function(x) ....)We have implemented a subset of matrix operations from base R. When you
do a matrix multiplication like A %*% B, R calls BLAS/LAPACK functions
linked to your R build. quickr is linked to the same BLAS/LAPACK
implementation, so you can expect the same results and slightly faster
computation due to lower overhead. There are multiple BLAS/LAPACK
implementations, and performance therefore depends on what you have
installed.
To illustrate the performance of matrix operations, we do a linear
regression using the normal equation, which is fast but not the most
numerically stable method. We compare with RcppArmadillo, which is a
popular library for matrix operations. The RcppArmadillo
implementation aims to match our R implementation.
lm <- function(X, y) {
declare(
type(X = double(n, k)),
type(y = double(n))
)
df <- nrow(X) - ncol(X)
XtX <- crossprod(X)
Xty <- crossprod(X, y)
coef <- solve(XtX, Xty)
fit_val <- X %*% coef
resid <- y - fit_val
s2 <- crossprod(resid)[1]
s2 <- s2 / df
U <- chol(XtX)
XtX_inv <- chol2inv(U)
std_err <- sqrt(diag(XtX_inv) * s2)
list(
coefficients = coef,
stderr = std_err,
df.residual = df,
fitted_values = fit_val,
residuals = resid
)
}
qlm <- quick(lm)
Rcpp::sourceCpp(
code = '#include <RcppArmadillo.h>
// [[Rcpp::depends(RcppArmadillo)]]
// [[Rcpp::export]]
Rcpp::List RcppLm(const arma::mat& X, const arma::colvec& y) {
int n = X.n_rows;
int k = X.n_cols;
int df = n - k;
arma::mat XtX = arma::trans(X) * X;
arma::colvec Xty = arma::trans(X) * y;
arma::colvec coef = arma::solve(XtX, Xty);
arma::colvec fit_val = X * coef;
arma::colvec resid = y - fit_val;
double s2 = arma::dot(resid, resid);
s2 = s2 / df;
arma::mat U = arma::chol(XtX);
arma::mat XtX_inv = arma::inv(arma::trimatu(U)) * arma::trans(arma::inv(arma::trimatu(U)));
arma::colvec std_err = arma::sqrt(arma::diagvec(XtX_inv) * s2);
return Rcpp::List::create(
Rcpp::Named("coefficients") = coef,
Rcpp::Named("stderr") = std_err,
Rcpp::Named("df.residual") = df,
Rcpp::Named("fitted_values") = fit_val,
Rcpp::Named("residuals") = resid
);
}'
)
beta <- c(0.5, 1.0, -2.0, 10, 5)
X <- cbind(1, matrix(rnorm(3 * 10^6), ncol = 4))
y <- as.vector(X %*% beta + rnorm(nrow(X), sd = 2))
timings <- bench::mark(
r = lm(X, y),
quickr = qlm(X, y),
RcppArmadillo = RcppLm(X, y),
check = FALSE, # stderr is a vector in R and matrix in RcppArmadillo
min_iterations = 30
)
timings
#> # A tibble: 3 × 6
#> expression min median `itr/sec` mem_alloc `gc/sec`
#> <bch:expr> <bch:tm> <bch:tm> <dbl> <bch:byt> <dbl>
#> 1 r 21.5ms 22ms 44.9 11.4MB 0
#> 2 quickr 14.4ms 14.6ms 67.3 11.4MB 2.04
#> 3 RcppArmadillo 15.3ms 25ms 42.3 11.4MB 0You can install quickr from CRAN with:
install.packages("quickr")You can install the development version of quickr from GitHub with:
# install.packages("pak")
pak::pak("t-kalinowski/quickr")You will also need a C and Fortran compiler, preferably the same ones used to build R itself.
On macOS:
-
Make sure xcode tools and gfortran are installed, as described in https://mac.r-project.org/tools/. In Terminal, run:
sudo xcode-select --install # curl -LO https://mac.r-project.org/tools/gfortran-12.2-universal.pkg # R 4.4 curl -LO https://mac.r-project.org/tools/gfortran-14.2-universal.pkg # R 4.5 sudo installer -pkg gfortran-12.2-universal.pkg -target /
-
Optional: install
flang-newvia Homebrew (used by quickr on macOS when available):brew install flang
On Windows:
- Install the latest version of Rtools
On Linux:
- The “Install Required Dependencies” section here provides detailed instructions for installing R build tools on various Linux flavors.