R Observatory: Metrics for cudaverse version 0.4.2

Balamuta, James Joseph

doi:10.5281/zenodo.21843040

cudaverse

0.4.2Package record

Lightweight 'CUDA' Numerical Computing

0packages depend

517downloads / year

69.1%test coverage

13/13flavors OK

Overview

About

Maintained by Yaoxiang LiCRAN page ↗GitHub ↗Website ↗Issues ↗

Provides a lightweight interface to graphics processing unit (GPU)-accelerated numerical computing using 'CUDA'. Dense tensors, sparse matrices, decompositions, distances, exact nearest neighbours, clustering, graph workflows, and embeddings share one consistent interface. The native backend discovers the 'NVIDIA CUDA Driver API', 'cuBLAS', and 'cuSOLVER' libraries at runtime without bundling 'LibTorch' or the 'CUDA Runtime'. Stage-level provenance records the backend, device, and data transfers used by each result. A portable implementation supports package validation on systems without 'CUDA'. Background for the included Leiden community detection and uniform manifold approximation and projection methods is given by Traag, Waltman and van Eck (2019) doi:10.1038/s41598-019-41695-z and McInnes et al. (2018) doi:10.21105/joss.00861, respectively.

Changes in 0.4.2

Native Euclidean distances use a dimension-aware cancellation check and direct coordinate differences for close pairs, preserving small distances in distance matrices and k-nearest-neighbour results.
Added an explicit precision = "tf32" option to native float32 tensor_matmul(). It permits cuBLAS TensorFloat-32 computation on supported GPUs, records the policy in provenance and leaves standard dtype arithmetic and %*% unchanged. Unsupported requests fail without changing later calls.
Standard float32 matmul benchmarks use a documented componentwise rounding model with conditioning, underflow and overflow checks. Reports retain the former strict tolerance and its failed-entry count as diagnostics.
Native PCA computes scores by projecting the centered/scaled input onto its fitted loadings. Identical observations keep identical scores and exact zero-distance neighbour ties, while the scores remain on the device.

Show all changes in 0.4.2Hide the rest of the changes in 0.4.2

Benchmark validation now applies numerical limits element by element, rejects malformed GPU activity reports, and checks the explicitly declared backends so that a native CUDA comparison does not require torch.
Retained benchmark timings check for competing GPU processes before and after individual samples, outside the measured interval.

From NEWS.md in the release

Install

R (CRAN)pak (dev)Ubuntu (r2u)Fedora (COPR)openSUSE (OBS)

install.packages("cudaverse", repos = "https://cloud.r-project.org/")

Installs the current CRAN release. Prebuilt binaries on Windows and macOS.

if (!requireNamespace("pak", quietly = TRUE))
  install.packages("pak", repos = "https://cloud.r-project.org/")
pak::pak("cudaverse/cudaverse")

Development version from cudaverse/cudaverse. Installs pak first if it is not already available.

Release

# enable cranapt (r2u) if it is not already set up, then install
grep -rqs r2u.stat.illinois.edu /etc/apt/sources.list /etc/apt/sources.list.d/ \
  || curl -fsSL https://raw.githubusercontent.com/eddelbuettel/r2u/master/inst/scripts/add_cranapt_noble.sh | sudo bash
sudo apt install -y r-cran-cudaverse

Fast Ubuntu binaries via r2u (no compilation). The cranapt setup runs once and is skipped when it is already enabled. We list this channel from recorded r2u download activity rather than from a live index of the repository, so treat it as last known rather than guaranteed.

bspm (used by r2u) also needs make: sudo apt install make

# enable the iucar/cran COPR if it is not already present, then install
ls /etc/yum.repos.d/*iucar*cran*.repo >/dev/null 2>&1 \
  || sudo dnf copr enable -y iucar/cran
sudo dnf install -y R-CRAN-cudaverse

RPMs from the iucar/cran COPR, rebuilt from CRAN.

For install.packages() integration in R, also: sudo dnf install R-CoprManager

# add the autoCRAN OBS repo (Tumbleweed) if it is missing, then install
zypper -q lr | grep -qi autoCRAN \
  || sudo zypper addrepo https://download.opensuse.org/repositories/devel:/languages:/R:/autoCRAN/openSUSE_Tumbleweed/devel:languages:R:autoCRAN.repo
sudo zypper install -y R-cudaverse

RPMs auto-built from CRAN on openSUSE OBS (Tumbleweed). We list this channel from recorded autoCRAN download activity rather than from a live index of the repository, so treat it as last known rather than guaranteed.

