how to run a function on PyCUDA

You need to run a function written in python in the GPU using PyCUDA. When I work with arrays, I allocate memory for them on the video card and can continue to work with them.enter a description of the image here

How to do this only with a function, for example, what would be soenter a description of the image here

Author: MaxU, 2018-05-17
Source

1 answers

If the function performs simple operations, you can use the abstraction pycuda. gpuarray. gpuArray:

import pycuda.gpuarray as gpuarray
import pycuda.driver as cuda
import pycuda.autoinit
import numpy as np

a_gpu = gpuarray.to_gpu(np.random.randn(4,4).astype(np.float32))
a_doubled = (2*a_gpu).get()
print(a_doubled)
print(a_gpu)

More complex functions will need to be implemented with inserts in the language C, as in the example with the function doublify(...)

 1
Author: MaxU, 2018-05-17 15:11:42