Chapter 11

1. C/C++ programming language is used to write kernel function inside SourceModule class, and this kernel function is compiled by nvcc (Nvidia C ) Compiler.
2. The kernel call function is as follows:

myfirst_kernel(block=(512,512,1),grid=(1024,1014,1))

3. False. The order of block execution is random in PyCUDA program, and it can't be determined by PyCUDA programmer.

4. The directives from driver class remove the need of separate allocation of memory for the Array, uploading it to the device and downloading the result back to host. All operations are performed simultaneously during a kernel call. This makes the code simpler and easy to read.
5. The PyCUDA code for adding two to every element in an array is shown below:

import pycuda.gpuarray as gpuarray
import numpy
import pycuda.driver as drv

start = drv.Event()
end=drv.Event()
start.record()
start.synchronize()
n=10
h_b = numpy...