Esempio n. 1
0
    void setup_buffer( double alpha, double beta, const std::string& path )
    {
        int fileError = sparseHeaderfromFile(&n_vals, &n_rows, &n_cols, path.c_str());
        if (fileError != 0)
        {
            throw clsparse::io_exception( "Could not read matrix market header from disk: " + path);
        }

        if (csrMatrixfromFile(row_offsets, col_indices, values, path.c_str(), explicit_zeroes))
        {
            throw clsparse::io_exception( "Could not read matrix market from disk: " + path);
        }

        //n_rows = row_offsets.size( );
        //n_cols = col_indices.size( );
        //n_vals = values.size( );

        cudaError_t err = cudaMalloc( (void**) &device_row_offsets, (n_rows + 1) * sizeof( clsparseIdx_t ) );
        CUDA_V_THROW( err, "cudaMalloc device_row_offsets" );

        err = cudaMalloc( (void**) &device_col_indices, n_vals * sizeof( clsparseIdx_t ) );
        CUDA_V_THROW( err, "cudaMalloc device_col_indices" );

        err = cudaMalloc( (void**) &device_values, n_vals * sizeof( T ) );
        CUDA_V_THROW( err, "cudaMalloc device_values" );

        err = cudaMalloc( (void**) &device_A, n_rows * n_cols * sizeof( T ) );
        CUDA_V_THROW( err, "cudaMalloc device_A" );
    }
Esempio n. 2
0
    void setup_buffer(double alpha, double beta, const std::string& path)
    {
        int fileError = sparseHeaderfromFile(&n_vals, &n_rows, &n_cols, path.c_str());
        if (fileError != 0)
        {
            throw clsparse::io_exception("Could not read matrix market header from disk");
        }

        if (csrMatrixfromFile(row_offsets, col_indices, values, path.c_str()))
        {
            throw clsparse::io_exception("Could not read matrix market header from disk");
        }

        cudaError_t err = cudaMalloc((void**)&device_row_offsets, (n_rows + 1) * sizeof(int));
        CUDA_V_THROW(err, "cudaMalloc device_row_offsets");

        err = cudaMalloc((void**)&device_col_indices, n_vals * sizeof(int));
        CUDA_V_THROW(err, "cudaMalloc device_col_indices");

        err = cudaMalloc((void**)&device_values, n_vals * sizeof(T));
        CUDA_V_THROW(err, "cudaMalloc device_values");

        err = cudaMalloc((void**)&device_A, n_rows * n_cols * sizeof(T));
        CUDA_V_THROW(err, "cudaMalloc device_A");

        // Output CSR
        err = cudaMalloc((void**)&devRowOffsets, (n_rows + 1) * sizeof(int));
        CUDA_V_THROW(err, "cudaMalloc devRowOffsets");

        err = cudaMalloc((void**)&devColIndices, n_vals * sizeof(int));
        CUDA_V_THROW(err, "cudaMalloc devColIndices");

        err = cudaMalloc((void**)&devValues, n_vals * sizeof(T));
        CUDA_V_THROW(err, "cudaMalloc devValues");

        // Allocate memory for nnzPerRow
        err = cudaMalloc((void**)&nnzPerRow, n_rows * sizeof(int));
        CUDA_V_THROW(err, "cudaMalloc nnzPerRow");

    }// end