TheRock/rocm-libraries/projects/rocalution/clients/samples/benchmark.cpp

/* ************************************************************************
 * Copyright (C) 2018-2020 Advanced Micro Devices, Inc. All rights Reserved.
 *
 * Permission is hereby granted, free of charge, to any person obtaining a copy
 * of this software and associated documentation files (the "Software"), to deal
 * in the Software without restriction, including without limitation the rights
 * to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
 * copies of the Software, and to permit persons to whom the Software is
 * furnished to do so, subject to the following conditions:
 *
 * The above copyright notice and this permission notice shall be included in
 * all copies or substantial portions of the Software.
 *
 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
 * AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
 * THE SOFTWARE.
 *
 * ************************************************************************ */

#include <cstdlib>
#include <iostream>
#include <rocalution/rocalution.hpp>

using namespace rocalution;

int main(int argc, char* argv[])
{
    // Check command line parameters
    if(argc == 1)
    {
        std::cerr << argv[0] << " <matrix> [Num threads]" << std::endl;
        exit(1);
    }

    // Initialize rocALUTION
    init_rocalution();

    // Check command line parameters for number of OMP threads
    if(argc > 2)
    {
        set_omp_threads_rocalution(atoi(argv[2]));
    }

    // Print rocALUTION info
    info_rocalution();

    // rocALUTION objects
    LocalVector<double> vec1;
    LocalVector<double> vec2;
    LocalVector<double> vec3;

    LocalMatrix<double> mat;

    // Time measurement
    double tick, tack;

    // Number of tests
    const int max_tests = 200;

    // Read matrix from MTX file
    mat.ReadFileMTX(std::string(argv[1]));

    // Allocate vectors
    vec1.Allocate("x", mat.GetN());
    vec2.Allocate("y", mat.GetM());
    vec3.Allocate("z", mat.GetM());

    int size = mat.GetM();
    int nnz  = mat.GetNnz();

    // Initialize objects
    vec1.Ones();
    vec2.Zeros();
    vec3.Ones();
    mat.Apply(vec1, &vec2);

    // Move objects to accelerator
    mat.MoveToAccelerator();
    vec1.MoveToAccelerator();
    vec2.MoveToAccelerator();
    vec3.MoveToAccelerator();

    // Print object info
    mat.Info();
    vec1.Info();
    vec2.Info();
    vec3.Info();

    std::cout << "----------------------------------------------------" << std::endl;
    std::cout << "Number of tests = " << max_tests << std::endl;

    std::cout << "----------------------------------------------------" << std::endl;
    std::cout << "Stand alone micro benchmarks" << std::endl;

    // Dot product
    // Size = 2*size
    // Flop = 2 per element
    vec3.Dot(vec2);

    _rocalution_sync();
    tick = rocalution_time();

    for(int i = 0; i < max_tests; ++i)
    {
        vec3.Dot(vec2);
        _rocalution_sync();
    }

    _rocalution_sync();
    tack = rocalution_time();
    std::cout << "Dot execution: " << (tack - tick) / max_tests / 1e3 << " msec"
              << "; " << max_tests * double(sizeof(double) * (size + size)) / (tack - tick) / 1e3
              << " Gbyte/sec; " << max_tests * double((2 * size)) / (tack - tick) / 1e3
              << " GFlop/sec" << std::endl;

    // Reduce
    // Size = size
    // Flop = 1 per element
    vec3.Reduce();

    _rocalution_sync();
    tick = rocalution_time();

    for(int i = 0; i < max_tests; ++i)
    {
        vec3.Reduce();
        _rocalution_sync();
    }

    _rocalution_sync();
    tack = rocalution_time();
    std::cout << "Reduce execution: " << (tack - tick) / max_tests / 1e3 << " msec"
              << "; " << max_tests * double(sizeof(double) * (size)) / (tack - tick) / 1e3
              << " Gbyte/sec; " << max_tests * double((size - 1)) / (tack - tick) / 1e3
              << " GFlop/sec" << std::endl;

    // Norm
    // Size = size
    // Flop = 2 per element
    vec3.Norm();

    _rocalution_sync();
    tick = rocalution_time();

    for(int i = 0; i < max_tests; ++i)
    {
        vec3.Norm();
        _rocalution_sync();
    }

