CUDA中本地内存比共享内存慢吗

CUDA中本地内存比共享内存慢吗？

cuda
#include 
__global__ void localMemoryAccess(float* array, int size) {
    int tid = threadIdx.x + blockIdx.x * blockDim.x;
    
    if (tid < size) {
        float value = array[tid];
        // 对本地内存进行读写操作
        array[tid] = value * 2;
    }
}
__global__ void sharedMemoryAccess(float* array, int size) {
    __shared__ float sharedArray[256];
    
    int tid = threadIdx.x + blockIdx.x * blockDim.x;
    
    if (tid < size) {
        float value = array[tid];
        // 对共享内存进行读写操作
        sharedArray[threadIdx.x] = value;
        __syncthreads();
        array[tid] = sharedArray[threadIdx.x] * 2;
    }
}
int main() {
    int size = 1024;
    float* array;
    float* d_array;
    
    array = (float*)malloc(size * sizeof(float));
    cudaMalloc((void**)&d_array, size * sizeof(float));
    
    for (int i = 0; i < size; i++) {
        array[i] = i;
    }
    
    cudaMemcpy(d_array, array, size * sizeof(float), cudaMemcpyHostToDevice);
    
    int blockSize = 256;
    int gridSize = (size + blockSize - 1) / blockSize;
    
    localMemoryAccess<<>>(d_array, size);
    cudaMemcpy(array, d_array, size * sizeof(float), cudaMemcpyDeviceToHost);
    
    for (int i = 0; i < size; i++) {
        printf("%f ", array[i]);
    }
    
    printf("\n");
    
    sharedMemoryAccess<<>>(d_array, size);
    cudaMemcpy(array, d_array, size * sizeof(float), cudaMemcpyDeviceToHost);
    
    for (int i = 0; i < size; i++) {
        printf("%f ", array[i]);
    }
    
    free(array);
    cudaFree(d_array);
    
    return 0;
}