<dependency>
    <groupId>org.jcuda</groupId>
    <artifactId>jcuda</artifactId>
    <version>10.2.0</version>
</dependency>
<dependency>
    <groupId>org.jcuda</groupId>
    <artifactId>jcublas</artifactId>
    <version>10.2.0</version>
</dependency>

import jcuda.*;
import jcuda.runtime.*;
import jcuda.jcublas.*;

public class MatrixMultiplication {
    public static void main(String[] args) {
        // 初始化JCuda
        JCuda.setExceptionsEnabled(true);

        // 定義矩陣的大小
        int N = 2; // 矩陣的行或列數(shù)
        float[] hostA = {1, 2, 3, 4}; // 矩陣A
        float[] hostB = {5, 6, 7, 8}; // 矩陣B
        float[] hostC = new float[N * N]; // 結果矩陣C

        // 在設備上分配內存
        Pointer d_A = new Pointer();
        Pointer d_B = new Pointer();
        Pointer d_C = new Pointer();
        JCuda.cudaMalloc(d_A, hostA.length * Sizeof.FLOAT);
        JCuda.cudaMalloc(d_B, hostB.length * Sizeof.FLOAT);
        JCuda.cudaMalloc(d_C, hostC.length * Sizeof.FLOAT);

        // 將主機數(shù)據(jù)復制到設備
        JCuda.cudaMemcpy(d_A, Pointer.to(hostA), hostA.length * Sizeof.FLOAT, cudaMemcpyHostToDevice);
        JCuda.cudaMemcpy(d_B, Pointer.to(hostB), hostB.length * Sizeof.FLOAT, cudaMemcpyHostToDevice);

        // 使用cublasSgemm進行矩陣乘法
        float alpha = 1; // 矩陣乘法的因子
        float beta = 0;  // 初始化結果矩陣的因子
        JCublas.cublasSgemm('N', 'N', N, N, N, alpha, d_A, N, d_B, N, beta, d_C, N);

        // 將結果從設備復制回主機
        JCuda.cudaMemcpy(Pointer.to(hostC), d_C, hostC.length * Sizeof.FLOAT, cudaMemcpyDeviceToHost);

        // 輸出結果
        System.out.println("Result C:");
        for (int i = 0; i < N; i++) {
            for (int j = 0; j < N; j++) {
                System.out.print(hostC[i * N + j] + " ");
            }
            System.out.println();
        }

        // 釋放設備的內存
        JCuda.cudaFree(d_A);
        JCuda.cudaFree(d_B);
        JCuda.cudaFree(d_C);
    }
}

javac -cp ".:jcublas.jar" MatrixMultiplication.java
java -cp ".:jcublas.jar" MatrixMultiplication