    _rocalution_sync();
    tack = rocalution_time();
    std::cout << "Norm execution: " << (tack - tick) / max_tests / 1e3 << " msec"
              << "; " << max_tests * double(sizeof(double) * (size)) / (tack - tick) / 1e3
              << " Gbyte/sec; " << max_tests * double((2 * size)) / (tack - tick) / 1e3
              << " GFlop/sec" << std::endl;

    // Vector Update 1
    // Size = 3xsize
    // Flop = 2 per element
    vec3.ScaleAdd(double(5.5), vec2);

    _rocalution_sync();
    tick = rocalution_time();

    for(int i = 0; i < max_tests; ++i)
    {
        vec3.ScaleAdd(double(5.5), vec2);
        _rocalution_sync();
    }

    _rocalution_sync();
    tack = rocalution_time();
    std::cout << "Vector update (scaleadd) execution: " << (tack - tick) / max_tests / 1e3
              << " msec"
              << "; "
              << max_tests * double(sizeof(double) * (size + size + size)) / (tack - tick) / 1e3
              << " Gbyte/sec; " << max_tests * double((2 * size)) / (tack - tick) / 1e3
              << " GFlop/sec" << std::endl;

    // Vector Update 2
    // Size = 3*size
    // Flop = 2 per element
    vec3.AddScale(vec2, double(5.5));

    _rocalution_sync();
    tick = rocalution_time();

    for(int i = 0; i < max_tests; ++i)
    {
        vec3.AddScale(vec2, double(5.5));
        _rocalution_sync();
    }

    _rocalution_sync();
    tack = rocalution_time();
    std::cout << "Vector update (addscale) execution: " << (tack - tick) / max_tests / 1e3
              << " msec"
              << "; "
              << max_tests * double(sizeof(double) * (size + size + size)) / (tack - tick) / 1e3
              << " Gbyte/sec; " << max_tests * double((2 * size)) / (tack - tick) / 1e3
              << " GFlop/sec" << std::endl;

    mat.ConvertToCSR();
    nnz = mat.GetNnz();

    mat.Info();
    // Matrix-Vector Multiplication
    // Size = int(size+nnz) [row_offset + col] + valuetype(2*size+nnz) [in + out + nnz]
    // Flop = 2 per entry (nnz)
    mat.Apply(vec1, &vec2);

    _rocalution_sync();
    tick = rocalution_time();

    for(int i = 0; i < max_tests; ++i)
    {
        mat.Apply(vec1, &vec2);
        _rocalution_sync();
    }

    _rocalution_sync();
    tack = rocalution_time();
    std::cout << "CSR SpMV execution: " << (tack - tick) / max_tests / 1e3 << " msec"
              << "; "
              << max_tests
                     * double((sizeof(double) * (size + size + nnz) + sizeof(int) * (size + nnz)))
                     / (tack - tick) / 1e3
              << " Gbyte/sec; " << max_tests * double((2 * nnz)) / (tack - tick) / 1e3
              << " GFlop/sec" << std::endl;

    mat.ConvertToMCSR();
    nnz = mat.GetNnz();

    mat.Info();
    // Matrix-Vector Multiplication
    // Size = int(size+(nnz-size)) [row_offset + col] + valuetype(2*size+nnz) [in + out + nnz]
    // Flop = 2 per entry (nnz)
    mat.Apply(vec1, &vec2);

    _rocalution_sync();
    tick = rocalution_time();

    for(int i = 0; i < max_tests; ++i)
    {
        mat.Apply(vec1, &vec2);
        _rocalution_sync();
    }

    _rocalution_sync();
    tack = rocalution_time();
    std::cout << "MCSR SpMV execution: " << (tack - tick) / max_tests / 1e3 << " msec"
              << "; "
              << max_tests
                     * double(
                         (sizeof(double) * (size + size + nnz - size) + sizeof(int) * (size + nnz)))
                     / (tack - tick) / 1e3
              << " Gbyte/sec; " << max_tests * double((2 * nnz)) / (tack - tick) / 1e3
              << " GFlop/sec" << std::endl;

    mat.ConvertToELL();
    nnz = mat.GetNnz();

    mat.Info();
    // Matrix-Vector Multiplication
    // Size = int(nnz) [col] + valuetype(2*size+nnz) [in + out + nnz]
    // Flop = 2 per entry (nnz)
    mat.Apply(vec1, &vec2);

