oneapi

[Darius_Nicholas]

I have tried many ways to copy the data from a two-dimensional vector to buffer, but I have failed.How does buffer store a two-dimensional vector？And how do we solve the problem in the diagram?

[VaishnaviV_Intel]

Hi,

Thanks for posting on Intel communities.

Could you please provide us with the complete code so that we can reproduce your issue at our end?

We have also tried below sample code,

#include <iostream>
#include <CL/sycl.hpp>

int matmul() {

    constexpr size_t M = 3, N = 3;
    constexpr size_t X = 2, Y = 2;

    std::vector<std::vector<int>> a = {{1,1,1,1}, {1,1,1,1}, {1,1,1,1}};
    std::vector<std::vector<int>> b = {{1,1,1,1}, {1,1,1,1}, {1,1,1,1}};
    std::vector<std::vector<int>> c = {{0,0,0,0}, {0,0,0,0}, {0,0,0,0}};

    sycl::queue Q;
    sycl::buffer Ab{a};
    sycl::buffer Bb{b};
    sycl::buffer Cb{c};

    Q.submit([&](sycl::handler& h) {
        sycl::accessor Ad{Ab,h};
        sycl::accessor Bd{Bb,h};
        sycl::accessor Cd{Cb,h};

        sycl::range global{M, N};
        sycl::range local{X, Y};

        h.parallel_for(sycl::nd_range{global,local}, [=](sycl::nd_item<2> it) {
            int j = it.get_global_id(0);
            int i = it.get_global_id(1);
            for (int k = 0; k < N; ++k) {
                Cd[j][i] += Ad[j][k] * Bd[k][i];
            }
        });
    }).wait();

    sycl::host_accessor Ch{Cb};
    for(int i = 0 ; i < M ; i++) {
        for(int j = 0 ; j < N ; j++)
            std::cout << Ch[i][j] << " ";
        std::cout << std::endl;
    }

    return 0;
}

int main()
{
    matmul();
    return 0;
}

Output:

Thanks & Regards,

Vankudothu Vaishnavi.

[Darius_Nicholas]

This my code !

#include <vector>

#include <iostream>

#include <CL/sycl.hpp>  // DPC++ header

std::vector<std::vector<double>> filter2d(const std::vector<double>& b, double a, const std::vector<std::vector<double>>& x) {

    int rows = x.size();

    int cols = x[0].size();

    std::vector<std::vector<double>> y(x.size(), std::vector<double>(x[0].size()));

    sycl::queue q(sycl::gpu_selector{});

    std::vector<double> flat_y(rows * cols);

    std::vector<double> flat_x(rows * cols);

    for (int i = 0; i < rows; i++) {

        for (int j = 0; j < cols; j++) {

            flat_x[i * cols + j] = x[i][j];

        }

    }

    sycl::range<1> num_items{rows* cols};

    sycl::buffer b_buffer(b);

    sycl::buffer y_buffer(flat_y.data(),num_items);

    sycl::buffer x_buffer(flat_x);

    //sycl::buffer<double, 2> x_buffer_1(x.data(), sycl::range<2>(rows, cols));  Why is this wrong？

    //sycl::buffer<double, 2> x_buffer((double*)x.data(), num_items);

    q.submit([&](sycl::handler& h) {

        sycl::accessor x_acc(x_buffer,h,sycl::read_only);

        sycl::accessor b_acc(b_buffer,h,sycl::read_only);

        sycl::accessor y_acc(y_buffer, h, sycl::write_only,sycl::no_init);

        /*auto b_acc = b_buffer.get_access<sycl::access::mode::read>(h);

        auto x_acc = x_buffer.get_access<sycl::access::mode::read>(h);

        auto y_acc = y_buffer.get_access<sycl::access::mode::write>(h);*/

        h.parallel_for(num_items, [=](sycl::item<1> idx) {

            int i = idx / cols;

            int j = idx % cols;

            y_acc[idx] = 0.0;

            for (int k = 0; k < b.size(); k++) {

                if (i - k >= 0) {

                    y_acc[idx] += b_acc[k] * x_acc[(i - k) * cols + j];

                }

            }

            y_acc[idx] /= a;

        });

    });

    q.wait();

    for (int i = 0; i < rows; i++) {

        for (int j = 0; j < cols; j++) {

            y[i][j] = flat_y[i * cols + j];

        }

    }

    return y;

}

[Darius_Nicholas]

