Dev cuda

[suzannejin]

No description provided.

[Copilot]

Pull request overview

This pull request introduces GPU acceleration support for the propr package through CUDA implementation, along with significant refactoring of the codebase architecture and enhanced testing infrastructure.

Changes:

• Added CUDA GPU acceleration for core computational functions (lrv, lrm, omega, ctzRcpp, comparison operations)
• Refactored C++ code into modular CPU/GPU dispatch architecture with separate interface, kernel, and trait layers
• Enhanced testing with permutation options (feature-wise vs sample-wise), additional test coverage, and GPU-specific tests
• Added profiling infrastructure (NVTX markers) and CI/CD improvements with containerized testing

Reviewed changes

Copilot reviewed 154 out of 160 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
thirdparty/Catch2	Added Catch2 testing framework as git submodule
tests/testthat/test-SHARED-updatepermutes.R	Added test for permutation conservation of gene/sample-wise totals
tests/testthat/test-SHARED-updateCutoffs-propr.R	Updated tests to use sample-wise permutation option and skip parallel tests on CI
tests/testthat/test-SHARED-updateCutoffs-propd.R	Added skip for parallel test on local CI
tests/testthat/test-SHARED-lrv.R	Added GPU vs CPU comparison test for lrv function
tests/testthat/test-SHARED-getCutoff-propr.R	Split test into feature-wise and sample-wise variants
tests/testthat/test-SHARED-getCutoff-propd.R	Updated tests to use non-moderated F-statistics
tests/testthat/test-SHARED-getAdjacency-propr.R	Fixed adjacency matrix logic to use directional comparison instead of absolute value
tests/testthat/test-SHARED-getAdjacency-propd.R	Enhanced test diagnostics and updated to use non-moderated F-statistics
tests/testthat/test-RCPP-ctzRcpp.R	Added GPU variant test for zero frequency counting
tests/testthat/test-RCPP-count.R	Added GPU variant tests for comparison operations
tests/testthat/test-PROPR-clr-alr-vlr.R	Added CPU vs GPU comparison tests for clr and alr transformations
tests/testthat/test-PROPD-weight.R	Refactored weight tests with new fetch_weights helper and comprehensive weight validation
tests/testthat/test-PROPD-updateF.R	Added comprehensive tests for F-statistic calculation with and without moderation
tests/testthat/test-PROPD-theta.R	Updated weight calculation to use sample reliability weights from limma
tests/testthat/test-PROPD-lrm-lrv.R	Added GPU vs CPU numerical equivalence test for propd
tests/testthat/test-GRAFLEX.R	Added GPU variant test for FDR calculation and updated permutation test
tests/testthat/test-GET-getMatrix.R	Enhanced test diagnostics with detailed error messages
tests/tests-gpu/test-PROPR-clr-alr-vlr.R	Added GPU-specific test suite for transformations
src/omega.cpp	Refactored to use dispatch pattern with CPU/GPU backend selection
src/lrv.cpp	Refactored to use dispatch pattern with CPU/GPU backend selection
src/lrm.cpp	Refactored to use dispatch pattern with CPU/GPU backend selection
src/lr2propr.cpp	Refactored to use dispatch pattern with CPU/GPU backend selection
src/host_profiler.cpp	Implemented host-side profiling infrastructure
src/host_exclusive_profiler.cpp	Implemented exclusive profiling to prevent nested measurements
src/dispatch/device_selector.cpp	Implemented runtime GPU backend selection based on R options
src/dispatch/cuda/omega.cu	CUDA implementation of omega calculations
src/dispatch/cuda/lrm.cu	CUDA implementation of log-ratio mean calculations
src/dispatch/cuda/lr2propr.cu	CUDA stubs for proportionality transformations
src/dispatch/cuda/ctzRcpp.cu	CUDA implementation of zero frequency counting
src/dispatch/cuda/comparison.cu	CUDA implementation using Thrust for comparison operations
src/dispatch/cuda/CMakeLists.txt	Build configuration for CUDA dispatch sources
src/dispatch/cpu/*.cpp	CPU implementations refactored into dispatch pattern
src/ctzRcpp.cpp	Refactored interface with GPU support
src/cpp_tests/*.cu	Added C++ unit test infrastructure
src/comparison.cpp	Refactored interface with GPU support
src/backend.h	Removed old header (functionality moved to dispatch pattern)
src/Makefile	Added minimal makefile
src/CMakeLists.txt	Main build configuration with CUDA support
scripts/git/pre-push	Added pre-push hook for automated testing
old_vignettes/e_differential.Rmd	Minor formatting correction
man/*.Rd	Updated documentation for new parameters and corrected examples
inst/include/propr/**/*.{h,hpp,cuh}	Added extensive header infrastructure for GPU/CPU implementations
configure	Added configuration script for CMake-based build
cmake/*.cmake	Added CMake modules for finding R, Rcpp, and testthat
ci/run-tests.R	Added CI test runner script
bench/nvbench	Added nvbench submodule for benchmarking
README.md	Fixed NULL to NA in examples and added local development section
R/*.R	Added NVTX profiling markers throughout R code
R/3-shared-updatePermutes.R	Added permutation_option parameter support
R/1-propr.R	Added permutation_option parameter and slot
R/1-propr-OOP.R	Added permutation_option slot to propr class
R/10-nvtx-shim.R	Added NVTX profiling shim functions
OLDCMakeLists.txt	Legacy CMake configuration preserved
NEWS.md	Updated changelog with new features
DESCRIPTION	Added nvtxR to suggests, fixed formatting
CMakeLists.txt	Main CMake configuration with CUDA detection
.gitmodules	Added submodule configuration
.gitignore	Added build and CI artifact directories

Comments suppressed due to low confidence (4)

tests/testthat/test-RCPP-count.R:1

• Missing space after comma before '4' in the expected value.
  src/dispatch/cpu/omega.cpp:1
• The function pad_matrix2 has an unclear suffix '2'. Consider renaming to something more descriptive like pad_matrix or explain why the '2' suffix is necessary.
  configure:1
• Misspelling in variable name 'ABSOLUE_PATH'. Should be 'ABSOLUTE_PATH'.
  inst/include/propr/interface/backend.hpp:1
• Parameter name 'nfeatsbool' is confusing - it appears to be an integer representing number of features but has 'bool' in the name. Should be renamed to 'nfeats'.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[Copilot]

Missing space around the default parameter assignment. Should be alpha = NA for consistency with R style guidelines.

Suggested change

	logratio <- function(counts, ivar, alpha=NA) {
	logratio <- function(counts, ivar, alpha = NA) {

[Copilot]

Misspelling of 'development' as 'developement'.

Suggested change

	## For Local Developement
	For local developement install the pre-push script `./scripts/git/pre-push` to run the tests.
	## For Local Development
	For local development install the pre-push script `./scripts/git/pre-push` to run the tests.