    _rocalution_sync();
    tick = rocalution_time();

    for(int i = 0; i < max_tests; ++i)
    {
        mat.Apply(vec1, &vec2);
        _rocalution_sync();
    }

    _rocalution_sync();
    tack = rocalution_time();
    std::cout << "ELL SpMV execution: " << (tack - tick) / max_tests / 1e3 << " msec"
              << "; "
              << max_tests * double((sizeof(double) * (size + size + nnz) + sizeof(int) * (nnz)))
                     / (tack - tick) / 1e3
              << " Gbyte/sec; " << max_tests * double((2 * nnz)) / (tack - tick) / 1e3
              << " GFlop/sec" << std::endl;

    mat.ConvertToCOO();
    nnz = mat.GetNnz();

    mat.Info();
    // Matrix-Vector Multiplication
    // Size = int(2*nnz) + valuetype(2*size+nnz)
    // Flop = 2 per entry (nnz)
    mat.Apply(vec1, &vec2);

    _rocalution_sync();
    tick = rocalution_time();

    for(int i = 0; i < max_tests; ++i)
    {
        mat.Apply(vec1, &vec2);
        _rocalution_sync();
    }

    _rocalution_sync();
    tack = rocalution_time();
    std::cout << "COO SpMV execution: " << (tack - tick) / max_tests / 1e3 << " msec"
              << "; "
              << max_tests
                     * double((sizeof(double) * (size + size + nnz) + sizeof(int) * (2 * nnz)))
                     / (tack - tick) / 1e3
              << " Gbyte/sec; " << max_tests * double((2 * nnz)) / (tack - tick) / 1e3
              << " GFlop/sec" << std::endl;

    mat.ConvertToHYB();
    nnz = mat.GetNnz();

    mat.Info();
    // Matrix-Vector Multiplication
    // Size = int(nnz) [col] + valuetype(2*size+nnz) [in + out + nnz]
    // Flop = 2 per entry (nnz)
    mat.Apply(vec1, &vec2);

    _rocalution_sync();
    tick = rocalution_time();

    for(int i = 0; i < max_tests; ++i)
    {
        mat.Apply(vec1, &vec2);
        _rocalution_sync();
    }

    _rocalution_sync();
    tack = rocalution_time();
    std::cout << "HYB SpMV execution: " << (tack - tick) / max_tests / 1e3 << " msec"
              << "; "
              // like O(ELL)
              << max_tests * double((sizeof(double) * (size + size + nnz) + sizeof(int) * (nnz)))
                     / (tack - tick) / 1e3
              << " Gbyte/sec; " << max_tests * double((2 * nnz)) / (tack - tick) / 1e3
              << " GFlop/sec" << std::endl;

    mat.ConvertToDIA();
    nnz = mat.GetNnz();

    mat.Info();
    // Matrix-Vector Multiplication
    // Size = int(size+nnz) + valuetype(2*size+nnz)
    // Flop = 2 per entry (nnz)
    mat.Apply(vec1, &vec2);

    _rocalution_sync();
    tick = rocalution_time();

    for(int i = 0; i < max_tests; ++i)
    {
        mat.Apply(vec1, &vec2);
        _rocalution_sync();
    }

    _rocalution_sync();
    tack = rocalution_time();
    std::cout << "DIA SpMV execution: " << (tack - tick) / max_tests / 1e3 << " msec"
              << "; "
              // assuming ndiag << size
              << max_tests * double((sizeof(double) * (nnz))) / (tack - tick) / 1e3
              << " Gbyte/sec; " << max_tests * double((2 * nnz)) / (tack - tick) / 1e3
              << " GFlop/sec" << std::endl;

    mat.ConvertToCSR();

    std::cout << "----------------------------------------------------" << std::endl;
    std::cout << "Combined micro benchmarks" << std::endl;

    double dot_tick = 0, dot_tack = 0;
    double norm_tick = 0, norm_tack = 0;
    double updatev1_tick = 0, updatev1_tack = 0;
    double updatev2_tick = 0, updatev2_tack = 0;
    double spmv_tick = 0, spmv_tack = 0;

    for(int i = 0; i < max_tests; ++i)
    {
        vec1.Ones();
        vec2.Zeros();
        vec3.Ones();
        mat.Apply(vec1, &vec2);

