CUDA之矩阵乘法——复数

做好矩阵乘法和转置之后本来开心得不行的！
准备上手做个最基本的波束形成了！
突然发现希尔伯特变换完以后需要进行各种复数的运算…所以临时补写了一个复数乘法…
学着学着好像有点感觉了~！还是蛮有意思的。当然前提是能调试成功。
用一句傅小姐的名言鼓励一下“只要心甘情愿任何事情都会变得简单！”。

代码

__device__ float GetReal(const Matrix A, int row, int col) {

    return A.real[row * A.stride + col];
}

__device__ float GetImag(const Matrix A, int row, int col) {
    return A.imag[row * A.stride + col];
}

__device__ void SetElement(Matrix A, int row, int col, float valueR, float valueI) {
    A.real[row * A.stride + col] = valueR;
    A.imag[row * A.stride + col] = valueI;
}

__device__ Matrix GetSubMatrix(Matrix A, int row, int col) {
    Matrix Asub;
    Asub.width = BLOCK_SIZE;
    Asub.height = BLOCK_SIZE;
    Asub.stride = A.stride;
    Asub.real = &A.real[A.stride * BLOCK_SIZE * row+ BLOCK_SIZE * col];
    Asub.imag = &A.imag[A.stride * BLOCK_SIZE * row+ BLOCK_SIZE * col];
    return Asub;
}

__global__ void CMatMulKernel(Matrix A, Matrix B, Matrix C) {
    int blockRow = blockIdx.y;
    int blockCol = blockIdx.x;
    Matrix Csub = GetSubMatrix(C, blockRow, blockCol);
    float CvalueR = 0;
    float CvalueI = 0;
    int row = threadIdx.y;
    int col = threadIdx.x;
    for (int m = 0; m < (A.width / BLOCK_SIZE); ++m) {
        Matrix Asub = GetSubMatrix(A, blockRow, m);
        Matrix Bsub = GetSubMatrix(B, m, blockCol);
        __shared__ float AsR[BLOCK_SIZE][BLOCK_SIZE];
        __shared__ float AsI[BLOCK_SIZE][BLOCK_SIZE];
        __shared__ float BsR[BLOCK_SIZE][BLOCK_SIZE];
        __shared__ float BsI[BLOCK_SIZE][BLOCK_SIZE];
        AsR[row][col] = GetReal(Asub, row, col);
        AsI[row][col] = GetImag(Asub, row, col);
        BsR[row][col] = GetReal(Bsub, row, col);
        BsI[row][col] = GetImag(Bsub, row, col);
        __syncthreads();
        for (int e = 0; e < BLOCK_SIZE; ++e)
        {
            CvalueR += AsR[row][e] * BsR[e][col]-AsI[row][e]*BsI[e][col];
            CvalueI += AsR[row][e] * BsI[e][col]+AsI[row][e]*BsR[e][col];
        }
        __syncthreads();
    }
    SetElement(Csub, row, col, CvalueR,CvalueI);
}

void CMatMul(const Matrix A, const Matrix B, Matrix C) {
    Matrix d_A;
    d_A.width = d_A.stride = A.width; 
    d_A.height = A.height;
    size_t size = A.width * A.height * sizeof(float);
    cudaMalloc((void**)&d_A.real, size);
    cudaMalloc((void**)&d_A.imag, size);
    cudaMemcpy(d_A.real, A.real, size,
    cudaMemcpyHostToDevice);
    cudaMemcpy(d_A.imag, A.imag, size,
    cudaMemcpyHostToDevice);

    Matrix d_B;
    d_B.width = d_B.stride = B.width; 
    d_B.height = B.height;
    size = B.width * B.height * sizeof(float);
    cudaMalloc((void**)&d_B.real, size);
    cudaMalloc((void**)&d_B.imag, size);
    cudaMemcpy(d_B.real, B.real, size,
    cudaMemcpyHostToDevice);
    cudaMemcpy(d_B.imag, B.imag, size,
    cudaMemcpyHostToDevice);

    Matrix d_C;
    d_C.width = d_C.stride = C.width; 
    d_C.height = C.height;
    size = C.width * C.height * sizeof(float);
    cudaMalloc((void**)&d_C.real, size);
    cudaMalloc((void**)&d_C.imag, size);

    dim3 dimBlock(BLOCK_SIZE, BLOCK_SIZE);
    dim3 dimGrid(B.width / dimBlock.x, A.height / dimBlock.y);
    CMatMulKernel<<>>(d_A, d_B, d_C);
    cudaMemcpy(C.real, d_C.real, size,
    cudaMemcpyDeviceToHost);
    cudaMemcpy(C.imag, d_C.imag, size,
    cudaMemcpyDeviceToHost);

    cudaFree(d_A.real);
    cudaFree(d_A.imag);
    cudaFree(d_B.real);
    cudaFree(d_B.imag);
    cudaFree(d_C.real);
    cudaFree(d_C.imag);
}

为什么这么久过去了我依然不能上传图片。