PyCuda学习三之--共享内存与Thread的同步

共享内存与Thread的同步

• 给出3072*3072大小的数组, 每一个元素都是整数, 现在要做的就是, 将每个元素的立方相加, 并求出最终的结果. 首先,我们先用PyCuda基础知识写出来一个可以运行的程序.

  import time
  import numpy as np
  
  import pycuda.autoinit
  import pycuda.driver as cuda
  from pycuda.compiler import SourceModule
  
  
  mod = SourceModule("""
  
      __global__ void sumOfSquares(int* num, int *result, size_t N)
      {
          int index = threadIdx.x + blockIdx.x * blockDim.x;
          int stride = blockDim.x * gridDim.x;
          int sum = 0;
          for (int i = index; i < N; i += stride) {
              sum += num[i]*num[i]*num[i];
          }
          result[index] = sum;
      }
      
  """)
  
  
  def test(N, np_seed):
      np.random.seed(np_seed)
      a = np.random.randint(1, 10, N)
      N = np.int32(N)
      thread_size = 256
      # block_size = int((N + thread_size - 1) / thread_size)
      # power = len(str(block_size)) - 2
      # block_size = int(block_size / (10**power))
      block_size = 32
      b = np.empty(thread_size*block_size, dtype=np.int32)
      sumOfSquares = mod