Skip to contents

bigalgebra

Arithmetic routines for native R matrices and big.matrix objects

Frédéric Bertrand, Michael J. Kane, Bryan Lewis, John W. Emerson

https://doi.org/10.32614/CRAN.package.bigalgebra

bigalgebra provides fast linear algebra primitives that operate seamlessly on base matrix objects and [bigmemory::big.matrix] containers. The package wraps BLAS and LAPACK routines with R-friendly helpers so that vector updates, matrix products, and classic decompositions work the same way in memory or on disk.

Package highlights

Package options

The package defines a number of global options that begin with bigalgebra:

Option Default value * bigalgebra.temp_pattern with default matrix_ * bigalgebra.tempdir with default tempdir * bigalgebra.mixed_arithmetic_returns_R_matrix with default TRUE * bigalgebra.DEBUG with default FALSE

The bigalgebra.tempdir option must be a function that returns a temporary directory path used to store big matrix results of BLAS and LAPACK operations. The default value is simply the base R tempdir() function.

The bigalgebra.temp_pattern option is a name prefix for file names of generated big matrix objects output as a result of BLAS and LAPACK operations.

The bigalgebra.mixed_arithmetic_returns_R_matrix option determines whether arithmetic operations involving an R matrix or vector and a big.matrix matrix or vector return a big matrix (when the option is FALSE), or return a normal R matrix (TRUE).

BLAS and LAPACK backends

The package is built, by default, with R’s native BLAS libraries, which use 32-bit signed integer indexing. The default build is limited to vectors of at most 2^31 − 1 entries and matrices with at most 2^31 − 1 rows and 2^31 − 1 columns (note that standard R matrices are limited to 2^31 − 1 total entries).

The package includes a reference BLAS implementation that supports 64-bit integer indexing, relaxing the limitation on vector lengths and matrix row and column limits. Installation of this package with the 64-bit reference BLAS implementation may be performed from the command-line install:

REFBLAS=1 R CMD INSTALL bigalgebra

where bigalgebra is the source package (for example, bigalgebra_0.9.0.tar.gz).

The package may also be built with user-supplied external BLAS and LAPACK libraries, in either 32- or 64-bit varieties. This is an advanced topic that requires additional Makevars modification, and may include adjustment of the low-level calling syntax depending on the library used.

Feel free to contact us for help installing and running the package.

This website, the unit tests, some C code fixes and improvements as well as these examples were created by F. Bertrand.

Maintainer: Frédéric Bertrand .

Installation

You can install the released version of bigalgebra from CRAN with:

install.packages("bigalgebra")

You can install the development version of bigalgebra from GitHub with:

devtools::install_github("fbertran/bigalgebra")

Quick tour of the functionality

The snippets below mirror the worked examples in the vignettes and show how the helpers behave with in-memory and file-backed matrices.

Level 1 BLAS helpers

These helpers cover vector updates, reductions, and element-wise transforms such as the in-place square root provided by dsqrt().

library(bigmemory)
library(bigalgebra)

x <- bigmemory::big.matrix(5, 1, init = 0)
dset(ALPHA = 9, X = x)
dsqrt(X = x)
x[]
#> [1] 3 3 3 3 3

y <- bigmemory::big.matrix(5, 1, init = 1)
dvcal(ALPHA = 0.5, X = x, BETA = 2, Y = y)
y[]
#> [1] 3.5 3.5 3.5 3.5 3.5

Matrix products with dgemm() 

A <- bigmemory::big.matrix(5, 4, init = 1)
B <- bigmemory::big.matrix(4, 4, init = 2)
C <- bigmemory::big.matrix(5, 4, init = 0)

dgemm(A = A, B = B, C = C, ALPHA = 1, BETA = 0)
C[]
#>      [,1] [,2] [,3] [,4]
#> [1,]    8    8    8    8
#> [2,]    8    8    8    8
#> [3,]    8    8    8    8
#> [4,]    8    8    8    8
#> [5,]    8    8    8    8

LAPACK decompositions 

set.seed(1)
M <- matrix(rnorm(9), 3)
SPD <- crossprod(M)
SPD_big <- as.big.matrix(SPD)
dpotrf(A = SPD_big)
#> [1] 0
chol_factor <- SPD_big[,]
chol_factor[lower.tri(chol_factor)] <- 0
chol_factor
#>          [,1]       [,2]       [,3]
#> [1,] 1.060398 -0.2388263 -0.6138286
#> [2,] 0.000000  1.8082109  0.2222424
#> [3,] 0.000000  0.0000000  0.8294922

File-backed big.matrix workflows 

tmpdir <- tempdir()
file_big <- filebacked.big.matrix(3, 3, init = diag(3),
                                  backingpath = tmpdir,
                                  backingfile = "example.bin")
#> Warning in filebacked.big.matrix(3, 3, init = diag(3), backingpath = tmpdir, : No
#> descriptor file given, it will be named example.bin.desc
file_big[1, 3] <- 5
file_big[]
#>      [,1] [,2] [,3]
#> [1,]    1    1    5
#> [2,]    1    1    1
#> [3,]    1    1    1
rm(file_big)
gc()
#>           used (Mb) gc trigger (Mb) limit (Mb) max used (Mb)
#> Ncells  900616 48.1    1699095 90.8         NA  1377768 73.6
#> Vcells 2253326 17.2    8388608 64.0      65536  3247956 24.8

Available vignettes

The full vignette set expands on the topics above and demonstrates how the routines interact:

  1. Working with big.matrix Objects – Managing shared memory, file backing, and clean-up for large datasets.
  2. Level 1 BLAS-Style Helpers – Filling vectors, Hadamard products, and reductions on disk-backed data.
  3. Matrix Wrapper Helpers – Symmetric and general matrix products, including strategies for chaining operations.
  4. LAPACK Decompositions with bigalgebra – QR, Cholesky, eigenvalue, and SVD workflows for both in-memory and file-backed matrices.