3.10。示例
原文: http://numba.pydata.org/numba-doc/latest/cuda/examples.html
3.10.1。矩阵乘法
这是使用 CUDA 内核的矩阵乘法的简单实现:
@cuda.jitdef matmul(A, B, C):"""Perform square matrix multiplication of C = A * B"""i, j = cuda.grid(2)if i < C.shape[0] and j < C.shape[1]:tmp = 0.for k in range(A.shape[1]):tmp += A[i, k] * B[k, j]C[i, j] = tmp
这种实现很简单直观但性能很差,因为相同的矩阵元素将从设备内存中多次加载,这很慢(某些设备可能有透明的数据缓存,但它们可能不够大,不能一次保存整个输入)。
如果我们使用阻塞算法来减少对设备内存的访问,则会更快。 CUDA 为块中的线程提供快速共享内存,以便在任务上协同计算。以下实现了使用共享内存的方形矩阵乘法的更快版本:
from numba import cuda, float32# Controls threads per block and shared memory usage.# The computation will be done on blocks of TPBxTPB elements.TPB = 16@cuda.jitdef fast_matmul(A, B, C):# Define an array in the shared memory# The size and type of the arrays must be known at compile timesA = cuda.shared.array(shape=(TPB, TPB), dtype=float32)sB = cuda.shared.array(shape=(TPB, TPB), dtype=float32)x, y = cuda.grid(2)tx = cuda.threadIdx.xty = cuda.threadIdx.ybpg = cuda.gridDim.x # blocks per gridif x >= C.shape[0] and y >= C.shape[1]:# Quit if (x, y) is outside of valid C boundaryreturn# Each thread computes one element in the result matrix.# The dot product is chunked into dot products of TPB-long vectors.tmp = 0.for i in range(bpg):# Preload data into shared memorysA[tx, ty] = A[x, ty + i * TPB]sB[tx, ty] = B[tx + i * TPB, y]# Wait until all threads finish preloadingcuda.syncthreads()# Computes partial product on the shared memoryfor j in range(TPB):tmp += sA[tx, j] * sB[j, ty]# Wait until all threads finish computingcuda.syncthreads()C[x, y] = tmp
由于共享内存是有限的资源,因此代码一次从输入数组预加载小块。然后,它调用 syncthreads() 等待所有线程完成预加载并在共享内存上进行计算之前。它在计算后再次同步,以确保所有线程在共享内存中完成数据,然后在下一次循环迭代中覆盖它。
