Rcpp_vector_add_hdf5

C++ Function Reference

1 Signature

BigDataStatMeth::hdf5Dataset * BigDataStatMeth::Rcpp_vector_add_hdf5(BigDataStatMeth::hdf5Dataset *dsA, BigDataStatMeth::hdf5Dataset *dsB, BigDataStatMeth::hdf5Dataset *dsC, bool bparal, Rcpp::Nullable< int > threads=R_NilValue)

2 Description

Pure vector addition for HDF5 vectors.

3 Parameters

  • dsA (BigDataStatMeth::hdf5Dataset *): First input vector dataset
  • dsB (BigDataStatMeth::hdf5Dataset *): Second input vector dataset
  • dsC (BigDataStatMeth::hdf5Dataset *): Output vector dataset
  • bparal (bool): Whether to use parallel processing
  • threads (Rcpp::Nullable< int >): Number of threads for parallel processing (optional)

4 Returns

Pointer to result dataset

5 Details

Performs optimized element-wise addition C = A + B where A, B, and C are HDF5 vector datasets. ~100x more efficient than matrix-vector approaches by avoiding unnecessary block processing and vector duplication.

6 Call Graph

Function dependencies

7 Source Code

File: inst/include/hdf5Algebra/vectorOperations.hpp • Lines 103-175

inline BigDataStatMeth::hdf5Dataset* Rcpp_vector_add_hdf5(
            BigDataStatMeth::hdf5Dataset* dsA, BigDataStatMeth::hdf5Dataset* dsB, BigDataStatMeth::hdf5Dataset* dsC,
            bool bparal, Rcpp::Nullable<int> threads = R_NilValue)
    {
        try {
            // Validate inputs are vectors
            hsize_t sizeA = validateVector(dsA);
            hsize_t sizeB = validateVector(dsB);
            
            if (sizeA == 0 || sizeB == 0) {
                Rcpp::Rcout << "vector add error: inputs are not vectors\n";
                return dsC;
            }
            
            if (sizeA != sizeB) {
                Rcpp::Rcout << "vector add error: non-conformable vector dimensions\n";
                return dsC;
            }
            
            // Create output vector with same dimensions as input
            hsize_t rowsA = dsA->nrows();
            hsize_t colsA = dsA->ncols();
            dsC->createDataset(colsA, rowsA, "real");
            
            // Direct vector I/O - no block processing needed
            std::vector<hsize_t> stride = {1, 1};
            std::vector<hsize_t> block = {1, 1};
            
            std::vector<double> vdA(sizeA);
            std::vector<double> vdB(sizeB);
            
            // Read entire vectors in single operations
            dsA->readDatasetBlock({0, 0}, {rowsA, colsA}, stride, block, vdA.data());
            dsB->readDatasetBlock({0, 0}, {dsB->nrows(), dsB->ncols()}, stride, block, vdB.data());
            
            // Perform element-wise addition using STL transform (highly optimized)
            if (bparal && sizeA > 10000) {
                // Parallel processing for large vectors
    #pragma omp parallel num_threads(get_threads(bparal, threads))
    {
    #pragma omp for schedule(static)
        for (hsize_t i = 0; i < sizeA; ++i) {
            vdA[i] += vdB[i];
        }
    }
            } else {
                // Sequential processing (optimal for smaller vectors)
                std::transform(vdA.begin(), vdA.end(), vdB.begin(), vdA.begin(), std::plus<double>());
            }
            
            // Write result in single operation
            dsC->writeDatasetBlock(vdA, {0, 0}, {rowsA, colsA}, stride, block);
            
        } catch(H5::FileIException& error) {
            checkClose_file(dsA, dsB, dsC);
            Rcpp::Rcerr << "\nc++ exception Rcpp_vector_add_hdf5 (File IException)";
            // return dsC;
        } catch(H5::DataSetIException& error) {
            checkClose_file(dsA, dsB, dsC);
            Rcpp::Rcerr << "\nc++ exception Rcpp_vector_add_hdf5 (DataSet IException)";
            // return dsC;
        } catch(std::exception& ex) {
            checkClose_file(dsA, dsB, dsC);
            Rcpp::Rcerr << "\nc++ exception Rcpp_vector_add_hdf5: " << ex.what();
            // return dsC;
        } catch (...) {
            checkClose_file(dsA, dsB, dsC);
            Rcpp::Rcerr << "\nC++ exception Rcpp_vector_add_hdf5 (unknown reason)";
            // return dsC;
        }
        
        return dsC;
    }

8 Usage Example

#include "BigDataStatMeth.hpp"

// Example usage
auto result = Rcpp_vector_add_hdf5(...);