Health

CRAN checks

13OK

CRAN check page ↗

Slowest check: 3.9 min · r-devel-windows-x86_64

Status changes

OK13 OKNow
OK6 OK2026-09-11

Advisories

none known

How this is measured·How advisories are found

Documentation

In release 0.4.2

READMEYes · 812 wordsVignettesYes · 5 · dynamicNEWSYes · 100% structured · mentions 0.4.2Citation fileNoWebsiteYes · pkgdown

Description cites a JOSS paper

Help pages: 44 (1 internal)includes a package overview page
Examples that run: 100%40 run in a plain check (2 of them under a condition)
Exported functions with examples: 100%
Documented parameters: 95%
Return-value docs: 100%
Help pages citing references: 0%

Vignettes (5)

Keep an R workflow on the GPU HTML
Performance and package size HTML
Set up CUDA for cudaverse HTML
What runs on CUDA? HTML
Your first CUDA workflow with cudaverse HTML

How these are measured

Downloads

517

CRAN downloads in the past year

Rank #23,930 · ~1/day · ~43/mo

Releases R versions

Daily download trend is not available in this view yet.

38430 days

51790 days

5171 year

Compare downloads with other packages →

Also on17 r2u14 autocran

How downloads are counted

Repository

cudaverse/cudaverse ↗Organization: cudaverse

2Stars

0Forks

0Open issues

0Open PRs

3Releases

244Commits

1Commit identities

cudagpurr-packagescientific-computingtensors

244 commits · Last activity 2026-10-17 · +100% stars, 30d

Responsiveness

How these are sampled

PRs merged (all time): 100%
Median days to close, 50 most recently updated closed issues: not measured
Median days to close, 50 most recently updated closed PRs: 0

Stars over time

2026-09-10 · 12026-09-26 · 2

How these are measured

PracticesBeta

10 development-tooling and community-health practices detected across 5 families in the upstream repository

Checks run against github.com/cudaverse/cudaverse on 2026-10-11.

Continuous integration (2)

GitHub ActionsR CMD check workflow

The R CMD check workflow runs on Linux, macOS and Windows, and on R-devel

Governance and community (3)

Code of conduct (inherited from cudaverse)Contributing guide (inherited from cudaverse)PR template (inherited from cudaverse)

Docs source (3)

pkgdown

GitHub Pages site last deployed 2026-09-26 · The repository has a site deploy workflow

pkgdown site last built 2026-09-26, before release 0.4.2 (2026-10-17)

CRAN release process (1)

cran-comments.md

Show all practicesHide 1 practice

Git structural (1)

.gitattributes

How this is detected·Detection ruleset v3 (2026-09-26)

Dependencies

Declared dependencies

10 external dependencies (excludes base and recommended)

Installing cudaverse adds no packages beyond those that come with R.

Depends (0)

none

Imports (3)

Matrixmethodsstats

LinkingTo (0)

none

Suggests (10)

igraph knitr rmarkdown RSpectra Rtsne S4Vectors SingleCellExperiment testthat torch uwot

Enhances (0)

none

Dependency license breakdown →

Reverse dependencies

0direct

0indirect

Nothing depends on this yet.

How these are counted

Code & Tests

People & History

People (1)

Yaoxiang Li

Author, Maintainer since 0.4.1 (2026)

Releases (2)

Pick two releases to compare their code metrics.

0.4.2Latest
record
2026-10-17 · current release · diff ↗
0.4.1
record
2026-09-10
R
R 4.6.0 released · 2026-04-24

Package metadata

First published: 2026-09-10
Total releases: 2 / 1 yrs
License: MIT + file LICENSE OSI
Download size: 348 KB

Cite

Cite this package

Run in R for the authors' preferred citation:

citation("cudaverse")

Format

Li, Y. (2026). cudaverse: Lightweight 'CUDA' Numerical Computing (Version 0.4.2) [Computer software]. https://doi.org/10.32614/CRAN.package.cudaverse

This package has no citation file, so citation() returns this entry built from DESCRIPTION.

Cite the R Observatory

For a number measured here: a download total, a coverage figure, an archival date.

Format

APA

Balamuta, J. J. (2026). R Observatory: Metrics for cudaverse version 0.4.2 [Data set]. HJJB, LLC. Data release v2026-10-19. https://doi.org/10.5281/zenodo.21843040

From data release v2026-10-19, which the citation names so these numbers can be found later. More on citing and the projects behind them.

Report a problem with this page →