This is my code！

#include <vector>
#include <iostream>
#include <CL/sycl.hpp> // DPC++ header

std::vector<std::vector<double>> filter2d(const std::vector<double>& b, double a, const std::vector<std::vector<double>>& x) {
int rows = x.size();
int cols = x[0].size();
std::vector<std::vector<double>> y(x.size(), std::vector<double>(x[0].size()));

sycl::queue q(sycl::gpu_selector{});

std::vector<double> flat_y(rows * cols);
std::vector<double> flat_x(rows * cols);
for (int i = 0; i < rows; i++) {
for (int j = 0; j < cols; j++) {
flat_x[i * cols + j] = x[i][j];
}
}
sycl::range<1> num_items{rows* cols};

sycl::buffer b_buffer(b);
sycl::buffer y_buffer(flat_y.data(),num_items);
sycl::buffer x_buffer(flat_x);
//sycl::buffer<double, 2> x_buffer_1(x.data(), sycl::range<2>(rows, cols));
//sycl::buffer<double, 2> x_buffer((double*)x.data(), num_items);
//sycl::buffer<std::vector<double>, 2> x_buffer_2(x.data(), sycl::range<2>(rows, cols));

q.submit([&](sycl::handler& h) {
sycl::accessor x_acc(x_buffer,h,sycl::read_only);
sycl::accessor b_acc(b_buffer,h,sycl::read_only);

sycl::accessor y_acc(y_buffer, h, sycl::write_only,sycl::no_init);
/*auto b_acc = b_buffer.get_access<sycl::access::mode::read>(h);
auto x_acc = x_buffer.get_access<sycl::access::mode::read>(h);
auto y_acc = y_buffer.get_access<sycl::access::mode::write>(h);*/

h.parallel_for(num_items, [=](sycl::item<1> idx) {
int i = idx / cols;
int j = idx % cols;
y_acc[idx] = 0.0;
for (int k = 0; k < b.size(); k++) {
if (i - k >= 0) {
y_acc[idx] += b_acc[k] * x_acc[(i - k) * cols + j];
}
}
y_acc[idx] /= a;
});
});

q.wait();

for (int i = 0; i < rows; i++) {
for (int j = 0; j < cols; j++) {
y[i][j] = flat_y[i * cols + j];
}
}
return y;
}

[VaishnaviV_Intel · Accepted solution]

Hi,

Thanks for sharing your code with us.

Please try to declare a separate variable(i.e. z) for b.size() outside parallel_for and replace b.size() in parallel_for with that variable (i.e. z).

Please find the below modified code,

#include <vector>
#include <iostream>
#include <CL/sycl.hpp> // DPC++ header

std::vector<std::vector<double>> filter2d(const std::vector<double>& b, double a, const std::vector<std::vector<double>>& x) {
int rows = x.size();
int cols = x[0].size();
std::vector<std::vector<double>> y(x.size(), std::vector<double>(x[0].size()));
auto z=b.size();

sycl::queue q(sycl::cpu_selector_v);


std::vector<double> flat_y(rows * cols);
std::vector<double> flat_x(rows * cols);
for (int i = 0; i < rows; i++) {
for (int j = 0; j < cols; j++) {
flat_x[i * cols + j] = x[i][j];
}
}
sycl::range<1> num_items{static_cast<size_t>(rows * cols)};

sycl::buffer b_buffer(b);
sycl::buffer y_buffer(flat_y.data(),num_items);
sycl::buffer x_buffer(flat_x);
//sycl::buffer<double, 2> x_buffer_1(x.data(), sycl::range<2>(rows, cols));
//sycl::buffer<double, 2> x_buffer((double*)x.data(), num_items);
//sycl::buffer<std::vector<double>, 2> x_buffer_2(x.data(), sycl::range<2>(rows, cols));

q.submit([&](sycl::handler& h) {
sycl::accessor x_acc(x_buffer,h,sycl::read_only);
sycl::accessor b_acc(b_buffer,h,sycl::read_only);

sycl::accessor y_acc(y_buffer, h, sycl::write_only,sycl::no_init);
/*auto b_acc = b_buffer.get_access<sycl::access::mode::read>(h);
auto x_acc = x_buffer.get_access<sycl::access::mode::read>(h);
auto y_acc = y_buffer.get_access<sycl::access::mode::write>(h);*/

h.parallel_for(num_items, [=](sycl::item<1> idx) {
int i = idx / cols;
int j = idx % cols;
y_acc[idx] = 0.0;
for (int k = 0; k < z; k++) {
if (i - k >= 0) {
y_acc[idx] += b_acc[k] * x_acc[(i - k) * cols + j];
}
}
y_acc[idx] /= a;
});
});


q.wait();


for (int i = 0; i < rows; i++) {
for (int j = 0; j < cols; j++) {
y[i][j] = flat_y[i * cols + j];
}
}
return y;
}


int main() {
    // Example usage:
    std::vector<double> b = { 1.0, 2.0, 1.0 }; // Some filter coefficients
    double a = 1.0; // Some normalization factor
    std::vector<std::vector<double>> x = {
        { 1.0, 2.0, 3.0 },
        { 4.0, 5.0, 6.0 },
        { 7.0, 8.0, 9.0 }
    }; // Some 2D input data

    std::vector<std::vector<double>> y = filter2d(b, a, x);

    // Print the result
    for (const auto& row : y) {
        for (const auto& val : row) {
            std::cout << val << " ";
        }
        std::cout << "\n";
    }

    return 0;
}

Please let us know if you still face any issues.

Thanks & Regards,

Vankudothu Vaishnavi.

[Darius_Nicholas]

The code has run successfully! Thanks a lot!

[VaishnaviV_Intel]

Hi,

Thanks for accepting our solution. If you need any additional information, please post a new question as this thread will no longer be monitored by Intel. 

Thanks & Regards,

Vankudothu Vaishnavi