        // Dot product
        // Size = 2*size
        // Flop = 2 per element
        vec3.Dot(vec2);

        _rocalution_sync();
        dot_tick += rocalution_time();

        vec3.Dot(vec2);

        _rocalution_sync();
        dot_tack += rocalution_time();

        vec1.Ones();
        vec2.Zeros();
        vec3.Ones();
        mat.Apply(vec1, &vec2);

        // Norm
        // Size = size
        // Flop = 2 per element
        vec3.Norm();

        _rocalution_sync();
        norm_tick += rocalution_time();

        vec3.Norm();

        _rocalution_sync();
        norm_tack += rocalution_time();

        vec1.Ones();
        vec2.Zeros();
        vec3.Ones();
        mat.Apply(vec1, &vec2);

        // Vector Update 1
        // Size = 3xsize
        // Flop = 2 per element
        vec3.ScaleAdd(double(5.5), vec2);

        _rocalution_sync();
        updatev1_tick += rocalution_time();

        vec3.ScaleAdd(double(5.5), vec2);

        _rocalution_sync();
        updatev1_tack += rocalution_time();

        vec1.Ones();
        vec2.Zeros();
        vec3.Ones();
        mat.Apply(vec1, &vec2);

        // Vector Update 2
        // Size = 3*size
        // Flop = 2 per element
        vec3.AddScale(vec2, double(5.5));

        _rocalution_sync();
        updatev2_tick += rocalution_time();

        vec3.AddScale(vec2, double(5.5));

        _rocalution_sync();
        updatev2_tack += rocalution_time();

        vec1.Ones();
        vec2.Zeros();
        vec3.Ones();
        mat.Apply(vec1, &vec2);

        // Matrix-Vector Multiplication
        // Size = int(size+nnz) + valuetype(2*size+nnz)
        // Flop = 2 per entry (nnz)
        mat.Apply(vec1, &vec2);

        _rocalution_sync();
        spmv_tick += rocalution_time();

        mat.Apply(vec1, &vec2);

        _rocalution_sync();
        spmv_tack += rocalution_time();
    }

    std::cout << "Dot execution: " << (dot_tack - dot_tick) / max_tests / 1e3 << " msec"
              << "; "
              << max_tests * double(sizeof(double) * (size + size)) / (dot_tack - dot_tick) / 1e3
              << " Gbyte/sec; " << max_tests * double((2 * size)) / (dot_tack - dot_tick) / 1e3
              << " GFlop/sec" << std::endl;

    std::cout << "Norm execution: " << (norm_tack - norm_tick) / max_tests / 1e3 << " msec"
              << "; " << max_tests * double(sizeof(double) * (size)) / (norm_tack - norm_tick) / 1e3
              << " Gbyte/sec; " << max_tests * double((2 * size)) / (norm_tack - norm_tick) / 1e3
              << " GFlop/sec" << std::endl;

    std::cout << "Vector update (scaleadd) execution: "
              << (updatev1_tack - updatev1_tick) / max_tests / 1e3 << " msec"
              << "; "
              << max_tests * double(sizeof(double) * (size + size + size))
                     / (updatev1_tack - updatev1_tick) / 1e3
              << " Gbyte/sec; "
              << max_tests * double((2 * size)) / (updatev1_tack - updatev1_tick) / 1e3
              << " GFlop/sec" << std::endl;

    std::cout << "Vector update (addscale) execution: "
              << (updatev2_tack - updatev2_tick) / max_tests / 1e3 << " msec"
              << "; "
              << max_tests * double(sizeof(double) * (size + size + size))
                     / (updatev2_tack - updatev2_tick) / 1e3
              << " Gbyte/sec; "
              << max_tests * double((2 * size)) / (updatev2_tack - updatev2_tick) / 1e3
              << " GFlop/sec" << std::endl;

    std::cout << "SpMV execution: " << (spmv_tack - spmv_tick) / max_tests / 1e3 << " msec"
              << "; "
              << max_tests
                     * double((sizeof(double) * (size + size + nnz) + sizeof(int) * (size + nnz)))
                     / (spmv_tack - spmv_tick) / 1e3
              << " Gbyte/sec; " << max_tests * double((2 * nnz) / (spmv_tack - spmv_tick)) / 1e3
              << " GFlop/sec" << std::endl;

    stop_rocalution();

    return 0;
}